rakina@chromium.org chrome-loading@google.com Metrics recorded each time a SRP navigation/page load is abandoned (e.g. the navigation got cancelled, or the user backgrounded the app while the page load is not finished yet, etc. See AbandonedPageLoadMetricsObserver for more details). The reason the navigation/page load gets abandoned. The time when the navigation/page load is abandoned, relative to the last milestone. The time when the navigation/page load is abandoned, relative to navigation start. The timestamp of the AFT End event for the page load, relative to navigation start. The timestamp of the AFT Start event for the page load, relative to navigation start. The timestamp of the body chunk end event for the page load, relative to navigation start. The timestamp of the body chunk start event for the page load, relative to navigation start. The time when the CommitNavigation IPC for the navigation is received in the renderer process, relative to navigation start. The time when the CommitNavigation IPC for the navigation is sent from the browser process, relative to navigation start. The time when the DidCommit IPC for the navigation is received, relative to navigation start. Whether the navigation involves non-SRP URL (which later redirected to SRP). The timestamp of when the "DOMContentLoaded" event is fired for the page load, relative to navigation start. The timestamp of the first contentful paint for the page load, relative to navigation start. The time when the first network request that got redirected for the navigation started, relative to navigation start. Whether the first redirect response is received yet. The timestamp of the header chunk end event for the page load, relative to navigation start. Note that this event is randomly sampled based on the JavaScript in the page, recorded at the frequency of 1%. The timestamp of the header chunk start event for the page load, relative to navigation start. The timestamp of the largest contentful paint for the page load, relative to navigation start. The last milestone before the navigation is abandoned. The time when the URLLoader for the navigation started, relative to navigation start. The timestamp of when the "load" event is fired for the page load, relative to navigation start. The time when the final network request that didn't got redirected for the navigation started, relative to navigation start. Whether the non-redirect/final response is received yet. The time when the parsing of the HTTP response body started, relative to navigation start. If we've experienced backgrounding while the navigation/page load is running before this abandonment event happened, this will have the time when the backgrounding happened, relative to navigation start. If we've experienced hiding while the navigation/page load is running before this abandonment event happened, this will have the time when the hiding happened, relative to navigation start. The time when the final renderer process for the navigation is initialized, relative to navigation start. The network connection RTT when the abandonment happens, in milliseconds. This is useful to see if the navigation happens on a slow network. olesiamarukhnno@google.com Recorded on each page info bubble open when "About this site" data is fetched. The status of the "About this site" info, including if the data is available and whether the proto is valid. See |page_info::about_this_site_validation::AboutThisSiteStatus| for enum elements. yaoxia@chromium.org An enum that specifies the dismissal reason for JavaScript popup dialog that can be triggered by window.alert(), window.confirm() or window.prompt(). See JavaScriptDialogTabHelper::DismissalCause for the enum elements. csharrison@chromium.org True if the page attempted a tab-under navigation. yaoxia@chromium.org Recorded whenever window.open() is called when AdTagging is enabled. True if the page called window.open() with an ad script in the stack. True if the page called window.open() from an ad subframe. rhalavati@google.com Recorded once for any visible canvas, or offscreen canvas which pushes its rendered output directly into a placeholder element in the DOM. Records whether the canvas listens to keyboard events or not. Records whether the canvas listens to mouse events or not. Records whether the canvas drew any text or not. Records whether the canvas is off-screen or not. Records whether the canvas is heuristically estimated to need accessibility support. Note that this value does not exactly represent the heuristic result when accessibility is not enabled because some of the properties are not computed by the accessibility layer. Type of the rendering contexts created for the canvas. dtseng@chromium.org nektar@chromium.org Tracks automatic image descriptions provided to blind users. Whether the image description included a description. Whether the image already had label text before it was augmented with an additional description. Whether the image description included OCR text. aldietz@google.com mschillaci@google.com chrome-a11y-core@google.com Tracks usage of Page Zoom on Chrome for Android. Recorded each time the slider is dismissed if the user changed the zoom level on this page before dismissal. For users who changed the individual zoom level on this page, tracks what zoom level has been set on the page at slider dismissal. There may be multiple instances where the slider is opened on the same page, so there may be multiple values logged per page. The value will be an integer between 25 and 500, inclusive, bucketed in intervals of 5. abigailbklein@google.com Records the time taken to distill a web page. The amount of time (ms) required to distill a web page with the merged algorithm if distillation failed. The amount of time (ms) required to distill a web page with the merged algorithm if distillation succeeded. The amount of time (ms) required to distill a web page with the rules based algorithm if distillation failed. The amount of time (ms) required to distill a web page with the rules based algorithm if distillation succeeded. lwinston@google.com kristislee@google.com Records a distillation success / failure after a new page has been opened with reading mode. The status of the distillation alfredohiguera@google.com Records content quality metrics for a page, including text similarity (RougeL), structural integrity, and formatting preservation. This event is recorded when the distillation of a page is completed. Evaluates the preservation of formatting elements (bold, italic, blockquotes, and code) in the distilled HTML compared to the accessibility tree. Value is a percentage (0-100). The ratio of link density in the distilled HTML compared to the link density in the original accessibility tree. Value is a percentage (0-100). The ROUGE-L F1 score of the distilled text compared to the original text extracted from the AXTree, measuring sequence overlap of words. Value is a percentage (0-100). The ROUGE-L F2 score of the distilled text compared to the original text extracted from the AXTree. Value is a percentage (0-100). The ROUGE-L precision score of the distilled text compared to the original text extracted from the AXTree. Value is a percentage (0-100). The ROUGE-L recall score of the distilled text compared to the original text extracted from the AXTree. Value is a percentage (0-100). Evaluates the preservation of structural elements (headings and lists) in the distilled HTML compared to the accessibility tree. Value is a percentage (0-100). abigailbklein@google.com Recorded when an empty state is shown in Read Anything. The empty state is shown whenever distillation fails. When this happens, the user can "select to distill", meaning they can manually select text in the main web page they would like to see in reading mode. Total number of selections made. lwinston@google.com kristislee@google.com komo-eng@google.com Recorded when the user interacts with or is shown the Read Anything omnibox entry point. Whether the omnibox chip is shown. Whether the chip was triggered (via mouse or keyboard) to open Reading mode. lwinston@google.com kristislee@google.com komo-eng@google.com Recorded when distillation completes after the Read Anything omnibox entry point was used to open reading mode. The distillation status in reading mode after the omnibox entry point was used to open reading mode. gusmartin@google.com lwinston@google.com komo-eng@google.com Records Read Anything side panel events A flag set to true when the Read Anything side panel is closed for a given source ID. Records the UI entry point that the user clicked to open the Read Anything side panel. A flag set to true when the Read Anything side panel is shown for a given source ID. lucasradaelli@google.com nektar@chromium.org Tracks the amount of time sending accessibility events from web renderers. The amount of time spent in RenderAccessibilityImpl.SendPendingAccessibilityEvents abigailbklein@google.com rhalavati@google.com kyungjunlee@google.com Records the time taken to distill a web page. The amount of time (ms) required to distill a web page if distillation failed. The amount of time (ms) required to distill a web page if distillation succeeded. salmaabdelaziz@google.com dembski@google.com Metrics recorded when the Actor executes the AttemptOtpFilling tool on a webpage to attempt to fill an OTP. Records the OTP source type (SMS, Email, Authenticator App, Unknown) predicted by the Actor for the current main frame URL. npm@google.com web-identity-eng@google.com Metrics recorded for an actor login request. Recorded once per request when it completes or reaches a significant milestone. Recorded when the account used to login is automatically selected due to a previous user choice to always allow logging in with that account. An enum of the result from a query to get credentials. Recorded when there was some deduplication performed in order to show the accounts. Recorded at most once per actor request to show accounts to sign in. Recorded when the FedCM continuation API is invoked during an actor autologin flow. Recorded when there is a hanging FedCM request during an actor login request. An enum with the final result from a federated actor login request. Enum recorded when there is an identity provider requested by the embedder. Returns the time (ms) from when a "get credentials" request completes until an account is chosen. Returns the time (ms) from actor login request to when the "get credentials" request is completed. The number of accounts shown to the user during an actor login request. An enum of the type of account chosen by the user to login during an actor login request. dylancutler@google.com chrome-ai-security@google.com Metrics related to actor Origin Gating, a security feature controlling which origins actors may interact with. The ExecutionEngine::State when this metric is emitted. We want to see where in the actor tool use lifecycle that these navigation confirmations occur most frequently. The result of confirming with the server if the actor may interact with the origin of a URL that Chrome detects the actor navigating to. If the result is kNotRequired then the origin was already previously evaluated. johnidel@chromium.org jkarlin@chromium.org Records metrics associated with a single ad frame for a page load. An ad frame includes itself and all of its child frames. Only recorded for frames with non-zero bytes. Recorded when a page is destroyed or when the app is backgrounded on mobile. The peak percentage of CPU utilization attributed to the frame in any 30 second window prior to user activation. Frames that exist for less than 30 seconds still use 30 seconds as the denominator for the calculation. Wall time of tasks attributed to the frame after the frame received user activation. Measured in millseconds. Total wall time of tasks attributed to the frame. Measured in milliseconds. The depth of this frame's frame tree. There is no set limit on a frame's maximum depth. Bytes loaded from the cache for all resources loaded within the frame. This includes bytes loaded from both the HTTP cache and memory cache. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Network bytes loaded for resources with an image mime type within the frame. Includes header bytes. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Network bytes loaded for resources with a javascript mime type within the frame. Includes header bytes. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Network bytes loaded for all resources within the frame. Includes header bytes. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Total number of resources loaded by the frame, including incomplete resources. Network bytes loaded for resources with a video/ mime type within the frame. Includes header bytes. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). An enum representing whether the frame was Cross-Origin or Same-Origin to its parent frame. See FrameData::OriginStatus. An enum representing whether media was played in the frame. See FrameData::MediaStatus. An enum representing whether the frame received user activation. See FrameData::UserActivationStatus. Number of milliseconds before FirstContentfulpaint was reached in the frame. This is relative to subframe navigation start. Number of millseconds before Time-To-Interactive was reached in the frame. This is relative to subframe navigation start. Last known height of the iframe in pixels. Not reported accurately when "Visibility.Hidden" is true. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.15). Last known width of the iframe in pixels. Not reported accurately when "Visibility.Hidden" is true. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.15). A 0 or 1 values representing whether the frame had a display=none styling set. Records the last known value. johnidel@chromium.org jkarlin@chromium.org Recorded when a page has loaded non-zero ad bytes, and the page is being destroyed/navigated. Amount of network bytes used to load ad resource on the page. Includes resources that did not finish, and top-level ads. Measured in kilobytes. Amount of CPU wall time inside of ad iframes over the duration of a page load, while the page is in the foreground. Measured in milliseconds. Amount of network bytes used to load ad resources with a supported javascript mime type on the page. Measured in kilobytes. Amount of network bytes used to load ad resources with a video mime type on the page. Measured in kilobytes. Amount of network bytes used to load ad resources in the main frame. Includes resources that did not finish but does not include resources in subframes. Measured in bytes. This is rounded to the nearest exponential bucket (with a bucket ratio of 1.3). The estimated maximum density of ads on a page by area. Calculated as the area of ads on the page * 100 / page area. This counts each overlapping area once, it may be inaccurate due to updates and calculations being throttled. The estimated maximum density of ads on a page by height. Calculated as the combined height of ads on the page * 100 / page height. This counts each overlapping area once, it may be inaccurate due to updates and calculations being throttled. Amount of network bytes used to load resources on the page. Includes resources that did not finish. Measured in kilobytes. yaoxia@chromium.org johnidel@chromium.org jkarlin@chromium.org Recorded when the page is being destroyed/navigated away or when the app enters the background on mobile. This is separate from the `AdPageLoad` event to allow it to be recorded on all pages (i.e. not only pages that have loaded non-zero ad byte), and to allow a potentially different sampling. Both iframe and img ads are covered. The estimated average viewport ad density. Each density value is calculated as the area of (iframe and img) ads within the viewport * 100 / viewport area, where each overlapping area is counted once. And this returns the average of the densities accumulated over the page load time rounded to the nearest integer. The estimated kurtosis of the viewport ad density. Each density value is calculated as the area of (iframe and img) ads within the viewport * 100 / viewport area, where each overlapping area is counted once. And this returns the kurtosis of the densities accumulated over the page load time rounded to the nearest exponential bucket integer with bucket spacing 1.3. For negative values, the bucketing will be taken on its magnitude part. The estimated skewness of the viewport ad density. Each density value is calculated as the area of (iframe and img) ads within the viewport * 100 / viewport area, where each overlapping area is counted once. And this returns the skewness of the densities accumulated over the page load time rounded to the nearest exponential bucket integer with bucket spacing 1.3. For negative values, the bucketing will be taken on its magnitude part. The estimated variance of the viewport ad density. Each density value is calculated as the area of (iframe and img) ads within the viewport * 100 / viewport area, where each overlapping area is counted once. And this returns the variance of the densities accumulated over the page load time rounded to the nearest exponential bucket integer with bucket spacing 1.3. For negative values, the bucketing will be taken on its magnitude part. yaoxia@chromium.org johnidel@chromium.org jkarlin@chromium.org Recorded when the page is being destroyed/navigated away or when the app enters the background on mobile. This is separate from the `AdPageLoad` event to allow it to be recorded on all pages (i.e. not only pages that have loaded non-zero ad byte), and to allow a potentially different sampling. For all metrics in this event: 1) all detected ad types (e.g., iframes, imgs) are included, and 2) ads obscured by non-ad content are excluded. The estimated average viewport ad count accumulated over the page load time, multiplied by 100 and rounded to the nearest exponential bucket integer with bucket spacing 1.15. This allows for 2 decimal places of precision. The estimated average viewport ad density. Each density value is calculated as the area of (iframe and img) ads within the viewport * 100 / viewport area, where each overlapping area is counted once. And this returns the average of the densities accumulated over the page load time rounded to the nearest integer. The estimated kurtosis of the viewport ad density. Each density value is calculated as the area of (iframe and img) ads within the viewport * 100 / viewport area, where each overlapping area is counted once. And this returns the kurtosis of the densities accumulated over the page load time rounded to the nearest exponential bucket integer with bucket spacing 1.3. For negative values, the bucketing will be taken on its magnitude part. The estimated skewness of the viewport ad density. Each density value is calculated as the area of (iframe and img) ads within the viewport * 100 / viewport area, where each overlapping area is counted once. And this returns the skewness of the densities accumulated over the page load time rounded to the nearest exponential bucket integer with bucket spacing 1.3. For negative values, the bucketing will be taken on its magnitude part. The estimated variance of the viewport ad density. Each density value is calculated as the area of (iframe and img) ads within the viewport * 100 / viewport area, where each overlapping area is counted once. And this returns the variance of the densities accumulated over the page load time rounded to the nearest exponential bucket integer with bucket spacing 1.3. For negative values, the bucketing will be taken on its magnitude part. morlovich@chromium.org chrome-ads-core@google.com Captures latency-relevant metrics associated with a single FLEDGE auction. Many of these metrics are latency measurements, aggregated (mean/max) for parts of the auction flow that happen multiple times in an auction, e.g. calls to GenerateBid. Others are potentially latency-impacting measures of the auction, e.g. the number of InterestGroups, buyers, component auctions, bids that were scored either together or separately for the enforced and non-enforced KAnon modes. Finally, some of these metrics look at points in the auction flow where multiple inputs come together to assess which of these inputs most frequently appears on the latency critical path. The time in milliseconds between the AuctionRunner constructor (the start of the auction) and the last call to BidderWorklet::DeliverBidCallbackOnUserThread (the end of the BidGenerationPhase). The BidGenerationPhase represents the time during which ad-tech provided generateBid JavaScript functions are being executed. This is one of a series of metrics representing the duration of each phase of the auction, where the phases may overlap or have space between them. This is bucketed in linear buckets of 10 ms. The time in milliseconds between the AuctionRunner constructor (the start of the auction) and the first call to BidderWorklet::GenerateBidIfReady (specifically where that call corresponds to a bidder worklet for which all of its GenerateBid dependencies have been satisfied and it's about to call the ad-tech provided generateBid JavaScript function; the start of the BidGenerationPhase). The BidGenerationPhase represents the time during which ad-tech provided generateBid JavaScript functions are being executed. This is one of a series of metrics representing the duration of each phase of the auction, where the phases may overlap or have space between them. This is bucketed in linear buckets of 10 ms. The time in milliseconds between the AuctionRunner constructor (the start of the auction) and the last call to BidderWorklet::GenerateBidIfReady (the end of the BidSignalsFetchPhase). The BidSignalsFetchPhase represents the time during which any bidder worklet is waiting for one or more of the GenerateBid dependencies, notably bidding signals, in anticipation of calling an ad-tech provided generateBid JavaScript function. This is one of a series of metrics representing the duration of each phase of the auction, where the phases may overlap or have space between them. This is bucketed in linear buckets of 10 ms. The time in milliseconds between the AuctionRunner constructor (the start of the auction) and the first call to BidderWorklet::BeginGenerateBid (the start of the BidSignalsFetchPhase). The BidSignalsFetchPhase represents the time during which any bidder worklet is waiting for one or more of the GenerateBid dependencies, notably bidding signals, in anticipation of calling an ad-tech provided generateBid JavaScript function. This is one of a series of metrics representing the duration of each phase of the auction, where the phases may overlap or have space between them. This is bucketed in linear buckets of 10 ms. Time in ms between AdAuctionServiceImpl::RunAdAuction and AdAuctionServiceImpl::OnAuctionComplete. Roughly equivalent to UMA metric, Ads.InterestGroup.Auction.AuctionWithWinnerTime. Indicates the KAnonymityBidMode used for this auction. The time in milliseconds between the AuctionRunner constructor (the start of the auction) and the last call to AuctionRunner::OnLoadInterestGroupsComplete (the end of the LoadInterestGroupPhase). This is one of a series of metrics representing the duration of each phase of the auction, where the phases may overlap or have space between them. Unlike other phase metrics, this phase has no associated StartTime metric, since this phase begins at the start of the auction. This is bucketed in linear buckets of 10 ms. This metric is identical to LoadInterestGroupPhaseLatencyInMillis except that this metric uses linear bucketing. The latency in milliseconds between the start of the auction (the AuctionRunner constructor), and the call to AuctionRunner::OnLoadInterestGroupsComplete. This is related to UMA metric, Ads.InterestGroup.Auction.LoadGroupsTime, except that, in a multi-seller auction, the UMA metric is recorded once per component auction and then again for the top-level auction, whereas this metric is recorded only once, for the top-level auction. This metric also isn't limited to those cases where there were InterestGroups to load, since we can independently filter this based on the value of NumInterestGroups. Max latency of the time spent decoding each additional bid. This only includes additional bids that were successfully decoded, so those negative targeted or sent for scoring, and not those that encountered any error while being decoded, e.g. invalid base64, JSON parsing errors, etc. Aggregated across all components for a multi-seller auctions, this records the max latency in millis between BuyerHelper::StartGeneratingBids and BuyerHelper::OnGenerateBidComplete, counting each BidState once. As such, this includes starting the BidderWorklet in that latency. This latency window is intentionally terminated in OnGenerateBidComplete, and not in OnGenerateBidCompleteInternal, so that this *excludes* bids filtered during reprioritization, as well as those that failed because of some fatal error, except that we explicitly *include* those that timed out. Only for multi-seller auctions, this is the highest latency of the bidding and scoring phase, observed across all of the component auctions. This latency is measured as the time between the call to StartBiddingAndScoringPhase and the call to OnComponentAuctionComplete. This metric only has a value for entries that represent multi-seller auctions, as indicated by a value for the NumSellersWithBidders metric of two or greater. Max latency between the start of the BiddingAndScoring phase and when config promises are resolved. In a single-seller auction, this is only recorded if the auction has config promises; in a multi-seller auction, this is only recorded for component auctions and the top-level auction for those that have config promises. In auctions for which the config promises are resolved before the bidding and auction phase, a critical path latency of zero is recorded. This shows how long additional bids would have had to wait for the config promises to be resolved, since they couldn't be processed until the BiddingAndScoring phase anyway. Max latency between the auction start time and when config promises are resolved. In a single-seller auction, this is only recorded if the auction has config promises; in a multi-seller auction, this is only recorded for component auctions and the top-level auction for those that have config promises. This is not the same as how long the auction needed to wait for a config promise, since other work happened simultaneously. That's better captured by the GenerateBid critical path metrics and the ConfigPromisesResolved critical path metrics. Aggregated across all components for multi-seller auctions, this records the max time in millis that GenerateBid waited for the Script and/or WASM load. If both Script and WASM were requested, this is the latency of the slower of the two to complete. Aggregated across all components for multi-seller auctions, this records the max time in millis that GenerateBid waited for the fulfillment of the config promises. Aggregated across all components for multi-seller auctions, this records the max time in millis that GenerateBid waited for the download of the DirectFromSellerSignals. This combines the PerBuyerSignals and AuctionSignals, and as such, this is the latency of the slower of the two to complete. Aggregated across all components for multi-seller auctions, this records the max time in millis that GenerateBid waited for the download of the TrustedBiddingSignals. This includes *all* TrustedBiddingSignals for that buyer, because of how those all need to be available in order to reprioritize InterestGroups and apply per-buyer group limits. Aggregated across all components for a multi-seller auctions, this records the max latency in millis of calls to GenerateSingleBid. This is related to UMA metric, Ads.InterestGroup.Auction.GenerateBidTime, except that it's aggregated across all GenerateSingleBid calls for this auction. Like the UMA metric, this measures each distinct call to GenerateSingleBid, for which there could be two for a given InterestGroup, one that satisfies the kAnon threshold, and one that doesn't. Aggregated across all components for multi-seller auctions, this records the max time in millis that ScoreAd waited for the Script load. Aggregated across all components for multi-seller auctions, this records the max time in millis that ScoreAd waited for the download of the DirectFromSellerSignals. This combines the SellerSignals and AuctionSignals, and as such, this is the latency of the slower of the two to complete. Aggregated across all components for a multi-seller auctions, this records, across all of the ads scored, the max latency in millis between InterestGroupAuction::ScoreBidIfReady (the second call to it, which in turn calls SellerWorklet::ScoreAd) and InterestGroupAuction::OnScoreAdComplete. Aggregated across all components for a multi-seller auctions, this records the max latency in millis of calls to V8State::ScoreAd. This is related to UMA metric, Ads.InterestGroup.Auction.ScoreAdTime, except that it's aggregated across all GenerateSingleBid calls for this auction. Aggregated across all components for multi-seller auctions, this records the max time in millis that ScoreAd waited for the download of the TrustedScoringSignals. This includes *all* TrustedScoringSignals for that buyer, because of how those all need to be available in order to reprioritize InterestGroups and apply per-buyer group limits. Mean latency of the time spent decoding each additional bid. This only includes additional bids that were successfully decoded, so those negative targeted or sent for scoring, and not those that encountered any error while being decoded, e.g. invalid base64, JSON parsing errors, etc. Aggregated across all components for a multi-seller auctions, this records the mean latency in millis between BuyerHelper::StartGeneratingBids and BuyerHelper::OnGenerateBidComplete, counting each BidState once. As such, this includes starting the BidderWorklet in that latency. This latency window is intentionally terminated in OnGenerateBidComplete, and not in OnGenerateBidCompleteInternal, so that this *excludes* bids filtered during reprioritization, as well as those that failed because of some fatal error, except that we explicitly *include* those that timed out. Only for multi-seller auctions, this is the mean latency of the bidding and scoring phase, aggregated across each of the component auctions. This latency is measured as the time between the call to StartBiddingAndScoringPhase and the call to OnComponentAuctionComplete. This metric only has a value for entries that represent multi-seller auctions, as indicated by a value for the NumSellersWithBidders metric of two or greater. Mean latency between the start of the BiddingAndScoring phase and when config promises are resolved. In a single-seller auction, this is only recorded if the auction has config promises; in a multi-seller auction, this is only recorded for component auctions and the top-level auction for those that have config promises. In auctions for which the config promises are resolved before the bidding and auction phase, a critical path latency of zero is recorded. This shows how long additional bids would have had to wait for the config promises to be resolved, since they couldn't be processed until the BiddingAndScoring phase anyway. Mean latency between the auction start time and when config promises are resolved. In a single-seller auction, this is only recorded if the auction has config promises; in a multi-seller auction, this is only recorded for component auctions and the top-level auction for those that have config promises. This is not the same as how long the auction needed to wait for a config promise, since other work happened simultaneously. That's better captured by the GenerateBid critical path metrics and the ConfigPromisesResolved critical path metrics. Aggregated across all components for multi-seller auctions, this records the mean critical path latency in millis, the time after all other GenerateBid dependencies were satisfied that GenerateBid was blocked waiting for the Script and/or WASM load. This combines Script and WASM loads. Aggregated across all components for multi-seller auctions, this records the mean time in millis that GenerateBid waited for the Script and/or WASM load. If both Script and WASM were requested, this is the latency of the slower of the two to complete. Aggregated across all components for multi-seller auctions, this records the mean critical path latency in millis, the time after all other GenerateBid dependencies were satisfied that GenerateBid was blocked waiting for the fulfillment of the config promises. Aggregated across all components for multi-seller auctions, this records the mean time in millis that GenerateBid waited for the fulfillment of the config promises. Aggregated across all components for multi-seller auctions, this records the mean critical path latency in millis, the time after all other GenerateBid dependencies were satisfied that GenerateBid was blocked waiting for the download of the DirectFromSellerSignals. This combines the PerBuyerSignals and AuctionSignals since they should arrive at roughly the same time. Aggregated across all components for multi-seller auctions, this records the mean time in millis that GenerateBid waited for the download of the DirectFromSellerSignals. This combines the PerBuyerSignals and AuctionSignals, and as such, this is the latency of the slower of the two to complete. Aggregated across all components for multi-seller auctions, this records the mean critical path latency in millis, the time after all other GenerateBid dependencies were satisfied that GenerateBid was blocked waiting for the download of the TrustedBiddingSignals. This includes *all* TrustedBiddingSignals for that buyer, because of how those all need to be available in order to reprioritize InterestGroups and apply per-buyer group limits. Aggregated across all components for multi-seller auctions, this records the mean time in millis that GenerateBid waited for the download of the TrustedBiddingSignals. This includes *all* TrustedBiddingSignals for that buyer, because of how those all need to be available in order to reprioritize InterestGroups and apply per-buyer group limits. Aggregated across all components for a multi-seller auctions, this records the mean latency in millis of calls to GenerateSingleBid. This is related to UMA metric, Ads.InterestGroup.Auction.GenerateBidTime, except that it's aggregated across all GenerateSingleBid calls for this auction. Like the UMA metric, this measures each distinct call to GenerateSingleBid, for which there could be two for a given InterestGroup, one that satisfies the kAnon threshold, and one that doesn't. Aggregated across all components for multi-seller auctions, this records the mean critical path latency in millis, the time after all other ScoreAd dependencies were satisfied that ScoreAd was blocked waiting for the Script load. Aggregated across all components for multi-seller auctions, this records the mean time in millis that ScoreAd waited for the Script load. Aggregated across all components for multi-seller auctions, this records the mean critical path latency in millis, the time after all other ScoreAd dependencies were satisfied that ScoreAd was blocked waiting for the download of the DirectFromSellerSignals. This combines the SellerSignals and AuctionSignals since they should arrive at roughly the same time. Aggregated across all components for multi-seller auctions, this records the mean time in millis that ScoreAd waited for the download of the DirectFromSellerSignals. This combines the SellerSignals and AuctionSignals, and as such, this is the latency of the slower of the two to complete. Aggregated across all components for a multi-seller auctions, this records, across all of the ads scored, the mean latency in millis between InterestGroupAuction::ScoreBidIfReady (the second call to it, which in turn calls SellerWorklet::ScoreAd) and InterestGroupAuction::OnScoreAdComplete. Aggregated across all components for a multi-seller auctions, this records the mean latency in millis of calls to V8State::ScoreAd. This is related to UMA metric, Ads.InterestGroup.Auction.ScoreAdTime, except that it's aggregated across all GenerateSingleBid calls for this auction. Aggregated across all components for multi-seller auctions, this records the mean critical path latency in millis, the time after all other ScoreAd dependencies were satisfied that ScoreAd was blocked waiting for the download of the TrustedScoringSignals. This includes *all* TrustedScoringSignals for that buyer, because of how those all need to be available in order to reprioritize InterestGroups and apply per-buyer group limits. Aggregated across all components for multi-seller auctions, this records the mean time in millis that ScoreAd waited for the download of the TrustedScoringSignals. This includes *all* TrustedScoringSignals for that buyer, because of how those all need to be available in order to reprioritize InterestGroups and apply per-buyer group limits. Aggregated across all components for multi-seller auctions, this records the mean time that bids that were ready for scoring before the config promises were fulfilled were blocked waiting on the config promises. For multi-seller auctions, this only includes bids in the component auctions, and not rescoring of bids in the top-level auction. Aggregated across all components for multi-seller auctions, this records the mean time that bids that were ready for scoring before the seller worklet became available were blocked waiting on the seller worklet. For multi-seller auctions, this only includes bids in the component auctions, and not rescoring of bids in the top-level auction. Only recorded for multi-seller auctions, this records the mean time that top-level bids that were ready for rescoring before the config promises were fulfilled were blocked waiting on the config promises. Only recorded for multi-seller auctions, this records the mean time that top-level bids that were ready for rescoring before the seller worklet became available were blocked waiting on the seller worklet. Number of additional bids intentionally suppressed because at least one of the additional bid's specified negative interest groups was present. This is not an error condition; this is the intentional behavior of negative targeting. Still so, these additional bids are not scored, and do not participate in the auction. Number of additional bids rejected because the additional bid's owner was not allowed to bid in that auction by IsInterestGroupAPIAllowed. These additional bids are not scored, and do not participate in the auction. Number of additional bids rejected because the additional bid specified a currency that didn't match the currency associated with that buyer. These additional bids are not scored, and do not participate in the auction. Number of additional bids rejected because the additional bid from inside the signed additional bid was not structured as expected (encountered an error in DecodeAdditionalBid). This includes additional bids whose replay-prevention fields - auctionNonce, seller and topLevelSeller - are missing or don't match those of the auction. This also includes additional bids whose owner was not found in the auction's interestGroupBuyers. These additional bids are not scored, and do not participate in the auction. Number of additional bids rejected because the signed additional bid in the Ad-Auction-Additional-Bid response header was not a valid base64 string, and so could not be base64-decoded to JSON. These additional bids are not scored, and do not participate in the auction. Number of additional bids rejected because the additional bid from inside the signed additional bid was invalid JSON, and so failed to parse. These additional bids are not scored, and do not participate in the auction. Number of additional bids rejected because the signed additional bid JSON was not structured as expected (encountered an error in DecodeSignedAdditionalBid). These additional bids are not scored, and do not participate in the auction. Number of additional bids rejected because the signed additional bid was invalid JSON, and so failed to parse. These additional bids are not scored, and do not participate in the auction. Number of additional bids scored. These additional bids participate in the auction alongside other additional bids and bids created by calls to generateBid(). In a single-seller auction, this is either zero if the auction doesn't have any config promises or one if it does. This could also be inferred from NumConfigPromises, which would be zero or non-zero respectively in those cases. In a multi-seller auction, this is the number of component auctions that have any config promises, which is different from NumConfigPromises, in that each component auction may independently have any number of config promises. The number of distinct BidderWorklets. Interest groups with the same owner can only share worklets if they have the same bidding script URL and KV server URL. As such, we might observe a non-standard setup in which, for example, a buyer has used a different bidding script URL for each InterestGroup, causing performance issues because of the fragmentation of BidderWorklets. BidderWorklets are shared across component auctions, and this measure counts the number of distinct BidderWorklets used within the scope of the scope of an entire auction, either single- or multi-seller. This may imply that the number of BidderWorklets created in the context of this auction is actually less than this number, because of reuse. However, it also means that this metric can be compared against the NumDistinctOwnersWithInterestGroups metric, which (unlike NumOwnersWithInterestGroups) counts each buyer involved in a multi-seller auction once, even if that buyer participates in multiple components. Total number of bids aborted because the BidderWorklet encountered a fatal error, such as a crash, a JS fetch error, or any other fatal error - excluding timeouts, which are counted by the metric below - aggregated across all Owners and component auctions. Total number of InterestGroups aborted because the per-buyer cumulative timeout was reached, aggregated across all Owners and component auctions. Total number of InterestGroups filtered as per-buyer limits are applied, aggregated across all Owners and component auctions. Total number of InterestGroups filtered based on priority during the loading of InterestGroups, aggregated across all Owners and component auctions. This specifically excludes InterestGroups filtered during load for other reasons, such as failing to satisfy requested seller capabilities. Total number of InterestGroups filtered during bid reprioritization, which happens once bidding signals are available, aggregated across all Owners and component auctions. Related to UMA metric, Ads.InterestGroup.Auction.BidFiltered, except that it's aggregated. To determine the fraction of bids filtered, compare this against the NumInterestGroups metric. Aggregated across all components for multi-seller auctions, this records the total number of bids that were ready for scoring before the config promises were fulfilled. For multi-seller auctions, this only includes bids in the component auctions, and not rescoring of bids in the top-level auction. Aggregated across all components for multi-seller auctions, this records the total number of bids that were ready for scoring before the seller worklet became available. For multi-seller auctions, this only includes bids in the component auctions, and not rescoring of bids in the top-level auction. The total number of config promises for this auction. For a multi-seller auction, this includes promises from both the top-level auction and all component auctions within. The number of buyers with InterestGroups participating in an auction, aggregated across all component auctions. This number excludes buyers with no ads or no script URL, and also those filtered because the priority for all InterestGroups calculated in the BuyerHelper constructor was negative. Unlike NumOwnersWithInterestGroups, it only counts each buyer once, even if it appears in multiple component auctions of a multi-seller auction. Also unlike NumOwnersWithInterestGroups, this metric has a value even if there are no bidders eligible to participate in this auction, though, in this case, that value is likely to be zero. Aggregated across all components for multi-seller auctions, this records the number of occurrences in which GenerateBid was blocked waiting for the Script and/or WASM load. Aggregated across all components for multi-seller auctions, this records the number of occurrences in which GenerateBid was blocked waiting for the fulfillment of the config promises. Aggregated across all components for multi-seller auctions, this records the number of occurrences in which GenerateBid was blocked waiting for the download of the DirectFromSellerSignals. This combines the PerBuyerSignals and AuctionSignals since they should arrive at roughly the same time. Aggregated across all components for multi-seller auctions, this records the number of occurrences in which GenerateBid was blocked waiting for the download of the TrustedBiddingSignals. This includes *all* TrustedBiddingSignals for that buyer, because of how those all need to be available in order to reprioritize InterestGroups and apply per-buyer group limits. Total number of InterestGroups aggregated across all owners and component auctions, equivalent to UMA metric, Ads.InterestGroup.Auction.NumInterestGroups. This is the number of InterestGroups loaded, before per-buyer limits are applied. Excluded from this (and all other metrics counting InterestGroups) are InterestGroups that are incapable of participating in an auction, e.g. those with no bidding logic URLs, or with no ads. No value is defined for this metric if there are no bidders eligible to participate in this auction. Number of instances in which the BidderWorklet doesn't bid, maybe because it failed to load or crashed. Because there's no bid, ScoreAd won't be called for these. Number of instances in which the BidderWorklet only generates one bid. If the KAnonymityBidMode is Simulate or Enforce, this means that the InterestGroup's top generated bid didn't satisfy the kAnonymity threshold, and no separate bid could be generated that did satisfy the kAnonymity threshold. If the KAnonymityBidMode is None, this just means that a bid was successfully generated. For each of these, ScoreAd will be called once. Number of instances in which an InterestGroup generated a complicated sequence of multiple bids that doesn't match any of the other 'NumInterestGroupsWith...' cases. Number of instances in which an InterestGroup's top generated bid did satisfy the kAnonymity threshold, and so no separate bid was needed. For each of these, ScoreAd will be called once. If the KAnonymityBidMode is None, this metric will be 0, because there won't be a separate kAnon bid. Number of instances in which an InterestGroup's top generated bid didn't satisfy the kAnonymity threshold, and a separate bid was generated that did satisfy the kAnonymity threshold. For each of these, ScoreAd will be called twice, once for each of these bids, which has latency and performance implications. If the KAnonymityBidMode is None, this metric will be 0, because there won't be a separate kAnon bid. Total number of negative targeting interest groups aggregated across all owners and component auctions, equivalent to UMA metric, Ads.InterestGroup.Auction.NumNegativeInterestGroups. This is the number of negative targeting interest groups loaded into the AdAuctionNegativeTargeter, regardless of which of these, if any, are later referenced by additional bids. No value is defined for this metric if the auction may not have additional bids, i.e. the additionalBids promise was not provided on the auction config. However, a value is defined for this metric even if no additional bids are provided to the auction, i.e. no additional bids appear in Ad-Auction-Additional-Bid response headers. Number of negative interest groups ignored because they had no matching valid signature on the signed additional bid. By ignoring this negative interest group, the additional bid may participate in the auction where the intent was for it be suppressed by negative targeting. This "failing open" ensures that only the owner of the negative interest group, who created the additionalBidKey, is allowed to negatively target the interest group, and that nobody else can learn whether the interest group is present on the device. Number of negative interest groups ignored because they had a joining origin different from that specified on the additional bid. By ignoring this negative interest group, the additional bid may participate in the auction where the intent was for it be suppressed by negative targeting. This "failing open" ensures that negative targeting can only use targeting data from a single origin. The number of buyers with InterestGroups participating in an auction, aggregated across all component auctions. Equivalent to UMA metric, Ads.InterestGroup.Auction.NumInterestGroups. This number excludes buyers with no ads or no script URL, and also those filtered because the priority for all InterestGroups calculated in the BuyerHelper constructor was negative. However, it counts a buyer repeatedly for each component auction that buyer participates in. No value is defined for this metric if there are no bidders eligible to participate in this auction. The number of owner origins listed in an auction configuration that were excluded from taking part in a FLEDGE auction because they have no eligible interest groups. Equivalent to the UMA histogram named Ads.InterestGroup.Auction.NumOwnersWithoutInterestGroups. This does not count owners that were filtered due to user settings. For multi-seller auctions, this is the sum of all such excluded owners across all component auctions. This double-counts owners that participate in multiple component auctions. Recorded after all interest groups participating in an auction have been retrieved from storage at the start of an auction. This is related to the sibling histogram named "NumOwnersWithInterestGroups". Aggregated across all components for multi-seller auctions, this records the number of occurrences in which ScoreAd was blocked waiting for the Script load. Aggregated across all components for multi-seller auctions, this records the number of occurrences in which ScoreAd was blocked waiting for the download of the DirectFromSellerSignals. This combines the SellerSignals and AuctionSignals since they should arrive at roughly the same time. Aggregated across all components for multi-seller auctions, this records the number of occurrences in which ScoreAd was blocked waiting for the download of the TrustedScoringSignals. This is either the number of component auctions *plus one* for a multi-seller auction, or 1 for a single-seller auction. Equivalent to UMA metric, Ads.InterestGroup.Auction.NumSellersWithBidders. No value is defined for this metric if there are no bidders eligible to participate in this auction. Only recorded for multi-seller auctions, this records the number of bids that were ready for rescoring before the config promises were fulfilled. Only recorded for multi-seller auctions, this records the number of bids that were ready for rescoring before the seller worklet became available. How the auction concluded, whether that's successfully selecting an ad to be shown, or finding no InterestGroups to load, or getting no bids that outbid the contextual bid, or any other outcome. The time in milliseconds between the AuctionRunner constructor (the start of the auction) and the last call to SellerWorklet::ScoreAdIfReady (the end of the ScoreSignalsFetchPhase). The ScoreSignalsFetchPhase represents the time during which any seller worklet is waiting for one or more of the ScoreAd dependencies, notably scoring signals, in anticipation of calling an ad-tech provided scoreAd JavaScript function. This is one of a series of metrics representing the duration of each phase of the auction, where the phases may overlap or have space between them. This is bucketed in linear buckets of 10 ms. The time in milliseconds between the AuctionRunner constructor (the start of the auction) and the first call to SellerWorklet::ScoreAd (the start of the ScoreSignalsFetchPhase). The ScoreSignalsFetchPhase represents the time during which any seller worklet is waiting for one or more of the ScoreAd dependencies, notably scoring signals, in anticipation of calling an ad-tech provided scoreAd JavaScript function. This is one of a series of metrics representing the duration of each phase of the auction, where the phases may overlap or have space between them. This is bucketed in linear buckets of 10 ms. The time in milliseconds between the AuctionRunner constructor (the start of the auction) and the last call to SellerWorklet::DeliverScoreAdCallbackOnUserThread (the end of the ScoringPhase). The ScoringPhase represents the time during which ad-tech provided scoreAd JavaScript functions are being executed. This is one of a series of metrics representing the duration of each phase of the auction, where the phases may overlap or have space between them. This is bucketed in linear buckets of 10 ms. The time in milliseconds between the AuctionRunner constructor (the start of the auction) and the first call to SellerWorklet::ScoreAdIfReady (specifically where that call corresponds to a seller worklet for which all of its ScoreAd dependencies have been satisfied and it's about to call the ad-tech provided scoreAd JavaScript function; the start of the ScoringPhase). The ScoringPhase represents the time during which ad-tech provided scoreAd JavaScript functions are being executed. This is one of a series of metrics representing the duration of each phase of the auction, where the phases may overlap or have space between them. This is bucketed in linear buckets of 10 ms. The time in milliseconds between the AuctionRunner constructor (the start of the auction) and the last call to DebuggableAuctionWorklet::OnHavePid (the end of the WorkletCreationPhase). The WorkletCreationPhase represents the time during which worklets are being started, and will extend throughout the auction if the maximum number of processes is reached for bidder and/or seller worklets. This is one of a series of metrics representing the duration of each phase of the auction, where the phases may overlap or have space between them. This is bucketed in linear buckets of 10 ms. The time in milliseconds between the AuctionRunner constructor (the start of the auction) and the first call to InterestGroupAuction::Request*Worklet (the start of the WorkletCreationPhase). The WorkletCreationPhase represents the time during which worklets are being started, and will extend throughout the auction if the maximum number of processes is reached for bidder and/or seller worklets. This is one of a series of metrics representing the duration of each phase of the auction, where the phases may overlap or have space between them. This is bucketed in linear buckets of 10 ms. yaoxia@google.com chrome-ads-core@google.com Recorded after computing the subresource filter activation for a page at process response time during navigation. This is only recorded when there has previously been an ads intervention triggered on the page within the site data's lifetime (currently 7 days). Clearing site data will reset intervention information. Records the status of the ads intervention, whether the intervention lead to subresource filter activation, measured during page load. Records the last measured intervention recorded for the page, measured during page load. sullivan@chromium.org True if the page loaded in the main frame is an AMP document (has an amp attribute in the html element). Will not be recorded for non-AMP frames. Note that most AMP documents are loaded in a subframe from a reader like Google Search; this metric measures sites that are using AMP as a site-wide framework and not a reader experience. Measures First Input Delay, the duration between the hardware timestamp and the start of event processing on the main thread for the first meaningful input per navigation, in the AMP subframe. See https://goo.gl/tr1oTZ for a detailed explanation. In milliseconds. The number of distinct user interactions in an AMP subframe. See definition of user interaction: https://web.dev/better-responsiveness-metric/#group-events-into-interactions An approximation of a high percentile of user interaction latency of an AMP subframe, in ms. For this metric, we measure the latency of user interactions by Responsiveness.UserInteraction.MaxEventDuration. The maximum value of high percentile of user interaction latency above budget of a page load in AMP subframes, in ms. For this metric, we measure the latency of user interactions by Responsiveness.UserInteraction.MaxEventDuration. Deprecated as of 04/23. It is still reported, but most users should use AmpPageLoad.SubFrame.LayoutInstability.MaxCumulativeShiftScore.SessionWindow.Gap1000ms.Max5000ms. Measures the cumulative layout shift (bit.ly/lsm-explainer) that has occurred during the session, in the AMP subframe. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. Measures the cumulative layout shift (bit.ly/lsm-explainer) that has occurred in the AMP subframe, before any user input or document scroll. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. Measures the maximum cumulative layout shift (bit.ly/cls-explainer) that has occurred in a session window, in the AMP subframe. The gap between two consecutive shifts in a window is not bigger than 1000ms and the maximum window size is 5000ms. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. Measures the time in milliseconds from the navigation in the main frame to the navigation in the AMP subframe. Will be negative in cases where the AMP subframe was prerendered. Will be positive in cases where the main frame was a non-same-document navigation. Should be near-zero when the main frame navigation is a same-document navigation and the AMP subframe was not prerendered. Measures the time in milliseconds from navigation timing's navigation start to the time when the page first paints content, in the AMP subframe. A contentful paint includes a paint of text, image, SVG, or canvas. Measures the time in milliseconds from navigation timing's navigation start to the time the first paint is performed, in the AMP subframe. Measures the time in milliseconds from navigation timing's navigation start to the time when the page first paints the experimental largest content (text or image) within viewport, in the AMP subframe. See http://bit.ly/largest_contentful_paint_explainer for more details. Differs from SubFrame.PaintTiming.NavigationToLargestContentfulPaint in that removed content is still considered a valid candidate. True if the page loaded in any subframe is an AMP document (has an amp attribute in the html element). Will not be recorded for non-AMP frames. Note that most AMP documents are loaded in a subframe from a reader like Google Search; this metric measures usage of such readers. ssid@google.com salg@google.com This event is recorded when page context provider is queried by the Assistant app. Whether the page context was requested based on the structured Assist data provided. ssid@google.com salg@google.com This event is recorded when AssistContent was requested by Android OS. Whether structured data is attached to the AssistContent. nemco@google.com wenyufu@chromium.org Recorded when the user changes auto-dark theming for a site through the app menu. A boolean signaling that auto-dark theming has been disabled. gangwu@chromium.org fgorski@chromium.org Records when the activity enter or exit Android N+ multi-window mode. Enter or exit the multi-window mode. gangwu@chromium.org fgorski@chromium.org Records the multi window state when the activity is shown. The state of the multi-window mode, single or multi-window. shuyng@google.com twellington@chromium.org Records on each page load when the User Agent is overridden with site exception. An enum that records the type of user agent. gangwu@chromium.org fgorski@chromium.org Records when a user rotates the device triggering a screen orientation change. The orientation of the device rotated to. gangwu@chromium.org fgorski@chromium.org Records when a user pressed 'Request Desktop Site' in the app menu. An enum that records the type of user agent. pdyson@chromium.org AppListAppClickData events contain click history metrics for an app. These events are recorded when an app is launched from the launcher on ChromeOS, but may be for an app different from the app launched. See event AppListAppLaunch for more detail on the keys used. The ID of the app launch event. The type of app. 1: CHROME, 2: PLAY, 3: PWA/Bookmark app. Rank apps by total number of clicks. The app with the most clicks has rank 1. The number of clicks on this app in the current session that occurred between 12:00am and 1:00am. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 1:00am and 2:00am. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 2:00am and 3:00am. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 3:00am and 4:00am. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 4:00am and 5:00am. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 5:00am and 6:00am. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 6:00am and 7:00am. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 7:00am and 8:00am. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 8:00am and 9:00am. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 9:00am and 10:00am. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 10:00am and 11:00am. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 11:00am and 12:00pm. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 12:00pm and 1:00pm. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 1:00pm and 2:00pm. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 2:00pm and 3:00pm. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 3:00pm and 4:00pm. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 4:00pm and 5:00pm. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 5:00pm and 6:00pm. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 6:00pm and 7:00pm. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 7:00pm and 8:00pm. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 8:00pm and 9:00pm. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 9:00pm and 10:00pm. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 10:00pm and 11:00pm. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the current session that occurred between 11:00pm and 12:00am. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the last 24 hours. Accurate to 15 minutes. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The number of clicks on this app in the last hour. Accurate to one minute. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. Where the app was last launched from. 1: GRID, 2: SUGGESTED, 3: SHELF. Index recording when in the sequence of app launch clicks the app was last clicked on. 0: most recent app clicked on, 1: app clicked immediately before the most recent, etc. The sequence number of the AppListAppLaunch event that corresponds to this AppListAppClickData event. There will be six AppListAppClickData events for each sequence number. Time since this app was last clicked on to launch. In seconds, bucketed: [1, 59] to the nearest second, [60, 599] to the nearest minute, [600, 1199] to the nearest 5 minutes, [1200, 3599] to the nearest 10 minutes, [3600, 17999] to the nearest 30 minutes, [18000, 86400] to the nearest hour. Total number of clicks launching this app in this session. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. pdyson@chromium.org Recorded when an app is launched from the launcher on ChromeOS. This can be from the suggestion chip, or from the grid of apps. The UKM metrics are not keyed by navigational urls. Instead, for Chrome apps the keys are based upon the app id, for Play apps the keys are based upon a hash of the package name and for PWAs the keys are the urls associated with the PWA. Total number of clicks launching logged apps in the last 24 hours. Accurate to the nearest 15 minutes. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. Total number of clicks launching logged apps in the last hour. Accurate to the nearest minute. Bucketing: values above 20 rounded down to the nearest 10. Maximum value of 200. The type of app. 1: CHROME, 2: PLAY, 3: PWA/Bookmark app. Click method. 1: MOUSE, 2: TOUCH, 3: SYTLUS, 4: KEYBOARD. An enum representing the day of the week that the data was logged in the local time zone. Sunday = 0. The mode of the device. 1: CLOSED_LID (Lid is closed), 2: LAPTOP (Lid is open, tablet mode off or unsupported) , 3: TABLET (Lid is open, tablet mode on or no lid at all). The type of the device. 1: TABLET, 2: LAPTOP. The hour of the day when the data is logged. Hours since midnight in the local time zone. Where the app was launched from. 1: GRID, 2: SUGGESTED, 3: SHELF. The position of the app within the grid/suggested apps/shelf. 0 is the first position. Number of hours in the current session up to this event. Bucketing: exponential buckets, increasing in size by 25%, rounded to the nearest integer. i.e. 0, 1, 2, 3, 4, 5, 6, 7, 9, 12, 15, 18, 23, 28, 36, 44... rahimis@google.com ireneuszszulc@google.com Describes an AtMemory search query and the users response to the results. Recorded when the UI containing the results is destroyed (due to a new search query or the overall UI closing). A bitmask of the data sources that provided an entry. Currently, the sources are local Autofill, Gmail, Calendar, and Photos. The accepted data type. Records the 0-based index (position) of the selected suggestion from the first-level list. In case of selected sub-suggestion from a fly-out menu, this records the position of the top-level suggestion. Records the 0-based index (position) of the selected secondary attribute from the fly-out menu when a suggestion is accepted. -1 means the user accepted the top-level primary attribute directly without selecting from the fly-out menu. 0, 1, 2... means the user opened the fly-out menu and accepted the secondary attribute at that 0-based position. Describes the return value of the search query. Time in milliseconds between when a query was submitted and the query response received. Whether a search result in response to this query was accepted. Whether a suggestion was autofilled after it was accepted. A randomly generated identifier that connects multiple AtMemory.SearchQuery to the same AtMemory.UiSession. Unique only for the duration that a popup is showing. A 0-based index that stores the order in which different AtMemory.SearchQuery entries are emitted. rahimis@google.com ireneuszszulc@google.com Contains information about an AtMemory UI session. A UI session corresponds to showing the Autofill popup (from initial show to hide) on Desktop or the AtMemory bottomsheet on Android. The metric is recorded when the AtMemory UI is closed. A random variable assigned to a specific instance of a form control (e.g. a HTML input element). This should be globally unique and suitable for distinguishing fields. Two different users would have different values. For the same user, the same field in the same form in two different tabs would have two different values. The field signature of the field that AtMemory was triggered on (value taken modulo 1021). The form signature of the field that AtMemory was triggered on (value taken modulo 1021). Whether a search query was submitted. Tracks how the suggestion UI was triggered. Whether at least one search result was accepted during the entire session. Whether a suggestion was filled. A randomly generated identifier that connects multiple AtMemory.SearchQuery to the same AtMemory.UiSession. jamessun@google.com payments-autofill-team@google.com Records whether the Annotated Page Content (APC) fetch was successful. APC fetching is triggered at the start of the AI-based amount extraction process to generate the structured page data that will subsequently be analyzed by the model to retrieve the checkout amount. Recorded once per page load when the APC fetch attempt completes. The result of the APC fetch. jamessun@google.com payments-autofill-team@google.com Records the final result of the AI-based checkout amount extraction process. Recorded once per page load, either when the model execution completes (successfully, with failure, or with invalid data) or when the client-side timeout is reached. Records duration in milliseconds from when the AI amount extraction is initialized to when an empty response is received from the model. Records duration in milliseconds from when the AI amount extraction is initialized to when an invalid response is received from the model. The reason the AI amount extraction response was invalid. The result of the AI amount extraction. Records duration in milliseconds from when the AI amount extraction is initialized to when a non-empty valid response is received from the model. jamessun@google.com longsheng@google.com Records the result of the amount extraction process. This process is to search for the final checkout amount from the merchants' checkout page and respond to the browser process. This event starts from the moment amount extraction is initiated in the browser process to when a response is received. Logged when the search is finished. Records duration in milliseconds from when the amount extraction is triggered to when an empty result of the search is returned to the browser process. An enum that records the result of the amount extraction: success, failure or timeout. Records duration in milliseconds from when the amount extraction is triggered to when a non-empty result of the search is returned to the browser process. schwering@google.com autofill-internal@google.com Recorded whenever the user hits "@" twice in quick succession. A random variable assigned to a specific instance of a form control (e.g. a HTML input element). This should be globally unique and suitable for distinguishing fields. Two different users would have different values. For the same user, the same field in the same form in two different tabs would have two different values. The field signature of the field where the "@@" sequence was typed (value taken modulo 1021). The form control type of the field where "@@" was typed. The form signature of the form where "@@" was typed (value taken modulo 1021). jamessun@google.com payments-autofill-team@google.com Records whether the AI-based checkout amount is within the issuer's transaction range. Recorded once per page load when AI amount extraction completes. The ID of the BNPL issuer. True if the amount is within the issuer's supported range, False otherwise. wilsonlow@google.com payments-autofill-team@google.com Recorded when a BNPL autofill suggestion was accepted on the current page. Only ever set to true, thus serving as a counter. wilsonlow@google.com payments-autofill-team@google.com Recorded when a BNPL suggestion was added to the payments autofill dropdown and shown to the user. Only ever set to true, thus serving as a counter. sebsg@chromium.org Whether the upload was proposed to the user or the reasons why it was not. The value is a bitmask of |CardUploadDecisionMetric|. schwering@google.com Recorded when a credit card form is autofilled. Logs the field types which Autofill would have filled if the credit card on file was complete and there was no security policy. Logs a qualitative summary of the field types which Autofill would have filled if the credit card on file was complete and there was no security policy. Logs the field types which Autofill would have filled into visible fields if the credit card on file was complete and there was no security policy. Logs a qualitative summary of the field types which Autofill would have filled into visible fields if the credit card on file was complete and there was no security policy. Logs the field types which Autofill actually filled. Logs a qualitative summary of the field types which Autofill actually filled. Logs the field types which Autofill actually filled into visible fields. Logs a qualitative summary of the field types which Autofill actually filled into visible fields. Logs the field types which Autofill would have filled if there was no security policy. Logs a qualitative summary of the field types which Autofill would have filled if there was no security policy. Logs the field types which Autofill would have filled into visible fields if there was no security policy. Logs a qualitative summary of the field types which Autofill would have filled into visible fields if there was no security policy. An approximately 10-bit hash of the form's actual signature. Disabled policy-controlled feature "autofill" prevented a cross-origin credit-card field from being filled. With "autofill", a frame can allow a child-frame to be autofilled across origin. koerber@google.com battre@google.com Records when a user edits a value that was filled by autofill. The signature of the field (mod 1021). This is the hash identifier used to denote this field for query and voting purposes. See components/autofill/core/common/signatures.cc for more details. The signature of the form (mod 1021). This is the hash identifier used to denote this form for query and voting purposes. See components/autofill/core/common/signatures.cc for more details. Field's first overall `FieldType`. See `AutofillType::GetTypes()`. Field's second overall `FieldType`. See `AutofillType::GetTypes()`. Field's third overall `FieldType`. See `AutofillType::GetTypes()`. Field's fourth overall `FieldType`. See `AutofillType::GetTypes()`. rogerm@chromium.org Recorded when Chrome validates field type predictions (on submit, on leaving a form, on receiving a form with autocomplete attributes). The signature of the field (mod 1021). This is the hash identifier used to denote this field for query and voting purposes. See components/autofill/core/common/signatures.cc for more details. The signature of the form (mod 1021). This is the hash identifier used to denote this form for query and voting purposes. See components/autofill/core/common/signatures.cc for more details. The field is currently autofilled. Time since form parse. The system which generated the prediction: heuristics, server, overall. The event which triggered the validation: Submission, Left Form, Form has Autocomplete attributes. The field was autofilled and subsequently edited. rogerm@chromium.org Recorded when Chrome validates field type predictions (on submit, on leaving a form, on receiving a form with autocomplete attributes). The observed type of the field. See: autofill::FieldType The signature of the field The signature of the form. Time since form parse. The predicted type of the field. See: autofill::FieldType The system which generated the prediction: heuristics, server, overall. The event which triggered the validation: Submission, Left Form, Form has Autocomplete attributes. osaul@google.com Recorded when user interacted with autofill. Logs autofill form events funnel. See |FormEvent|. Type of form. Stored as bitvector. A set ith bit implies enum FormTypeNameForLogging's ith type was detected. Multiple types are possible. Time since form parse. kazinova@google.com Records whether the form was successfully filled by Autofill on iOS. The metric is used to measure silent renderer fails when the filling request is executed successfully but the results reported in a callback are empty. A boolean indicating success or failure. osaul@google.com Recorded when we parse a form to log form metadata and autofill settings that apply to all subsequent events for this form. FormSignature is approximately 10 bit hash of form structure. Type of form. Stored as bitvector. A set ith bit implies enum FormType's ith type was detected. Multiple types are possible. True for credit card forms, false for address/profile forms. Number of local credit cards or local autofill profiles. Number of masked and full server credit cards or server autofill profiles. battre@chromium.org kolos@chromium.org nepper@chromium.org These metrics are a clone of Autofill's UMA key metrics to measure Autofill quality. They are recorded on navigation if a form was submitted before. Number of Autofill fills invoked by the user on the current form. Has value 1 if user accepted the suggestion to fill values on a submitted form, 0 otherwise. This is counted as 1 if the user accepted such a suggestion at any time, regardless whether other suggestions were ignored or whether the accepted suggestion was reverted afterwards. Only recorded if suggestions were shown to the user. Has value 1 if the user filled any fields via Autofill on a submitted form, 0 otherwise. If the user reverts the filling, the form is still counted as filled. This metric is different from FillingAcceptance in the sense that the latter is only recorded if the user has actually seen a prompt to fill data. Has value 1 for submitted and autofilled forms if user edited at least one filled value before submitting the form, 0 otherwise. Only recorded if any suggestion was filled. Has value 1 for submitted forms when Chromium had any address or payments data to fill, 0 otherwise. It does not matter whether the user actually requested to fill the data. It does not matter either whether the data available for filling corresponded to the submitted data. Number of times a user focused on a form element and changed its value. Consecutive value changes on the same form element are not counted, neither are JS invoked changes. Also, re-focusing a form element and changing its value is only counted if another form element's value was changed in the meantime. Type of form. Stored as bitvector. A set ith bit implies enum FormTypeNameForLogging's ith type was detected. Multiple types are possible. yishuil@google.com wilsonlow@google.com payments-autofill-team@google.com Recorded when a user selects the Pay Later tab from the payments autofill dropdown. Only ever set to true, thus serving as a counter. yishuil@google.com wilsonlow@google.com payments-autofill-team@google.com Recorded when the Pay Later tab was shown in the payments autofill dropdown. Only ever set to true, thus serving as a counter. yishuil@google.com wilsonlow@google.com payments-autofill-team@google.com Recorded when a BNPL suggestion from the Pay Later tab was accepted by users. Only ever set to true, thus serving as a counter. osaul@google.com Recorded when user selects a suggestion and we fill the form with that suggestion. The signature of the field The signature of the form. True for credit card forms, false for address/profile forms. Time since form parse. Whether the suggestion was from a local card/autofill profile or from a server card/autofill profile. osaul@google.com Recorded when a suggestion is shown to user. The signature of the field The signature of the form. Field's |FieldType| based on heuristics. See |AutofillField.heuristic_type()|. Field's autocomplete field type hint. See |AutofillField.html_type()|. Time since form parse. Field's |FieldType| returned by server. See |AutofillField.server_type()|. osaul@google.com Recorded when user edits a text field. The text field may have been autofilled. The signature of the field Field's first `FieldTypeGroup`. See `AutofillType::GetGroups()`. Field's second `FieldTypeGroup`. See `AutofillType::GetGroups()`. Field's third `FieldTypeGroup`. See `AutofillType::GetGroups()`. Field's fourth `FieldTypeGroup`. See `AutofillType::GetGroups()`. The signature of the form. Field's |FieldType| based on heuristics. See |AutofillField.heuristic_type()|. Whether the field's autocomplete hint specified 'billing' or 'shipping'. See |AutofillField.html_mode()|. Field's autocomplete field type hint. See |AutofillField.html_type()|. True if the field was autofilled. See |AutofillField.is_autofilled|. True if the field was empty. See |AutofillField.IsEmpty()|. Time since form parse. Field's |FieldType| returned by server. See |AutofillField.server_type()|. vidhanj@google.com koerber@google.com chrome-autofill-alerts@google.com Recorded after a form submission that either resulted in a save prompt for addresses or in a silent update of an already existing profile. Logs if data was imported from a field that has autocomplete=UNRECOGNIZED. Minimum number of fields whose removal makes the import candidate a duplicate of any existing profile. If the duplication rank is 1, the import candidate can be called a quasi-duplicate of an existing profile. Only recorded for new profiles and if a stored profile existed. Logs the import type of the storage prompt. An import type can be either the import of a new profile, an update of an existing profile that results in a confirmation dialog, or a silent update that only changes the structure of an already stored address but does not affect any settings-visible stored data. Logs the count for the number of fields that have been edited by the user in the storage dialog before the storage was confirmed. Logs the phone number validity status based on whether the phone number was imported in the profile or not or removed beforehand. Logs the user decision based on the storage prompt. True if the user had already saved an address profile, false otherwise. battre@chromium.org chrome-autofill-team@google.com All information about a specific field, recorded at form submission or form destruction time (whatever comes first). (As of M132) Info are recorded only for fields that have at most `autofill::kMaxLogEventsPerField` events. The field's autocomplete attribute value, the values are from |AutofillMetrics::AutocompleteState|. Whether filling the field was skipped during autofill, and if it was skipped, what the skipping reasons were. Recorded as a DenseSet of |FieldFillingSkipReason|. A bit vector of autofill statuses for a field, listed in |AutofillMetrics::AutofillStatus|. The first bit represents whether the field is focusable, the second means whether the field was focused.... These autofill statuses will not be stored in UKM individually. We create functions to read the individual autofill status from a bit vector in queries. go/zmwtc The total number of all the field log events created for this field. A random variable assigned to a specific instance of a form control (e.g. a HTML input element). This should be globally unique and suitable for distinguishing fields. Two different users would have different values. For the same user, the same field in the same form in two different tabs would have two different values. The signature of the field, a 10 bit hash value (0-1021) of the field name and field type used to denote this field for query and voting purposes. A form control type of a web form-associated element. We only list the ones that are handled by Autofill, see |autofill::FormControlType|. A random variable assigned to a specific instance of a HTML form in a specific renderer. This should be globally unique and suitable for counting distinct forms. Two different users would have different values. A single user loading the same form in two different tabs would have two different values. This is used to link the fields to a form in Autofill2.FormSummary. The signature of the form, which is a 10-bit hash of the form's domain, name and all the supported form fields (including empty names). The type of the field predicted the regular expressions. The types are from |autofill::FieldType|. The mode according to the autocomplete attribute of a field. Only set if a field has a valid autocomplete attribute. Currently this is used to specify if the field is part of the billing or shipping address. See autofill::mojom::HtmlFieldMode. The field type specified in the autocomplete attribute of the field. Only set if a field has a valid autocomplete attribute. See autofill::mojom::HtmlFieldType. The type of the field predicted by the ML model after consolidation with regex heuristics (the ML model is not allowed to override the regex heuristics in some cases). The types are from |autofill::FieldType|. The final field type from the list of |autofill::FieldType| that we choose after rationalization, which is used to determine the autofill suggestion when the user triggers autofilling. If multiple fields have the same signature, this indicates the position within this set of fields. This allows us to understand problems related to duplicated field signatures. A single HTML form can consist of multiple sections. A section is a set of fields that are filled together. For example a single form can contain a shipping and a billing address. Each of those corresponds to one section. The sections are mapped to consecutive natural numbers starting at 1. The section IDs are unique per form. The major source of predictions from the Autofill crowdsourced server, see AutofillQueryResponse::FormSuggestion::FieldSuggestion::FieldPrediction::Source. The additional source of predictions from the Autofill crowdsourced server if there is an additional prediction, see AutofillQueryResponse::FormSuggestion::FieldSuggestion::FieldPrediction::Source. The type of the field predicted from the Autofill crowdsourced server from majority voting (and some more complex stuff). The types are from |autofill::FieldType|. The additional type of the field predicted from the Autofill crowdsourced server if there is an additional prediction. The types are from |autofill::FieldType|. This is an annotation for server predicted field types which indicates that a manual override defines the server type. This allows us to assess the impact of server overrides and detect possible problems with it. Whether the field's type was changed by running the rationalization. battre@chromium.org chrome-autofill-team@google.com Information about a specific field, recorded after form submission. A random variable assigned to a specific instance of a form control (e.g. a HTML input element). This should be globally unique and suitable for distinguishing fields. Two different users would have different values. For the same user, the same field in the same form in two different tabs would have two different values. FormSessionIdentifier and FieldSessionIdentifier together are used to link to the field in Autofill2.FieldInfo. The signature of the field, a 10 bit hash value (0-1021) of the field name and field type used to denote this field for query and voting purposes. A random variable assigned to a specific instance of a HTML form in a specific renderer. This should be globally unique and suitable for counting distinct forms. Two different users would have different values. A single user loading the same form in two different tabs would have two different values. FormSessionIdentifier and FieldSessionIdentifier together are used to link to the field in Autofill2.FieldInfo. The signature of the form, which is a 10-bit hash of the form's domain, name and all the supported form fields (including empty names). The time interval from form parse to form submission in ms in exponential buckets from |GetSemanticBucketMinForDurationTiming()|. If multiple fields have the same signature, this indicates the position within this set of fields. This allows us to understand problems related to duplicated field signatures. The source of form submission. The source is from autofill::mojom::SubmissionSource. The first possible field submitted type is determined by comparing the submitted value in the field with the stored user profiles. We will have at most three possible field submitted types. The types are from |autofill::FieldType|. The second possible field submitted type is determined by comparing the submitted value in the field with the stored user profiles. We will have at most three possible field submitted types. The types are from |autofill::FieldType|. The third possible field submitted type is determined by comparing the submitted value in the field with the stored user profiles. We will have at most three possible field submitted types. The types are from |autofill::FieldType|. battre@chromium.org chrome-autofill-team@google.com Tracks metrics of forms of type FormTypeNameForLogging::kPostalAddress and FormTypeNameForLogging::kCreditCard - if at least one field was focused whose type did not belong to FormType::kUnknownFormType. Autofocus on page load, focus-by-pressing-the-tab-key and focus-by-tapping/clicking count as focus events. The event name contains the term "complex forms" because only forms of type FormTypeNameForLogging::kPostalAddress and FormTypeNameForLogging::kCreditCard are logged (to save space and bandwidth). The event is recorded when the form object is deallocated. A bit-vector with bits set according to `FormType`. A set bit indicates that at least one field of the `FormType` was focused via a tap or click such that the field was eligible for an autofill if data existed. A bit-vector with bits set according to `FormType`. A set bit indicates that at least one field of type `FormType` was autofilled at least once (if the autofill is reverted or modified, this is ignored). For an ablation experiment we define a grid of 14-day windows. The ablation state is determined by a hash value of a user-seed (random number), the main frame origin, the ablation window. If, for example, an ablation window lasts 14 days, this can be a value between 0 and 13. It indicates whether the user is in the begining or at the end of an ablation window. A bit-vector with bits set according to `FormType`. A set bit indicates that at least one field of type `FormType` was autofilled and later modified by the user. A random variable assigned to a specific instance of a HTML form in a specific renderer. This should be globally unique and suitable for counting distinct forms. Two different users would have different values. A single user loading the same form in two different tabs would have two different values. This is used to link the fields in this form in Autofill2.FieldInfo and more details in Autofill2.FormSummary. The signature of the form (mod 1021). This is the hash identifier used to denote this form. See components/autofill/core/common/signatures.cc for more details. Bit vector with the following bits from FormTypeNameForLogging: - kUnknownFormType (1<<0), - kAddressForm (1<<1), - kCreditCardForm (1<<2) (A form with any field classified as a CC field), - kPasswordForm (1<<3), - kStandaloneCvcForm (1<<4), - kEmailOnlyForm (1<<5), - kPostalAddressForm (1<<6), Note that a form needs to be a kCreditCardForm or kPostalAddressForm to be logged. A bit-vector with bits set according to `FormType`. A set bit indicates that at least one field of type `FormType` was non-empty at submission time. If 1, the ablation study is in dry run mode, meaning that clients are diverted into ablation and control groups, but they observe no behavioral change. A bit-vector with bits set according to `FormType`. A set bit indicates that that for this form and the respective form type, the user was in the ablation group, meaning that the user did not get the autofill UI, regardless whether they had data available for filling. A bit-vector with bits set according to `FormType`. A set bit indicates that that for this form and the respective form type, the user was in the conditional ablation group, meaning that the user had data that could be filled, but the filling was suppressed due to the ablation. A bit-vector with bits set according to `FormType`. A set bit indicates that that for this form and the respective form type, the user was in the ablation control group. This is the corresponding control group to IsInAblationGroupOfAblation. Autofill is not disabled but metrics are recorded. A bit-vector with bits set according to `FormType`. A set bit indicates that that for this form and the respective form type, the user was in the conditional ablation control group. This is the corresponding control group to IsInAblationGroupOfConditionalAblation. Autofill is not disabled but metrics are recorded. The time interval from first interaction on the form to form submission in ms in buckets from |GetSemanticBucketMinForAutofillDurationTiming()|. An interaction is defined as modifying the value of any field in the form. A bit-vector with bits set according to `FormType`. A set bit indicates that for at least one field of type `FormType` a suggestion was queried and available for the user. A bit is unset if no suggestions were queried for a form type. A bit-vector with bits set according to `FormType`. A set bit indicates that at least one field of the `FormType` was modified by the user (typed text, pasted text or autofilled). Whether the form was submitted. battre@chromium.org chrome-autofill-team@google.com Information recorded for a HTML form element on the HTML page, recorded at form submission or form destruction time (whatever comes first). A bit vector of some autofill form events that records a funnel of this autofill form. The autofill form events are from an enum list specifying the possible status of the autofill forms, such as if a dropdown with suggestions was shown, if the user interacted with a field in the form. Some of the events are triggered separately for address and credit card forms, e.g. if all the fillable fields in the credit card forms were filled. New form events can be added over time and a form can have multiple form events. This bit vector stored only events 0..63. A bit vector of some autofill form events that records a funnel of this autofill form. This contains the entries 64...127 shifted right by 64 bits. A random variable assigned to a specific instance of a HTML form in a specific renderer. This should be globally unique and suitable for counting distinct forms. Two different users would have different values. A single user loading the same form in two different tabs would have two different values. This is used to link the fields in this form in Autofill2.FieldInfo. The signature of the form, which is a 10-bit hash of the form's domain, name and all the supported form fields (including empty names). The time interval from first interaction on the form to form submission in ms in exponential buckets from |GetSemanticBucketMinForDurationTiming()|. An interaction is defined as modifying the value of any field in the form. The time interval from form parse to form submission in ms in exponential buckets from |GetSemanticBucketMinForDurationTiming()|. A sampling rate is a precomputed positive integer and is applied to each form and adaptive to the popularity of forms. The most popular forms are classified into a `very popular` category and maybe apply a 1:100 sampling while the less popular forms have a sampling of 1:1. If the sampling is 1:X, this sample rate value stores the `X`. Whether the form was submitted. jihadghanna@google.com Recorded when the user accepts a suggestion on some trigger field. This includes only suggestions going through the Autofill codebase, meaning that password suggestions are excluded. The index of the accepted suggestion in the list of shown suggestions. A random variable assigned to a specific instance of a form control (e.g. a HTML input element). This should be globally unique and suitable for distinguishing fields. Two different users would have different values. For the same user, the same field in the same form in two different tabs would have two different values. The signature of the field. A random variable assigned to a specific instance of a HTML form in a specific renderer. This should be globally unique and suitable for counting distinct forms. Two different users would have different values. A single user loading the same form in two different tabs would have two different values. The signature of the form. FieldType of the trigger field where the suggestions where accepted. For fields that have multiple field types, one of them is chosen arbitrarily. Length of the value in the field before accepting the suggestion. The type of the accepted suggestion. jihadghanna@google.com jkeitel@google.com schwering@google.com Records when a user edits a value that was filled by an Autofill AI suggestion. The first server AI type of the field as classified by Autofill's ML model. The second server AI type of the field as classified by Autofill's ML model. The third server AI type of the field as classified by Autofill's ML model. The fourth server AI type of the field as classified by Autofill's ML model. The type of entity for which the current metric record is being logged (E.g. Passport, Driver's License, etc.). Type of the event being recorded, see `AutofillAiUkmLogger::EventType` for more details. A random variable assigned to a specific instance of a form control (e.g. a HTML input element). This should be globally unique and suitable for distinguishing fields. Two different users would have different values. For the same user, the same field in the same form in two different tabs would have two different values. The signature of the field (mod 1021). This is the hash identifier used to denote this field for query and voting purposes. See components/autofill/core/common/signatures.cc for more details. The first `FieldType` as classified by Autofill. See `AutofillType::GetAutofillAiTypes()`. The second `FieldType` as classified by Autofill. See `AutofillType::GetAutofillAiTypes()`. The third `FieldType` as classified by Autofill. See `AutofillType::GetAutofillAiTypes()`. The fourth `FieldType` as classified by Autofill. See `AutofillType::GetAutofillAiTypes()`. The source of the format string of the field, see `FormatStringSource` in components/autofill/core/browser/autofill_field.h for more info. The order of the event in the sequence of event sharing the same FormSessionIdentifier. A random variable assigned to a specific instance of a HTML form in a specific renderer. This should be globally unique and suitable for counting distinct forms. Two different users would have different values. A single user loading the same form in two different tabs would have two different values. This is used to link the fields to a form in AutofillAi.KeyMetrics. The signature of the form (mod 1021). This is the hash identifier used to denote this form for query and voting purposes. See components/autofill/core/common/signatures.cc for more details. jihadghanna@google.com jkeitel@google.com schwering@google.com These metrics are a clone of Autofill's UMA key metrics to measure Autofill quality. They are recorded on navigation if a form was submitted before, only if the submitted form had Autofill AI prediction. The number of fields at form submission that are autofilled and that were autofilled by AutofillAi. Fields filled with any other source will be excluded from this count. The number of fields at form submission that are autofilled. This will include fields that are autofilled by any source except Autocomplete, including AutofillAi. The type of entity for which the current metric record is being logged (E.g. Passport, Driver's License, etc.). Has value 1 if user accepted the suggestion to fill values on a submitted form, 0 otherwise. This is counted as 1 if the user accepted such a suggestion at any time, regardless whether other suggestions were ignored or whether the accepted suggestion was reverted afterwards. Only recorded if suggestions were shown to the user. Has value 1 if the user filled any fields via Autofill on a submitted form, 0 otherwise. If the user reverts the filling, the form is still counted as filled. This metric is different from FillingAcceptance in the sense that the latter is only recorded if the user has actually seen a prompt to fill data. Has value 1 for submitted and autofilled forms if user edited at least one filled value before submitting the form, 0 otherwise. Only recorded if any suggestion was filled. Has value 1 for submitted forms when Chromium had any address or payments data to fill, 0 otherwise. It does not matter whether the user actually requested to fill the data. It does not matter either whether the data available for filling corresponded to the submitted data. A random variable assigned to a specific instance of a HTML form in a specific renderer. This should be globally unique and suitable for counting distinct forms. Two different users would have different values. A single user loading the same form in two different tabs would have two different values. This is used to link the fields to a form in AutofillAi.FieldInfo. The signature of the form (mod 1021). This is the hash identifier used to denote this form for query and voting purposes. See components/autofill/core/common/signatures.cc for more details. Has value 1 if the user is opted in into Autofill AI suggestions and filling and 0 otherwise. Has value 1 if the user didn't edit any field manually at submission time and 0 otherwise. jihadghanna@google.com jkeitel@google.com brunobraga@google.com Records when an AutofillAi prompt is shown to the user, such as saving or updating entities. The type of entity for which the current metric record is being logged (E.g. Passport, Driver's License, etc.). A random variable assigned to a specific instance of a HTML form in a specific renderer. This should be globally unique and suitable for counting distinct forms. Two different users would have different values. A single user loading the same form in two different tabs would have two different values. This is used to link the fields to a form in AutofillAi.KeyMetrics. The signature of the form (mod 1021). This is the hash identifier used to denote this form. See components/autofill/core/common/signatures.cc for more details. Describes the type of prompt shown to the user. This could be for example a save or update entity prompt. The decision made by the user after being shown a prompt. This could be for example, accepted, declined or ignored. Describes to which storage the data will be saved if the user accepts the prompt, such locally or in the wallet server. hajimehoshi@chromium.org Recorded at a history commit for each reason passed to BackForwardCache::DisableForRenderFrameHost. The source IDs match with the source IDs of HistoryNavigation events. For history navigations, records the reason passed to BackForwardCache::DisableForRenderFrameHost calls. The reason is specified as a 48-bit int representing the source (blink or an embedder) and a source-specific 16-bit int representing the reason specific to that source. The metric is recorded when a history navigation commits and emits a sample for each reason. One history navigation can emit multiple reasons, and these reasons are recorded as multiple events with the same source ID based on the navigation ID. hajimehoshi@chromium.org fergal@chromium.org Recorded at a history commit for each reason passed to RenderFrameHost::IsInactiveAndDisallowActivation. The source IDs match with the source IDs of HistoryNavigation events. For history navigations, records the reason passed to RenderFrameHost::IsInactiveAndDisallowActivation. The metric is recorded when a history navigation commits and emits a sample for each reason. One history navigation can emit multiple reasons, and these reasons are recorded as multiple events with the same source ID based on the navigation ID. nator@chromium.org rayankans@chromium.org peter@chromium.org A BackgroundFetch event is logged before a background fetch is started from a document context. Boolean for whether the background fetch was denied due to permission. This includes only the content setting permission. The value of downloadTotal provided with the background fetch. This is the number of bytes that the developer expects to be downloaded with the background fetch. This number is exponentially bucketed for privacy reasons, and uses the UKM GetExponentialBucketMin method with a value of 2.0 for spacing. Boolean for whether a title was provided with the background fetch. Count of icons provided with the background fetch. Number of requests in the background fetch. This number is exponentially bucketed for privacy reasons, and uses the UKM GetExponentialBucketMin method with a value of 2.0 for spacing. Ratio of the ideal icon to the chosen icon size, times hundred. This will be set to -1 if either ideal icon size is 0, or if none of the provided icons are suitable. nator@chromium.org rayankans@chromium.org peter@chromium.org A BackgroundFetchDeletingRegistration event is logged when a background fetch job is being deleted. Boolean for whether the background fetch job was cancelled from the UI. nator@chromium.org rayankans@chromium.org peter@chromium.org Logged when a Background Sync one-shot registration is completed. Integer value of the maximum number of allowed retries. Integer value of the number of retries / sync events dispatched. The enum value of the completion status, defined in blink::ServiceWorkerStatusCode. nator@chromium.org rayankans@chromium.org peter@chromium.org Logged when a one-shot Background Sync registration is created. Boolean for whether a sync event can be immediately dispatched for this registration. Boolean for whether this is a duplicate registration. phillis@chromium.org cmumford@chromium.org Metrics that measure the use of Badging API Logged when navigator.setAppBadge or navigator.clearAppBadge is called. felixarntz@google.com nrosenthal@chromium.org Metrics stating the detected major+minor version of known content management systems when detected. This event is recorded when the page is loaded. For interoperability with the framework detection above, this reports each content management system individually instead of a single number. The version numbers here are a bit mask: 0xMMmm (Major+minor), e.g. version 16.3 would be 0x1003 A bitmask (0xMMmm) representing the Drupal version A bitmask (0xMMmm) representing the WordPress version npm@google.com web-identity-eng@google.com Records metrics for the Email Verification Protocol (EVP) flow. Surfaced once per EVP request attempt when the request flow finishes. Records the outcomes and failure reasons of the high-level Email Verification Protocol (EVP) flow from autofill side. Recorded once per autofill attempt when the verification flow finishes. Records the outcome and dismissal reasons of the Email Verification Permission UI. Recorded once per permission UI showing when it is dismissed or accepted. Records the status of the IsVerifiable stage of an Email Verification Protocol (EVP) request. Recorded once per request when the IsVerifiable check completes. Records the status of the Verify stage of an Email Verification Protocol (EVP) request. Recorded once per request when the Verify operation completes. Records the time taken (in milliseconds) from the start of checking if an email is verifiable until the pre-prompt background verification check completes. Records the time taken (in milliseconds) from the start of the token verification call until the post-prompt token verification and fetch completes. npm@google.com web-identity-eng@google.com Records metrics for the Email Verification Protocol (EVP) when a form containing a verified email verification token is submitted. Records when a form containing a verified email verification token is submitted. Recorded once when the form submission occurs after successful email verification. yigu@chromium.org web-identity-eng@google.com Records performance metrics for FedCM (Federated Credential Management) API. Recorded when a FedCmMetrics object is destroyed, typically when cleaning up a RequestImpl object which is created and associated with a call to the FedCM API. Records the number of accounts that match an account label. Records at the time the API call is completed. Records the number of times an accounts dialog is shown per API call. Records at the time the API call is completed. This is in units of dialogs. Records the page scroll (y-axis) position upon account selection. Records once per user selecting an account by clicking on a FedCM dialog. Bucketized exponentially with a standard bucket spacing of 1.15. Records at the time the API call is completed. This is in units of pixels. Records the number of requests sent to the accounts endpoint per API call. Records at the time the API call is completed. This is in units of requests. Records the size of the accounts received from the IdP before applying any filters. Records one sample per accounts fetch if there's at least one account returned. The units is number of accounts (using exponential bucketing). Records the size of the accounts received from the IdP after applying the filters. Records one sample per accounts fetch if there's at least one account left after filtering. The units is number of accounts (using exponential bucketing). Records whether the FedCM auto re-authn call is blocked because the auto re-authn content settings permission is disabled. Records at most one sample per FedCM API with auto re-authn enabled: some failures could occur before this metric is recorded. Records at the time the API call is completed. Records whether the FedCM auto re-authn call is blocked because the auto re-authn is under embargo, i.e. due to cooldown. Records at most one sample per FedCM API with auto re-authn enabled: some failures could occur before this metric is recorded. Records at the time the API call is completed. Records whether auto-reauthn was disabled by an embedder (e.g. due to an active actor task). Records at most one sample per FedCM API with auto re-authn enabled: some failures could occur before this metric is recorded Records whether the FedCM auto re-authn call is blocked because preventSilentAccess() has been called on this site. Records at most one sample per FedCM API with auto re-authn enabled: some failures could occur before this metric is recorded. Records at the time the API call is completed. Records whether there was zero, one, or multiple returning accounts when a site calls the FedCM API with autoReauthn: true. Records at most one sample per failed API call: some failures could occur before this metric is recorded. Records at the time the API call is completed. Records whether auto re-authn checks succeeded or not. The count would provide roughly the number of FedCM re-authn attempts, the true count would provide the number of times the re-authn UI is shown to the user, and the false count would provide the number of attempts where auto re-authn was blocked for some non-network related reason. Records at most one sample per FedCM API with auto re-authn enabled: some failures could occur before this metric is recorded. Records at the time the API call is completed. Records the outcome of the last account chooser the user sees in a button flow. Recorded at most once per button flow. For example, if the user selects an account on the account chooser and then clicks on the back button in the next dialog where an account chooser is shown again, leaving would only record one sample. Records the outcome of the last disclosure dialog the user sees in a button flow. Recorded at most once per button flow. For example, if the user clicks on the back button in the disclosure dialog and then clicks on an account where the disclosure dialog is shown again, leaving would only record one sample. Records the outcome of the loading dialog in button flow. Button flows always start with a loading dialog. Recorded once per button flow dialog destroyed. For example, when the dialog gets updated into an account chooser, the user can then select an account and subsequently cause the dialog to be destroyed once they are signed in. Upon destroyal, this metric is recorded. Records the reason that the FedCM dialog was closed (e.g. user clicked the dialog close button). Recorded when the FedCM dialog is closed without the user selecting an account. This includes instances where the close action is less intentional such as the virtual keyboard showing triggering the dialog close. Records at the time the API call is completed. Records whether the FedCM dialog was shown. Recorded once per API call when the request is fulfilled (either resolved or rejected). Records the type of frame that invokes disconnect(). Records once for each disconnect() call. Possible values are: main frame, same-site iframe, and cross-site iframe, where the site of the caller frame is compared with the site of the main frame. Records at the time the API call is completed. Records the number of accounts that match a domain hint. Records at the time the API call is completed. Records the outcome of the error dialog. Recorded at most once per API call when the API call is completed. Records the type of error dialog shown. Recorded at most once per API call when the API call is completed. Records the type of token response received. Recorded at most once per API call when the API call is completed. Records the session ID associated to the FedCM call for which this event is recording metrics. Each FedCM call gets a random integer session id, which helps group UKM events by the session id. Records at the time the API call is completed. Records the type of frame that invokes the FedCM API. Records once per API call. Possible values are: main frame, same-site iframe, and cross-site iframe, where the site of the caller frame is compared with the site of the main frame. Same-origin iframes are recorded under the same-site category. Records at the time the API call is completed. Records a 1 whenever a FedCM API call includes the nonce parameter. Records towards the beginning of the call, except for some cases where the call is immediately rejected. Records a 1 whenever a FedCM API call includes the nonce parameter, but does not include it inside the params JSON blob. Records towards the beginning of the call, except for some cases where the call is immediately rejected. Records whether a user is signing up with the API. Records once per successful account fetch from the IdP server after the accounts filtering. If there are multiple accounts, record the user as sign-in if IdP claims at least one of the accounts is sign-in. If an IdP does not share the user's sign-in state, record the state stored in the browser. Records at the time the API call is completed. Records the count of identity providers right before fetching endpoints with multi IDP flag enabled. It counts distinct providers including the ones a user is not signed in to. Recorded at most once per API call. Records at the time the API call is completed. Records the number of accounts that match a login hint. Records at the time the API call is completed. Records the number of times a mismatch dialog is shown per API call. Records at the time the API call is completed. This is in units of dialogs. Records whether an API call is rejected due to there's an existing in-flight API call initiated by other IdPs. Recorded once per promise rejection because of an existing API call. Records at the time the API call is completed. Records the RpMode of two consecutive requests. Recorded every time when a token request is invoked while there is another one pending. Records the number of IDPs that result in a 'mismatch' status from a single FedCM get() invocation. Records at the time the API call is completed. Records some integer 0 or greater. Uses exponential count bucketing. Records the number of IDPs that are requested from a single FedCM get() invocation. Records at the time the API call is completed. Records some integer greater than or equal to 1. Uses exponential count bucketing. Records the number of navigator.credentials.get() requests made in a document. Requests made when FedCM is disabled or when there is a pending FedCM request are not counted. Records at most one sample for every document where at least one navigator.credentials.get() request is made. No samples are recorded if the count is 0. Records at the time the API call is completed. The bucketing method used is base::UmaHistogramCounts100. Records the RP mode of the FedCM API. Recorded once per API call together with Status.RequestIdToken when the request is fulfilled (either resolved or rejected). Records at the time the API call is completed. Whether the accounts dialog is shown on an RP URL which has a path. Recorded each time the user is shown an accounts dialog. Records at the time the API call is completed. Records the status of a disconnect call to the FedCM API. Recorded once per IdentityCredential.disconnect() call, once the browser knows the result. Records the mediation requirement of a request token call to the FedCM API. Records a sample when the call is completed, i.e. at the same time as Status.RequestIdToken. Records the status of a request id token call to the FedCM API. Recorded once per API call when the request is fulfilled (either resolved or rejected). Records whether user sign-in states between IDP and browser match after browser receives an account list from IDP prior to showing UI. Records at the time the API call is completed. Records the status of third party cookies when a FedCM API call is invoked. Recorded at most once per API call. i.e. at the same time as Blink.FedCm.Status.RequestIdToken. Records the duration from when an accounts dialog is shown to when dialog is destroyed e.g. user selects an account, user closes the dialog, or the developer aborts the request. This duration includes the time when the accounts dialog is inactive e.g. user switches to a different tab. Samples are at most 10 minutes. Records a sample when an accounts dialog which is not the verifying UI for auto re-authentication is shown. Records at the time the API call is completed. Records the time (in milliseconds) from when the accounts dialog is shown to when the user closes the dialog without selecting any account. Records at the time the API call is completed. Records the time (in milliseconds) from when the accounts dialog is shown to when the user presses the Continue button. Records at the time the API call is completed. Records the time in milliseconds from when the disconnect() request is received by the browser process to the time in which it is completed, e.g. right before sending the response back to the renderer process. Records one sample for every disconnect call which sends a disconnect request. Records at the time the API call is completed. Records the time (in milliseconds) from when the user presses the Continue button to when the id token response is received. Records at the time the API call is completed. Records the duration from when a mismatch dialog is shown to when user triggers IDP sign-in pop-up window or when dialog is destroyed e.g. user closes the dialog. This duration includes the time when the mismatch dialog is inactive e.g. user switches to a different tab. Samples are at most 10 minutes. Records a sample when a mismatch dialog is shown, at the time the API call is completed. Records the time (in milliseconds) from when a call to the API was made to when the accounts dialog is shown. Records at the time the API call is completed. Records the time (in milliseconds) from when the accounts request was made to when the fetch response is received. Only recorded when the accounts are ready to show which means all fetches are done. In case of multiple identity providers, the fetch that takes the longest will be recorded. Records at the time the API call is completed. Records the time (in milliseconds) from when the client metadata request was made to when the fetch response is received. Only recorded when the accounts are ready to show which means all fetches are done. In case of multiple identity providers, the fetch that takes the longest will be recorded. Records at the time the API call is completed. Records the time (in milliseconds) from when a call to the API was made to when the well-known and config files fetch are received. Only recorded when the accounts are ready to show which means all fetches are done. In case of multiple identity providers, the fetch that takes the longest will be recorded. Records at the time the API call is completed. Records the overall time (in milliseconds) from when the API is called to when the id token response is received. Records at the time the API call is completed. Records the outcome of the user clicking on the "use other account" button in FedCM. Recorded once per API call when the user attempts to login with a new account via the "use other account" button, when the request is fulfilled (either resolved or rejected). Only considers the last "use other account" button clicked if there are multiple attempts. Records at the time the API call is completed. Records the outcome of the verifying dialog. Recorded once per verifying dialog shown when the FedCM call is completed, i.e. at the same time as Blink.FedCm.Status.RequestIdToken. Records a 1 whenever a FedCM API call includes a well-known file without `accounts` or `login_url` but where the config file has `client_metadata`. Records when the config and well-known files have been fetched. npm@chromium.org web-identity-eng@google.com Records performance metrics for FedCM (Federated Credential Management) API. Recorded when a FedCmMetrics object is destroyed, typically when cleaning up a RequestImpl object which is created and associated with a call to the FedCM API. This event uses the specific WEB_IDENTITY ukm_source_id which is generated from the identity provider's origin. Records the number of times an accounts dialog is shown per API call. Records at the time the API call is completed. This is in units of dialogs. Records the number of requests sent to the accounts endpoint per API call. This is in units of requests. Records the type of frame that invokes disconnect(). Records once for each disconnect() call. Possible values are: main frame, same-site iframe, and cross-site iframe, where the site of the caller frame is compared with the site of the main frame. Records at the time the API call is completed. Records the outcome of the error dialog. Recorded at most once per API call when the error dialog is displayed. Records at the time the API call is completed. Records the type of error dialog shown. Recorded at most once per API call when the error dialog is displayed. Records at the time the API call is completed. Records whether the error URL is same-origin, cross-origin but same-site or cross-site with the config URL. Recorded at most once per API call when a valid error URL is received. Records at the time the API call is completed. Records the type of token response received. Recorded at most once per API call when a token response received. Records at the time the API call is completed. Records the session ID associated to the FedCM call for which this event is recording metrics. Each FedCM call gets a random integer session id, which helps group UKM events by the session id. Records at the time the API call is completed. Records the type of frame that invokes the FedCM API. Records once per API call. Possible values are: main frame, same-site iframe, and cross-site iframe, where the site of the caller frame is compared with the site of the main frame. Same-origin iframes are recorded under the same-site category. Records at the time the API call is completed. Records a 1 whenever a FedCM API call includes the nonce parameter. Records towards the beginning of the call, except for some cases where the call is immediately rejected. Records a 1 whenever a FedCM API call includes the nonce parameter, but does not include it inside the params JSON blob. Records towards the beginning of the call, except for some cases where the call is immediately rejected. Records whether a user is signing up with the API. Records once per successful account fetch from the IdP server after the accounts filtering. If there are multiple accounts, record the user as sign-in if IdP claims at least one of the accounts is sign-in. If an IdP does not share the user's sign-in state, record the state stored in the browser. Records at the time the API call is completed. Records the number of times a mismatch dialog is shown per API call. Records at the time the API call is completed. This is in units of dialogs. Records the RpMode of two consecutive requests. Recorded every time when a token request is invoked while there is another one pending. Records the number of IDPs that result in a 'mismatch' status from a single FedCM get() invocation. Recorded every time a FedCM get() call is resolved. Records at the time the API call is completed. Records some integer 0 or greater. Uses exponential count bucketing. Records the number of IDPs that are requested from a single FedCM get() invocation. Recorded every time a FedCM get() call is resolved. Records some integer greater than or equal to 1. Uses exponential count bucketing. Records the RP mode of the FedCM API. Recorded once per API call together with Status.RequestIdToken when the request is fulfilled (either resolved or rejected). Records the status of a disconnect call to the FedCM API. Recorded once per IdentityCredential.disconnect() call, once the browser knows the result. Records the mediation requirement of a request token call to the FedCM API. Records a sample when the call is completed, i.e. at the same time as Status.RequestIdToken. Records the status of a request id token call to the FedCM API. Recorded once per API call when the request is fulfilled (either resolved or rejected). Records whether user sign-in states between IDP and browser match after browser receives an account list from IDP prior to showing UI. Records at the time the API call is completed. Records the status of third party cookies when a FedCM API call is invoked. Recorded at most once per API call. i.e. at the same time as Blink.FedCm.Status.RequestIdToken. Records the duration from when an accounts dialog is shown to when dialog is destroyed e.g. user selects an account, user closes the dialog, or the developer aborts the request. This duration includes the time when the accounts dialog is inactive e.g. user switches to a different tab. Samples are at most 10 minutes. Records a sample when an accounts dialog which is not the verifying UI for auto re-authentication is shown. Records at the time the API call is completed. Records the time (in milliseconds) from when the accounts dialog is shown to when the user closes the dialog without selecting any account. Records at the time the API call is completed. Records the time (in milliseconds) from when the accounts dialog is shown to when the user presses the Continue button. Records at the time the API call is completed. Records the time in milliseconds from when the disconnect() request is received by the browser process to the time in which it is completed, e.g. right before sending the response back to the renderer process. Records one sample for every disconnect call which sends a disconnect request. Records at the time the API call is completed. Records the time (in milliseconds) from when the user presses the Continue button to when the id token response is received. Records at the time the API call is completed. Records the duration from when a mismatch dialog is shown to when user triggers IDP sign-in pop-up window or when dialog is destroyed e.g. user closes the dialog. This duration includes the time when the mismatch dialog is inactive e.g. user switches to a different tab. Samples are at most 10 minutes. Records a sample when a mismatch dialog is shown. Records at the time the API call is completed. Records the time (in milliseconds) from when a call to the API was made to when the accounts dialog is shown. Records at the time the API call is completed. Records the overall time (in milliseconds) from when the API is called to when the id token response is received. Records at the time the API call is completed. Records the outcome of the user clicking on the "use other account" button in FedCM. Recorded once per API call when the user attempts to login with a new account via the "use other account" button, when the request is fulfilled (either resolved or rejected). Only considers the last "use other account" button clicked if there are multiple attempts. Records at the time the API call is completed. Records the outcome of the verifying dialog. Recorded once per verifying dialog shown when the FedCM call is completed, i.e. at the same time as Blink.FedCm.Status.RequestIdToken. Records a 1 whenever a FedCM API call includes a well-known file without `accounts` or `login_url` but where the config file has `client_metadata`. Records when the config and well-known files have been fetched. vmpstr@chromium.org chrishtr@chromium.org paint-dev@chromium.org Logged once per page; indicates the type of find-in-page behaviors users utilizied. Boolean indicating that a find-in-page match was hidden but became visible after the beforematch event was fired. Boolean indicating whether there was an active match in content-visibility unrendered elements. Boolean indicating that find-in-page happened. nidhijaju@chromium.org chrome-loading@google.com Timing metrics associated with the frame loader in Blink. Time taken to commit a new document in Blink in microseconds. Collected once per navigation when a frame is loaded. This is rounded down to the nearest exponential bucket. dom-dev@google.com nidhijaju@chromium.org chrome-loading@google.com Logged once per page load, to capture information about invocation of the HTMLDocumentParser for the initial document load. Does not include other aspects of document parsing, such as style sheet parsing or script execution during parsing. Also does not include innerText parsing. This event is not recorded on machines with low-resolution clocks. The number of chunks parsed while loading the page. Each chunk represents a piece of the HTML content that can be loaded without blocking or yielding. Time taken to fetch queued subresource preloads for a document in microseconds. This is rounded down to the nearest exponential bucket. The maximum time used to parse a single chunk before being blocked, yielding or completing, in microseconds. The minimum time used to parse a single chunk before being blocked, yielding or completing, in microseconds. The total amount of time spent in the HTMLDocumentParser parsing the page, in microseconds. Time spent on preloading subresources for a document, including mojo operations, in microseconds. This is rounded down to the nearest exponential bucket. Time taken to tokenize the HTML document, if necessary, and perform other tasks around parsing completion in microseconds. This is rounded down to the nearest exponential bucket. Time taken to tokenize an HTML document in Blink in microseconds. This is rounded down to the nearest exponential bucket. Time taken to scan and preload subresources for a document in microseconds. This is rounded down to the nearest exponential bucket. Time spent on the preload scanning process for an HTML document, in microseconds. This is rounded down to the nearest exponential bucket. The average number of tokens parsed across all chunks. The maximum number of tokens parsed as a chunk. The minimum number of tokens parsed as a chunk. The total number of tokens parsed across all chunks. The average time between parsing of two chunks, in microseconds. Will be zero if only one chunk was parsed. The minimum time between parsing of two chunks, in microseconds. Will be zero if only one chunk was parsed. The maximum time between parsing of two chunks, in microseconds. Will be zero if only one chunk was parsed. houssein@chromium.org nrosenthal@chromium.org Metrics stating the detected major+minor version of known frameworks when detected. This event is recorded when the page is loaded. Note that a page can use multiple frameworks at the same time, hence why this reports each framework individually instead of a single number. The version numbers here are a bit mask: 0xMMmm (Major+minor), e.g. version 16.3 would be 0x1003 A bitmask (0xMMmm) representing the Angular version A bitmask (0xMMmm) representing the NextJS version A bitmask (0xMMmm) representing the Nuxt version A bitmask (0xMMmm) representing the Vue version pdr@chromium.org paint-dev@chromium.org The aggregated time spent in each metric's component from creation of the local frame view to the First Contentful Paint signal. The event is reported once per page load when First Contentful Paint is signaled. This event is not recorded on machines with low-resolution clocks. The time spent in accessibility tree code, between navigation and First Contentful Paint, in microseconds. Time spent computing IntersectionObservations for anchor elements, in microseconds. The time spent in main frame animation, between navigation and First Contentful Paint, in microseconds. The time spent committing main thread compositing results, between navigation and First Contentful Paint, in microseconds. The time spent in main thread compositing inputs, between navigation and First Contentful Paint, in microseconds. The time spent doing forced document updates due to update sub-document content, between navigation and First Contentful Paint, in microseconds. Includes SVG, Canvas and Plugins. Time spent computing IntersectionObservations for DisplayLockDocumentState, in microseconds. The time spent in forced style recalc and layouts, between navigation and First Contentful Paint, in microseconds. The time spent in processing rAF-aligned input for the main frame, between navigation and First Contentful Paint, in microseconds. The time spent updating the document in preparation for a hit test for the main frame, between navigation and First Contentful Paint, in microseconds. The time spent committing layer tree impl data in the compositor thread, between navigation and First Contentful Paint, in microseconds. The time spent computing intersection observer state, between navigation and First Contentful Paint, in microseconds. The total number of times a blink-internal IntersectionObservation was recomputed between navigation and First Contentful Paint. The total number of times a javascript-instantiated IntersectionObservation was recomputed between navigation and First Contentful Paint. The time spent doing forced document updates due to generic Javascript API requests, between navigation and First Contentful Paint, in microseconds. Note this primarily captures Javascript that explicitly requires Style and Layout, but not implicit requests such as through setting focus or scroll position. Time spent computing IntersectionObservations for javascript observers, in microseconds. The time spent in layout document lifecycle work, between navigation and First Contentful Paint, in microseconds. Time spent computing IntersectionObservations for lazy loading, in microseconds. The time spent in ProxyMain::BeginMainFrame, between navigation and First Contentful Paint, in microseconds. Time spent computing IntersectionObservations for media elements, in microseconds. The time spent in paint document lifecycle work, between navigation and First Contentful Paint, in microseconds. The time spent parsing style sheets. Time spent computing IntersectionObservations for permission elements, in microseconds. Counts whether a possible sync-scroll update is attempted during an animation update of the outermost main frame of a page between navigation and First Contentful Paint (see sync_scroll_attempt_heuristic.h for details). The heuristic is checked once per main frame update. The recorded value will be either 1 or 0 for a sampled frame depending on whether the heuristic determined that there was a possible sync-scroll update. The time spent in pre-paint document lifecycle work, between navigation and First Contentful Paint, in microseconds. The time spent doing forced document updates due to browser requested actions, such as accessibility, intersection observer or spell checking, between navigation and First Contentful Paint, in microseconds. The time spent in style document lifecycle work, between navigation and First Contentful Paint, in microseconds. The time spent in for LayerTreeHost::UpdateLayers, between navigation and First Contentful Paint, in microseconds. Time spent computing internal geometry information about iframes, between navigation and First Contentful Paint, in microseconds. The time spent doing forced document updates due to a user action such as input, find-in-page or printing, between navigation and First Contentful Paint, in microseconds. As of 04/2023, updates for scrolling are included in this metric. Time elapsed, in microseconds, between when an animation update is first scheduled and when the resulting BeginMainFrame starts. The time spent waiting for the compositor thread to begin processing a commit, between navigation and First Contentful Paint, in microseconds. npm@chromium.org sullivan@chromium.org Paint timing metrics recorded in Blink, usually used as debugging metrics for PaintTiming metrics in the PageLoad event. Records whether a page has an image whose size equals the full viewport. Recorded when the LCP algorithm has stopped. tnak@chromium.org chrome-loading@google.com Events taken when HTTP responses are received and they have Link headers. Multiple events should be involved if one response contains multiple Link headers. This event provides insights into potential information leakage related to Link headers on subresources. Records the cross-origin usage related to Link headers on subresource fetches. This is only recorded for subresource fetches. dom@chromium.org chrome-loading@google.com Metrics associated with the async script count associated with a document, recorded when the HTML parsing of a mainframe has completed. pdr@chromium.org paint-dev@chromium.org Metrics that measure various update times in Blink. This includes paint, compositing, and layout update times. This event is reported once in the first main frame update after First Contentful Paint, with data from a randomly chosen frame up to and including First Contentful Paint, and once on LocalFrameView destruction with data from a random frame between FirstContentfulPaint and shutdown. This event is not recorded on machines with low-resolution clocks. The time taken for accessibility for the main frame in microseconds during the sampled frame. The total main frame time used in accessibility, in microseconds. Time spent computing IntersectionObservations for anchor elements, in microseconds. The total main frame time used to compute IntersectionObservations for anchor elements, in microseconds. The time taken for main frame animation in microseconds during the sampled frame. The total main frame time used by by main thread animations, in microseconds. The time taken by the compositing commit in microseconds during the sampled frame. The total main frame time used by the compositing commit, in microseconds. The time taken by the compositing inputs phase in microseconds during the sampled frame. The total main frame time used by the compositing inputs phase, in microseconds. The time spent doing forced document updates due to update sub-document content, during a sampled frame, in microseconds. Includes SVG, Canvas and Plugins. The time spent doing forced document updates due to update sub-document content, during the BeginMainFrame block of a sampled frame, in microseconds. Includes SVG, Canvas and Plugins. Time spent computing IntersectionObservations for DisplayLockDocumentState, in microseconds. The total main frame time used to compute IntersectionObservations for DisplayLockDocumentState, in microseconds. The time taken by forced style recalc and layouts in microseconds between the end of the last main frame update and the current sampled frame, in microseconds. The total main frame time used by forced style updates and layouts, in microseconds. This metric may include time that is also included in the AnimateBeginMainFrame metric. The time taken to process rAF-aligned input for the main frame in microseconds during the sampled frame. The total main frame time used in processing rAF-aligned input, in microseconds. The time taken to perform hit test document updates for the main frame in microseconds during the sampled frame. It does not compute the actual hit test time, rather the style, layout, compositing, etc time to prepare for the test. The total main frame time used in performing hit test document updates, in microseconds. The time spent committing layer tree impl data in the compositor thread, in microseconds. The total main frame time spent committing layer tree impl data in the compositor thread, in microseconds. This should always match ImplCompositorCommit and serves as a check. The time taken to compute intersection observer state for the main frame in microseconds during the sampled frame. The total main frame time used to compute IntersectionObserver observations, in microseconds. The number of blink-internal IntersectionObservations that were recomputed during the sampled frame. The number of blink-internal IntersectionObservations that were recomputed during the sampled frame. The number of javascript-instantiated IntersectionObservations that were recomputed during the sampled frame. The number of javascript-instantiated IntersectionObservations that were recomputed during the sampled frame. The time spent doing forced document updates due to generic Javascript API requests, during a sampled frame, in microseconds. Note this primarily captures Javascript that explicitly requires Style and Layout, but not implicit requests such as through setting focus or scroll position. The time spent doing forced document updates due to generic Javascript API requests, during the BeginMainFrame block of a sampled frame, in microseconds. Note this primarily captures Javascript that explicitly requires Style and Layout, but not implicit requests such as through setting focus or scroll position. Time spent computing IntersectionObservations for javascript observers, in microseconds. The total main frame time used to compute IntersectionObservations for Javascript observers, in microseconds. The time taken for layout for the main frame in microseconds during the sampled frame. The total main frame time used in layout, in microseconds. Time spent computing IntersectionObservations for lazy loading, in microseconds. The total main frame time used to compute IntersectionObservations for lazy loading, in microseconds. The time between a BeginMainFrame and the commit of paint results for the sampled frame, in microseconds. True if the lifecycle update in this event occurred before FCP (First Contentful Paint) was registered. A bitfield representation of the reasons for a main frame update. The fields are based on the FrameSequenceTrackerType enum, so bitwise and with the following values to determine if a reason was present: CompositorAnimation: 1, MainThreadAnimation: 2, PinchZoom: 4, RAF: 8, TouchScroll: 16, Universal: 32, Video: 64 and WheelScroll: 128. Time spent computing IntersectionObservations for media elements, in microseconds. The total main frame time used to compute IntersectionObservations for media elements, in microseconds. The time taken to paint for the main frame in microseconds during the sampled frame. The total main frame time used in paint, in microseconds. The time spent parsing style sheets. The total main frame time spent parsing style sheets. Time spent computing IntersectionObservations for permission elements, in microseconds. The total main frame time used to compute IntersectionObservations for permission elements, in microseconds. Counts whether a possible sync-scroll update is attempted during an animation update of the outermost main frame of a page (see sync_scroll_attempt_heuristic.h for details). The heuristic is checked once per main frame update. The recorded value will be either 1 or 0 for a sampled frame depending on whether the heuristic determined that there was a possible sync-scroll update. Counts whether a possible sync-scroll update is attempted during an animation update of the outermost main frame of a page (see sync_scroll_attempt_heuristic.h for details). The heuristic is checked once per main frame update. The recorded value will be either 1 or 0 for a sampled frame depending on whether the heuristic determined that there was a possible sync-scroll update. Note that since the check happens during main frame updates, there should be no difference between this metric and PossibleSynchronizedScrollCount2 above. The time taken for pre-paint for the main frame in microseconds during the sampled frame. The total main frame time used in pre-paint, in microseconds. The time spent doing forced document updates due to browser requested actions, such as accessibility, intersection observer or spell checking, during a sampled frame, in microseconds. The time spent doing forced document updates due to browser requested actions, such as accessibility, intersection observer or spell checking, during the BeginMainFrame block of a sampled frame, in microseconds. The time taken for style update for the main frame in microseconds during the sampled frame. The total main frame time spent updating style information and building the layout tree, in microseconds. The percentage of the main frame time used in style work. An int in the range [0,100]. The time taken for LayerTreeHost::UpdateLayers in microseconds during the sampled frame The total main frame time spent in LayerTreeHost::UpdateLayer, in microseconds. This should always match UpdateLayers and serves as a check. Time spent computing internal geometry information about iframes, in microseconds. The total main frame time used to compute internal geometry information about iframes, in microseconds. The time spent doing forced document updates due to a user action such as input, find-in-page or printing, during a sampled frame, in microseconds. As of 04/2023, updates for scrolling are included in this metric. The time spent doing forced document updates due to a user action such as input, find-in-page or printing, during the BeginMainFrame block of a sampled frame, in microseconds. As of 04/2023, updates for scrolling are included in this metric. Time elapsed between when an animation update is first scheduled and when the resulting BeginMainFrame starts. Time elapsed between when an animation update is first scheduled and when the resulting BeginMainFrame starts. The time spent waiting for the compositor thread to begin processing a commit, in microseconds. The total main frame time spent waiting for the compositor thread to begin processing a commit, in microseconds. This should always match WaitForCommit and serves as a check. chasej@chromium.org Collects data for a subset of UseCounter features where UMA UseCounter data shows a behaviour that is rare but too common to blindly change. This metric should be used to reason about whether a breaking change is acceptable or not. Note: In M87, a number of use counters were removed from UKM reporting due to being reported too frequently (high proportion across page loads). Opt-in UseCounter feature. Emits True(1) or False(0) to indicate whether the Feature recorded is in the main frame or not. yaoxia@chromium.org jkarlin@chromium.org The document.browsingTopics() result and other metadata. Recorded whenever document.browsingTopics() is invoked from a context. If there are more than 3 candidate topics (e.g. due to custom configuration FeatureParam), only the first 3 candidate topics will be recorded. The candidate topic ID at index 0. This metric won't be recorded if no topics were derived from the candidate epochs (which can be due to insufficient epochs, failed epoch topics calculation, cleared history, or cleared/blocked individual topics). Whether `CandidateTopic0` is a true top topic (as opposed to a padded or random one). This metric will be recorded iff `CandidateTopic0` is recorded. The version of the model used to calculate `CandidateTopic0`. This metric will be recorded iff `CandidateTopic0` is recorded. Whether `CandidateTopic0` should be filtered. This metric will be recorded iff `CandidateTopic0` is recorded. The version of the taxonomy applicable to `CandidateTopic0`. This metric will be recorded iff `CandidateTopic0` is recorded. The candidate topic ID at index 1. This metric won't be recorded if less than 2 topics were derived from the candidate epochs (which can be due to insufficient epochs, failed epoch topics calculation, cleared history, or cleared/blocked individual topics). Whether `CandidateTopic1` is a true top topic (as opposed to a padded or random one). This metric will be recorded iff `CandidateTopic1` is recorded. The version of the model used to calculate `CandidateTopic1`. This metric will be recorded iff `CandidateTopic1` is recorded. Whether `CandidateTopic1` should be filtered. This metric will be recorded iff `CandidateTopic1` is recorded. The version of the taxonomy applicable to `CandidateTopic1`. This metric will be recorded iff `CandidateTopic1` is recorded. The candidate topic ID at index 2. This metric won't be recorded if less than 3 topics were derived from the candidate epochs (which can be due to insufficient epochs, failed epoch topics calculation, cleared history, or cleared/blocked individual topics). Whether `CandidateTopic2` is a true top topic (as opposed to a padded or random one). This metric will be recorded iff `CandidateTopic2` is recorded. The version of the model used to calculate `CandidateTopic2`. This metric will be recorded iff `CandidateTopic2` is recorded. Whether `CandidateTopic2` should be filtered. This metric will be recorded iff `CandidateTopic2` is recorded. The version of the taxonomy applicable to `CandidateTopic2`. This metric will be recorded iff `CandidateTopic2` is recorded. The reason why the API access is not allowed. yaoxia@chromium.org jkarlin@chromium.org The epoch topics calculation result, including the top 5 topics and other metadata. Recorded whenever the periodic (weekly) topics calculation has finished. If the result has more than 5 topics (e.g. due to custom configuration FeatureParam), only the first 5 topics will be recorded. The version of the model used to calculate the result topics. This metric won't be recorded if the topics calculation failed in general. The start index of the randomly padded top topics. This metric won't be recorded if the topics calculation failed in general. If all topics are real, `PaddedTopicsStartIndex` will equal the feature param `kBrowsingTopicsNumberOfTopTopicsPerEpoch`. The version of the taxonomy applicable to the result topics. This metric won't be recorded if the topics calculation failed in general. The top topic ID at index 0. This metric won't be recorded if the candidate topic was blocked, or if the topics calculation failed in general. The top topic ID at index 1. This metric won't be recorded if the candidate topic was blocked, or if the topics calculation failed in general. The top topic ID at index 2. This metric won't be recorded if the candidate topic was blocked, or if the topics calculation failed in general. The top topic ID at index 3. This metric won't be recorded if the candidate topic was blocked, or if the topics calculation failed in general. The top topic ID at index 4. This metric won't be recorded if the candidate topic was blocked, or if the topics calculation failed in general. yaoxia@chromium.org jkarlin@chromium.org PageLoad metrics for BrowsingTopics. Recorded when a page has Topics API usages (that pass the user permission check), and the page is being destroyed/navigated. The number of unique context domains have requested the topics. Rounded to the nearest exponential bucket (with a bucket ratio of 1.15). navtracking-team@google.com dc-komics@google.com Emitted for the page visited at the beginning of a redirect chain (i.e. the initial URL), when third-party cookies are disabled. The redirect chain is used as a part of Bounce Tracking Mitigations (BTM). A randomly-generated 32-bit id to connect this event with the BTM.Redirect and BTM.ChainEnd events emitted within the same redirect chain. True when the site before the redirect chain is the same as the site at the end of the chain. navtracking-team@google.com dc-komics@google.com Emitted for the page visited at the end of a redirect chain (i.e. the final URL), when third-party cookies are disabled. The redirect chain is used as a part of Bounce Tracking Mitigations (BTM). A randomly-generated 32-bit id to connect this event with the BTM.ChainBegin and BTM.Redirect events emitted within the same redirect chain. True when the site before the redirect chain is the same as the site at the end of the chain. navtracking-team@google.com dc-komics@google.com Emitted on every HTTP redirect or client-side bounce within a redirect chain, when third-party cookies are disabled. A client-side bounce is a non-user-initiated navigation starting soon after the previous navigation committed. The redirect chain is used as a part of Bounce Tracking Mitigations (BTM). A randomly-generated 32-bit id to connect this event with other BTM.Redirect events and the BTM.ChainBegin/ChainEnd events emitted within the same redirect chain. Populated for client redirects only. The time in seconds (capped at 10) between the page committing and the redirect navigation starting. An int in the range [0, 10]. Whether the redirect read cookies, wrote cookies, both, or neither. Populated for client redirects only. Whether the user interacted with the page (See the HTML spec for a definition). This info is persisted within BTM for a duration of `btm::kInteractionTtl`. True when the site before the redirect chain is the same as the site at the end of the chain. True if the chain is missing the end URL. Any comparisons to the final site are invalid for this redirect. True when the redirect site is the same as the site at the end of the redirect chain. True when the redirect site is the same as the site before the redirect chain. 0-based index of this individual redirect within its chain. The length of the overall redirect chain. The type of the redirect: client or server. The site engagement level of the redirect URL, at the time the redirect took place. Whether the site has previously had a user activation. Whether the site has previously had a WebAuthn assertion. Populated for client redirects only. Whether the user triggered a successful web authn assertion. This info is persisted within BTM for a duration of `btm::kInteractionTtl`. thanhdng@chromium.org Records history of charge events including start time and duration of charges. Can be used with other ChargeEventHistory to track the last 30 days of charging events. If a "SmartCharging" event and a "ChargeEventHistory" event has them same source ID and "EventId", it means "ChargeEventHistory" is the charge history when "SmartCharging" event is logged. ChargeEventHistory records an array. Each ChargeEventHistory event records an item in the array, specified by its index ChargeEventHistoryIndex. Items in the same array will have same source ID and EventId. Duration of a history charge event. Index of this event in the ChargeEventHistory array. Please see the event summary for more details. Number of items in ChargeEventHistory array. Please see the event summary for more details. Duration (in minutes) between start time of a charge event and logging time. A unique number that represent order of an event. Used combined with "SmartCharging"'s EventId to reconstruct charge history. siakabaro@google.com Records the ChromeOS device management state. This event is recorded once per UKM report and is not associated with a source URL. An enum value representing the enrollment status of a ChromeOS device. nancylingwang@chromium.org dominickn@chromium.org Records the amount of time that the focused and activated website was accessed in last two hours on Chrome OS. This is logged each two hours, and only when the website is not installed as a PWA or Chrome app. The milliseconds that the website was accessed in the focused tab and the activated browser window. The accessing time is calculated each two hours, then the noise is applied, and sum the running time for two hours with noise to calculate the two hours accessing time. For example the running time in the past 2 hours = time1 * noise1 + time2 * noise2 + time3 * noise3.... Whether the metrics is read from the user pref on disk, recorded for the last login. Whether the website can be an installable web app. An enum that records the user type and the device type. nancylingwang@chromium.org dominickn@chromium.org Records each two hours in Chrome OS to log the input events happened on the app window. This is logged only when the user allows app syncing or allows sync of everything, and the user does not set up a sync passphrase. The number of the input events happened in each two hours. An enum that records the input event source. An enum that records the app type. An enum that records the user type and the device type. nancylingwang@chromium.org dominickn@chromium.org Records the installed app in Chrome OS, only when the user allows app syncing or allows sync of everything, and the user does not set up a sync passphrase. Only ARC apps, Chrome apps and web apps(PWA), system web apps, buildin apps are recorded. This is logged when the user installs a new app. An enum that records the app type. An enum that records the reason why the app was installed. An enum that records where the app was installed. An enum that records when the app was installed. An enum that records the user type and the device type. nancylingwang@chromium.org dominickn@chromium.org Records a launched app in Chrome OS, only when the user allows app syncing or allows sync of everything, and the user does not set up a sync passphrase. Only ARC apps, Chrome apps and web apps(PWA), system web apps, buildin apps are recorded. This is logged when an app is launched. An enum that records the app type. An enum that records where the app was launched. An enum that records the user type and the device type. nancylingwang@chromium.org dominickn@chromium.org Records the uninstalled app in Chrome OS, only when the user allows app syncing or allows sync of everything, and the user does not set up a sync passphrase. Only ARC apps, Chrome apps and web apps(PWA), system web apps, buildin apps are recorded. This is logged when the user uninstalls an app. An enum that records the app type. An enum that records how the app was uninstalled. An enum that records the user type and the device type. nancylingwang@chromium.org dominickn@chromium.org Records the amount of time that the app was used in the last two hours on Chrome OS, only when the user allows app syncing or allows sync of everything, and the user does not set up a sync passphrase. Only ARC apps, Chrome apps and web apps(PWA), system web apps, builtin apps are recorded. This is logged every two hours and at shutdown. An enum that records the app type. The milliseconds that the app was used. An enum that records the user type and the device type. nancylingwang@chromium.org dominickn@chromium.org Records the amount of time that the app were used in the last two hours on Chrome OS, only when the user allows app syncing or allows sync of everything, and the user does not set up a sync passphrase. Only ARC apps, Chrome apps and web apps(PWA), system web apps, buildin apps are recorded. This is logged every two hours. An enum that records the app type. The milliseconds that the app was used. An enum that records the user type and the device type. dougarnett@chromium.org Metrics associated with user click input. In particular, captures when user has a burst of clicks at the same location which may be a signal of user unhappiness for an experiment if it differs from the baseline data for the site. The count of the largest burst of click inputs for a page load. Only recorded if it meets a minimum threshold to be considered a burst. arichiv@chromium.org victortan@chromium.org Record when a Client Hint is requested in an ACCEPT_CH TLS Frame. Records an enum representing the specific Client Hint type that was requested. arichiv@chromium.org victortan@chromium.org Record when a Client Hint is requested in a Accept-CH HTTP Response Header. Records an enum representing the specific Client Hint type that was requested. arichiv@chromium.org victortan@chromium.org Record when a Client Hint is requested in a Accept-CH HTML Meta Tag. Records an enum representing the specific Client Hint type that was requested. arichiv@chromium.org victortan@chromium.org Record when a Client Hint is requested in a Critical-CH HTTP Response Header. Records an enum representing the specific Client Hint type that was requested. arichiv@chromium.org victortan@chromium.org Record when a Client Hint is requested in a Delegate-CH HTML Meta Tag. Records an enum representing the specific Client Hint type that was requested. aaronhk@chromium.org cwallez@chromium.org enga@chromium.org fserb@chromium.org jdarpinian@chromium.org kainino@chromium.org kbr@chromium.org shrekshao@google.com Metrics collecting types of rendering context created with canvas/offscreencanvas. It also collects requests of GPU device via WebGPU API. Type of rendering contexts created for HTMLCanvasElement. Type of rendering contexts created for HTMLCanvasElement that something has actually been drawn to. For WebGL 1 and 2 context this means there is at least one draw call issued. For other context types we leave the implementation blank at this moment. Event fired when a GPUDevice from the WebGPU API is created. Type of rendering contexts created for OffscreenCanvas. Type of rendering contexts created for OffscreenCanvas that something has actually been drawn to. For WebGL 1 and 2 context this means there is at least one draw call issued. For other context types we leave the implementation blank at this moment. picarella@google.com mmaryia@google.com As of M153, recorded when either a copy/cut or a subsequent paste event occurs for a user with the AutofillAtMemory feature disabled (prior to M153, it was recorded regardless of the feature status). A separate event entry is logged for each action: a copy or cut from a tab records an entry with `CopyFromTabNonce`, while a subsequent paste of that data from a different tab and within 10 minutes, records a separate entry with `PasteFromTabNonce`. A random 64-bit integer recorded when a user copies or cuts data from a non-incognito tab. The 64-bit nonce of the matching copy event, recorded when the user pastes the copied data into a different tab within 10 minutes. Used to link the copy and paste events. picarella@google.com mmaryia@google.com Added in M153. Recorded when either a copy/cut or a subsequent paste event occurs for a user with the AutofillAtMemory feature enabled. A separate event entry is logged for each action: a copy or cut from a tab records an entry with `CopyFromTabNonce`, while a subsequent paste of that data from a different tab and within 10 minutes, records a separate entry with `PasteFromTabNonce`. A random 64-bit integer recorded when a user copies or cuts data from a non-incognito tab. The 64-bit nonce of the matching copy event, recorded when the user pastes the copied data into a different tab within 10 minutes. Used to link the copy and paste events. shaktisahu@chromium.org Records metrics for the companion side panel. The events correspond to a single page load in the main frame and the metrics are flushed when the page navigates away. For most of the UI surfaces, only the last event (e.g. displayed or clicked) is recorded. The position of the ATX surface in the page. The last event on the ATX surface. The position of the CQ element that was clicked with respect to its parent list. The position of the CQ surface in the page. The last event on the CQ surface. The number of child elements available within the CQ surface after eligibility check after text finder API. Clamped to a max value of 10. The number of child elements actually shown within the CQ surface after eligibility check through text finder API. Clamped to a max value of 10. The UI location from which the companion page was opened. The position of the page entities element that was clicked with respect to its parent list. The position of the page entities surface in the page. The last event on the page entities surface. The number of child elements available within the page entities surface. Clamped to a max value of 10. The number of child elements shown within the page entities surface. Clamped to a max value of 10. The position of the PH surface in the page. The last user feedback event on the PH surface. The last event on the PH surface. The position of the PH element that was clicked with respect to its parent list. The position of the PH surface in the page. The last event on the PH surface. The number of child elements available within the PH surface, regardless of text finder API results. Clamped to a max value of 10. The number of child elements within the PH surface that pass text finder API elibigility check, this is the count bullet points that have jump tags. Clamped to a max value of 10. The last event on the promo surface. The number of times region search in the multimodal box was clicked. The position of the RelQr element that was clicked with respect to its parent list. The position of the RelQr surface in the page. The last event on the RelQr surface. The number of child elements available within the RelQr surface. Clamped to a max value of 10. The number of child elements shown within the RelQr surface. Clamped to a max value of 10. The position of the RelQs element that was clicked with respect to its parent list. The position of the RelQs surface in the page. The last event on the RelQs surface. The number of child elements available within the RelQs surface. Clamped to a max value of 10. The number of child elements shown within the RelQs surface. Clamped to a max value of 10. The number of times text search box was used, clamped to a max of 10. The position of the VQ element that was clicked with respect to its parent list. The position of the VQ surface in the page. The last event on the VQ surface. The number of child elements available within the VQ surface. Clamped to a max value of 10. The number of child elements shown within the VQ surface. Clamped to a max value of 10. The number of images classified as sensitive. Uses exponential bucket of 1.3 for privacy. The number of images classified as shoppy. Uses exponential bucket of 1.3 for privacy. The number of images classified as shoppy and non-sensitive. Uses exponential bucket of 1.3 for privacy. The number of images eligible for visual classification. Uses exponential bucket of 1.3 for privacy. The number of images that qualify for visual suggestions. Uses exponential bucket of 1.3 for privacy. chrome-compose-frontend@google.com dewittj@chromium.org carlosk@chromium.org Reported at PrimaryPageChanged (or tab destruction) if any metrics have been recorded. This collects user actions in the compose feature (and some of its effects) for this page. The number of times the Compose result was inserted into the page. Uses exponential buckets for privacy. The number of times the Compose context menu item was clicked for an element on the page. Uses exponential buckets for privacy. The number of times the Compose context menu item was shown for an element on the page. Uses exponential buckets for privacy. The number of times the Compose context menu item was clicked but the expected dialog was not presented because Autofill could not obtain the text field's respective FormData. Uses exponential buckets for privacy. The number of times the Compose context menu item was clicked but the expected dialog was not presented because Autofill could not obtain the text field's respective FormFieldData. Uses exponential buckets for privacy. True if the proactive nudge was disabled for this site during this page load. Added in M126. True if the proactive nudge was globally disabled this page load. Added in M126. The number of times the Compose dialog is opened from the proactive nudge during this page load. Recorded everytime the nudge is opened. Added in M126. The number of times the Compose proactive nudge should show during this page load. Recorded any time Compose is asked to show the proactive nudge even if it is eventually not shown. Added in M125. The number of times the Compose proactive nudge is shown during this page load. Recorded everytime a request to show the nudge returns true. Added in M126. chrome-compose-frontend@google.com petewil@chromium.org dewittj@chromium.org Reported at the end of every compose session. There may be more than one compose session per pageload. This collects user actions in the compose dialogs for this URL. If the user canceled this session with the "X" button. This does not include dismissing the dialog by losing focus. How many times the user pressed the "shorten" button in this session. Uses exponential buckets for privacy. How many times the user pressed the "generate" button in this session. Uses exponential buckets for privacy. How many times the compose dialog was shown to the user. Uses exponential buckets for privacy. How many times the user pressed the "formal" button in this session. Uses exponential buckets for privacy. If the user inserted results back to the web page in this session, either by pressing the "insert" or "replace" button. How many times the user pressed the "elaborate" button in this session. Uses exponential buckets for privacy. How many times the user pressed the "regenerate" button in this session. Uses exponential buckets for privacy. How many times the user pressed the "shorten" button in this session. Uses exponential buckets for privacy. True if the session started with the proactive nudge rather than the context menu option. Added in M126. How many times the user pressed the "undo" button in this session. Uses exponential buckets for privacy. harringtond@google.com carlosk@chromium.org Reported once per page load per each multiline text element that received input. Excludes potentially sensitive fields including input type=password or when autofill determines the field is sensitive. The control type, such as text or textarea. See |autofill::FormControlType|. Number of seconds the user spent editing the field. Includes time between key presses, up to 5 seconds each. Bucketed using GetExponentialBucketMinForUserTiming(). The signature of the field (mod 1021). This is the hash identifier used to denote this field. See components/autofill/core/common/signatures.cc for more details. The signature of the form (mod 1021). This is the hash identifier used to denote this form. See components/autofill/core/common/signatures.cc for more details. Whether the field is detected and supported by Autofill. Total number of characters typed into the text element. Bucketed with an exponential factor of 2. This is reported as -1 for sensitive field types, like passwords or credit cards numbers. Note that this number attempts to count only typed changes, but is not always accurate. Additionally, there is a limit to the total possible length observed. Total number of space-delimited words typed into the text element. Bucketed with an exponential factor of 2. This is reported as -1 for sensitive field types like passwords or credit cards numbers. Note that this number attempts to count only typed changes, but is not always accurate. Additionally, there is a limit to the total possible length observed. khushalsagar@chromium.org Metrics related to rendering in the compositor. The number of images checker-imaged and re-rasterized on this page. khushalsagar@chromium.org Metrics related to user interaction handled by the compositor. This includes user gestures like scrolling, pinch-zoom and page-scale animations triggered in response to a double-tap event. The length of the user interaction is defined as the time the compositor is running an animation resulting from user input. The number of visible pixels per frame checkerboarded during this interaction. The percentage of visible pixels per frame checkerboarded during this interaction. This value should be between [0, 100]. The number of images checker-imaged and re-rasterized during this interaction. The number of visible tiles per frame checkerboarded during this interaction. finnur@chromium.org peter@chromium.org Measures statistics about the contacts shared in the contacts picker (when the Contacts Picker dialog is dismissed.) Measures the contact properties requested by the website (names, emails, telephone numbers). Bit 1 - telephone numbers, bit 2 - emails, bit 3 - names. Measures the total number of contacts selected in the contacts picker. Measures the percentage of contacts selected in the contacts picker. rayankans@chromium.org Collected when a new Content Index entry is registered. The type of the content registered. Corresponds to blink.mojom.ContentCategory. rayankans@chromium.org Collected when a Content Index entry is deleted by the user from the Downloads page. Always true. rayankans@chromium.org Collected when a Content Index entry is opened from the Downloads page. Whether the browser was offline when the content was opened. felixarntz@google.com npm@chromium.org Metrics stating whether the main frame of a page uses certain content management systems. This event is recorded when the page is going to be unloaded or when the app enters the background on Android. The content management system detection is done when the document is loaded (similar to when load event is fired). If a CMS is used after this then it would not be detected. For interoperability with the JavaScript framework detection above, this reports each CMS individually instead of a single enum. True if the page loaded in the main frame uses the Drupal CMS. True if the page loaded in the main frame uses the Joomloa CMS. True if the page loaded in the main frame uses the Shopify CMS. True if the page loaded in the main frame uses the Squarespace CMS. True if the page loaded in the main frame uses the Wix CMS. True if the page loaded in the main frame uses the WordPress CMS. benwgold@google.com twellington@chromium.org Emits an integer id for a selection corresponding to the action enum defined in ChromeContextMenuPopulator.java (within the ContextMenuUma class). benwgold@google.com twellington@chromium.org Recorded when the context menu is constructed and logged if the search by image option was displayed. Recorded when the context menu is constructed and logged if the Lens option was displayed. Recorded when the context menu is constructed and logged if the Lens shopping option was displayed. Recorded when the context menu is constructed and logged if the Lens shopping chip was displayed. Recorded when the context menu is constructed and logged if the Lens translate chip was displayed. iwells@chromium.org sophiechang@chromium.org Metrics recorded when the user interacts with or dismisses a v2 cue. Records user interaction or non-interaction with the cue. Duration in milliseconds between showing the cue and user interaction or dismissal. Exponentially bucketed. iwells@chromium.org sophiechang@chromium.org Metrics recorded when the v2 cue is shown to the user. Number of background tabs suggested by the model that are still valid and relevant (e.g. not navigated away or closed). Exponentially bucketed. Number of background tabs suggested by the model that weren’t valid by the time the model response was received. Exponentially bucketed. Number of background tabs suggested by the model that are no longer relevant because they navigated to another URL before the model response was received. Exponentially bucketed. Records kSuccess when a cue is shown, or a different value indicating the reason why the cue was not shown. Duration in milliseconds between page load (WebContentsObserver::DidFinishNavigation()) and a cue being shown. Hash of the name of the critical user journey category output by the model. sophiechang@chromium.org zekunjiang@google.com Recorded when a decision to nudge an optimization type was made for the page load. The nudge decision made for optimization type for the page load. The optimization type a nudge decision was made for. sophiechang@chromium.org zekunjiang@google.com Recorded only if a contextual cueing nudge was shown to the user. The interaction the user had with the contextual cueing nudge. Whether the nudge is showing a dynamic cue (vs. a static cue). The amount of time it takes to display a nudge. This time is measured in milliseconds, and measured from page load to when the nudge is shown. The amount of time in milliseconds that a nudge is shown for, before user takes action on the nudge. Bucketed using GetExponentialBucketMinForUserTiming(). linnan@google.com measurement-api-dev+metrics@google.com We consider a non-user-initiated navigation to be a client bounce if it navigates the page away within a short period of time without any user activation/interaction. Recorded when client bounce occurs. The event is associated with the UKM source ID from the navigation ID of the most recent primary page navigation since M135. Measures how many non-navigation data hosts where registered on the page when a page was navigated away within 10s without any user activation by a non-user-initiated navigation. Measures how many non-navigation data hosts where registered on the page when a page was navigated away within 1s without any user activation by a non-user-initiated navigation. Note that these data hosts are also counted in "UserActivation.5s" and "UserActivation.10s". Measures how many non-navigation data hosts where registered on the page when a page was navigated away within 5s without any user activation by a non-user-initiated navigation. NNote that these data hosts are also counted in "UserActivation.10s". Measures how many non-navigation data hosts where registered on the page when a page was navigated away within 10s without any user interaction by a non-user-initiated navigation. Measures how many non-navigation data hosts where registered on the page when a page was navigated away within 1s without any user interaction by a non-user-initiated navigation. Note that these data hosts are also counted in "UserInteraction.5s" and "UserInteraction.10s". Measures how many non-navigation data hosts where registered on the page when a page was navigated away within 5s without any user interaction by a non-user-initiated navigation. Note that these data hosts are also counted in "UserInteraction.10s". bingler@chromium.org A collection of metrics which are recorded for cookies which include at least one non-ascii unicode character in either their name or value fields. This metric is only interested in an event being generated, the value it records is meaningless. Records events when a cookie has a non-ascii unicode character in its name field. Records events when a cookie has a non-ascii unicode character in its value field. amarjotgill@chromium.org bingler@chromium.org A collection of metrics which are recorded for cookies which will be blocked due to an origin mismatch, this means either the port or scheme it is being accessed by does not match that of the origin which originally set the cookie. True if the cookie was accessed by port which was mismatched. True if the cookie was accessed by scheme which was mismatched. chrisha@chromium.org siggi@chromium.org A CPU usage measurement that corresponds to a tick in a CPU usage time-series. There is also some metadata stored to determine how faithful the CPU usage measurement is (e.g. NumberOfCoresidentTabs). CPU usage measurement. This is in units of clock-cycles. For example, 1.2 seconds of 100% CPU usage on a 2.5GHz CPU corresponds to a value of 1.2 x 1.0 x 2.5e9 = 3e6. The event associated with the measurement. This is an integer value corresponding to the enum defined in resource_coordinator::RenderProcessProbe. Indicates if any frame hosted in this tab is visible at the moment the measurement is taken. Number of co-resident tabs in all of the render processes a tab is associated with when the CPU usage measurement is taken. The size of the observation window (time since last CPU measurement) in milliseconds. The uptime of the process being measured, in milliseconds. clamy@chromium.org Recorded when a subframe is embedded in a cross-origin context without having a CSP frame-ancestor header or an XFO header. Mesaures the number of pages with subframes embedded in cross-origin contexts without having a CSP frame-ancestor header or an XFO header. thanhdng@chromium.org Records daily summaries of charging. DailyChargeSummary summarizes charging information (e.g. amount of time on full charge, amount of time on AC adapter, etc) in a day. It can be used in combination with other DailyChargeSummary events to get information about the last 30 days of charging. If a "SmartCharging" event and a "DailyChargeSummary" event has them same source ID and "EventId", it means "DailyChargeSummary" is the summary when "SmartCharging" event is logged. DailyChargeSummary records an array. Each event records an item in the array, specified by its index DailySummaryIndex. Items in the same array will have same source ID and EventId. Amount of time (in minutes) the device was on AC charger and the charge was hold due to adaptive charging, at an index specified by DailySummaryIndex. Index of this event in the array. Please see the event summary for more details. Distance in days bettwen current day and the day that the daily summary was logged. Number of items in array. Please see the event summary for more details. Amount of time (in minutes) the device was keep on AC charger with full battery (100%). Amount of time (in minutes) the device was keep on AC charger. A number that represent order of an event. Used combined with "SmartCharging"'s EventId to reconstruct charge history. robertogden@chromium.org Data Reduction Proxy related metrics associated with a page load. See //components/data_reduction_proxy/. A unique identifier of this page load to the proxy server. Recorded for each navigation that uses the proxy. The estimated number of bytes that would have been used if that proxy was disabled on this page load. Recorded for each navigation that uses the proxy. skrakowi@chromium.org chrome-counter-abuse-alerts@google.com Records that the DevTools were opened by the user. A boolean signaling that the DevTools menu was opened by the user; only records true values. navtracking-team@google.com dc-komics@google.com This metric has been superceded by the BTM.ChainBegin UKM and is no longer emitted as of M138. Emitted after a redirect chain terminates, but for the page visited before the chain began -- the initial URL. A randomly-generated 32-bit id to connect this event with the DIPS.Redirect and DIPS.ChainEnd events emitted for the same redirect chain. True when the site before the redirect chain is the same as the site at the end of the chain. navtracking-team@google.com dc-komics@google.com This metric has been superceded by the BTM.ChainEnd UKM and is no longer emitted as of M138. Emitted for the page visited at the end of a redirect chain -- the final URL. A randomly-generated 32-bit id to connect this event with the DIPS.ChainBegin and DIPS.Redirect events emitted for the same redirect chain. True when the site before the redirect chain is the same as the site at the end of the chain. navtracking-team@google.com dc-komics@google.com Recorded when DIPS judges a site to be an incidental party and deletes its state if specific conditions are met. True iff the user has a third-party cookies blocking exception for the site (as a third-party). True iff site data deletion is enabled for the DIPS service. True when third-party cookies are blocked. navtracking-team@google.com dc-komics@google.com This metric has been superceded by the BTM.Redirect UKM and is no longer emitted as of M138. Recorded on every HTTP redirect or client-side bounce, where a client-side bounce is a non-user-initiated navigation starting soon after the previous navigation committed. A randomly-generated 32-bit id to connect this event with other DIPS.Redirect events and the DIPS.ChainBegin/ChainEnd events emitted for the same redirect chain. Populated for client redirects only. The time in seconds (capped at 10) between the page committing and the redirect navigation starting. An int in the range [0, 10]. Whether the redirect read cookies, wrote cookies, both, or neither. Populated for client redirects only. Whether the user interacted with the page (See the HTML spec for a definition). This info is persisted within BTM for a duration of `btm::kInteractionTtl`. True when the site before the redirect chain is the same as the site at the end of the chain. True if the chain is missing the end URL. Any comparisons to the final site are invalid for this redirect. True when the redirect site is the same as the site at the end of the redirect chain. True when the redirect site is the same as the site before the redirect chain. 0-based index of this individual redirect within its chain. The length of the overall redirect chain. The type of the redirect: client or server. Deprecated as of M139. The site engagement level of the redirect URL, at the time the redirect took place. Populated for client redirects only. Whether the user triggered a successful web authn assertion. This info is persisted within BTM for a duration of `btm::kInteractionTtl`. kaklilu@google.com dc-komics@google.com Recorded on a repeated interval when DIPS may want to clear site storage to improve user privacy. These metrics will be used to determine which triggering action is most appropriate for our use case. When these cases occur, the DIPS instance recording these metrics will keep track of which triggering actions have already happened to prevent logging duplicate metrics. True for sites that redirected the user since the last time the DIPS instance attempted to clear site storage. True for sites that redirected the user while accessing storage since the last time the DIPS instance attempted to clear site storage. True for sites wrote to the browser's storage since the last time the DIPS instance attempted to clear site storage. rakina@chromium.org navigation-dev@chromium.org Recorded when a Document object is created and records the source ID of the navigation and whether the document was in the main frame. This can be used to link subframe UKM events to the parent document. Whether the document was in a cross origin iframe. This can either be 0 or 1. Whether the document was in a cross site iframe. This can either be 0 or 1. Whether the document was in the main frame. This is can either be 0 or 1. Contains the UKM source id of the navigation as an integer. wylieb@google.com chrome-reader-mode@google.com Recorded for a subset of page loads (default: 1/1000 page loads). Records the result of the triggering model which determines if the current page is considered distillable. Recorded for the android-specific code which uses the results. The duration (in milliseconds) for how long it took from the time the distillability result to be returned. wylieb@google.com chrome-reader-mode@google.com Recorded for a subset of page loads (default: 1/1000 page loads). Records the result of the triggering model which determines if the current page is considered distillable. Recorded for the android-specific code which uses the result. The distillation result from the current page. wylieb@google.com chrome-reader-mode@google.com Recorded for a subset of page loads (default: 1/1000 page loads). Records when the user interacts with reader mode directly, after it's been offered. Recorded for the android-specific code which uses the results. Whether reader mode is shown for the current page. wylieb@google.com chrome-reader-mode@google.com Recorded for a subset of page loads (default: 1/1000 page loads). Records the result of the triggering model which determines if the current page is considered distillable. Whether the current page is distillable. clank-downloads@google.com Metrics taken when a download completes. Its parent event is Download.Started. The number of bytes that have been used in an excess (ie. download restarting the middle of a download). The id of the download that is used to associate separate events. The size of the file that is downloaded, expressed in kilobytes with exponentially growing buckets. The difference in the time between when the download finished and when it began, expressed in milliseconds. clank-downloads@google.com Metrics taken when a download is interrupted. Its parent event is Download.Started. The number of bytes that have been used in an excess (ie. download restarting the middle of a download). The difference in the size of the file that is downloaded compared to the initial reported size, expressed in kilobytes with exponentially growing buckets. The id of the download that is used to associate separate events. The reason the download was interrupted, expressed as an enum defined in DownloadInterruptReason. The size of the file that is downloaded, expressed in kilobytes with exponentially growing buckets. The difference in the time between when the download interrupted and when it began, expressed in milliseconds. clank-downloads@google.com Metrics taken when a download is resumed. Its parent event is Download.Started. The id of the download that is used to associate separate events. The mode by which the download was resumed, expressed as an enum defined in ResumeMode. The difference in the time between when the download resumed and when it began, expressed in milliseconds. clank-downloads@google.com Metrics taken when a download begins. It has one Download.Ended and none to multiple Download.Interrupted/Download.Resumed events associated with it. The state of the security of the final download URL and all the redirects leading to it. Expressed as an enum defined in DownloadConnectionSecurity. The id of the download that is used to associate separate events. The source of the download, expressed as an enum defined in DownloadEntry. The type of file that is downloaded, expressed as an enum defined in DownloadContentType. A boolean denoting if the final download URL is the same host as the initiating frame (i.e., whether the initiating site likely controls the download itself). cbe-device-trust-team@google.com Metrics related to Profiles enrolling into Enterprise cloud management. A boolean capturing that a Profile enrollment flow was blocked due to a lack of trust in the chain of navigation redirects. erikchen@chromium.org Recorded each time a background contents creates its first renderer by navigating its initial WebContents. Dummy Boolean as UKM does not support events with no metrics. quarz@chromium.org extensions-core@chromium.org Recorded when an extension redirects search away from the default search engine (DSE) page using the DeclarativeNetRequest API. Dummy Boolean as UKM does not support events with no metrics. howardcha@google.com webstore-eng@google.com Records user actions related to an extension, providing usage metrics for the Chrome Web Store. Tracked actions include pinning, unpinning, enabling, disabling, context menu access, and toolbar icon clicks. The action the user took on the extension. chikamune@chromium.org yyanagisawa@chromium.org sisidovski@chromium.org hiroshige@chromium.org kouhei@chromium.org Records the extension related metrics that might have negative impacts on LCP. This is recorded when navigation is finished. The number of enabled extensions that have a DeclarativeNetRequestFeedback permission. The exponential bucketing is applied for privacy reasons by using ukm::GetExponentialBucketMin(value, 2). The number of enabled extensions that have a DeclarativeNetRequest permission. The exponential bucketing is applied for privacy reasons by using ukm::GetExponentialBucketMin(value, 2). The number of enabled extensions that have a DeclarativeNetRequestWithHostAccess permission. The exponential bucketing is applied for privacy reasons by using ukm::GetExponentialBucketMin(value, 2). The number of enabled extensions that have a DeclarativeWebRequest permission. The exponential bucketing is applied for privacy reasons by using ukm::GetExponentialBucketMin(value, 2). The number of enabled extensions that are not themes nor apps nor components. The exponential bucketing is applied for privacy reasons by using ukm::GetExponentialBucketMin(value, 2). The number of enabled extensions that have host permissions for the navigation target URL. The exponential bucketing is applied for privacy reasons by using ukm::GetExponentialBucketMin(value, 2). The number of enabled extensions that inject content scripts. The exponential bucketing is applied for privacy reasons by using ukm::GetExponentialBucketMin(value, 2). The number of enabled extensions that have a WebRequestAuthProvider permission. The exponential bucketing is applied for privacy reasons by using ukm::GetExponentialBucketMin(value, 2). The number of enabled extensions that have a WebRequestBlocking permission. The exponential bucketing is applied for privacy reasons by using ukm::GetExponentialBucketMin(value, 2). The number of enabled extensions that have a WebRequest permission. The exponential bucketing is applied for privacy reasons by using ukm::GetExponentialBucketMin(value, 2). quarz@chromium.org extensions-core@chromium.org Recorded when an extension redirects a search away from the default search engine (DSE) page. The event is associated with the destination URL of the redirect. The API used to perform the redirect. quarz@chromium.org extensions-core@chromium.org Recorded when an extension uses the Tabs API to update the default search engine url of the tab. Dummy Boolean as UKM does not support events with no metrics. quarz@chromium.org extensions-core@chromium.org Recorded when an extension uses the Tabs API to remove a tab whose url is the default search engine. Dummy Boolean as UKM does not support events with no metrics. yhirano@chromium.org kinuko@chromium.org Whether and how keepalive requests are blocked. Recorded when a keepalive request finishes. The number of redirects the request experienced. This is up to 21, as specified at https://fetch.spec.whatwg.org/#http-redirect-fetch . The final state of the request. junhuihe@google.com qihuizhao@google.com payments-autofill-team@google.com An event that is logged when the eWallet FoP selector UI is closed. Set to true after user accepted the FoP selector UI and selected a FoP to use, and to false after user rejected/dismissed the FoP selector UI. The scheme of the payment link. junhuihe@google.com qihuizhao@google.com payments-autofill-team@google.com An event that is logged after the eWallet FoP selector UI is shown. The scheme of the payment link. Always set to true. So acts as a counter for the number of times a FoP selector is shown for eWallet. junhuihe@google.com qihuizhao@google.com payments-autofill-team@google.com An event that is logged after a given payment link in a certain page for an eWallet push payment flow is detected. This metric is different from blink::WebFeature::kLinkRelPayment because the blink::WebFeature::kLinkRelPayment metric counts when a link with payment rel is present, no matter whether it triggers eWallet payment flow or not, while this metric counts only when the eWallet payment flow is triggered. Always set to true and serves as a counter when analyzing the vents. siashah@google.com payments-autofill-team@google.com An event that is logged when the FoP selector UI for Pix is closed. Set to true after user accepted the FoP selector UI and selected a FoP to use, and to false after user rejected/dismissed the FoP selector UI. siashah@google.com payments-autofill-team@google.com An event that is logged after the FoP selector UI is shown. Always set to true. So acts as a counter for the number of times a FoP selector is shown for Pix. siashah@google.com payments-autofill-team@google.com An event that is logged after getting response from the InitiatePurchaseAction call to the server. The result of the InitiatePurchaseAction call. siashah@google.com longsheng@google.com payments-autofill-team@google.com An event that is logged after a likely Pix code is copied to the clipboard on any website. It includes whether the copy event occurred within an iframe. A boolean denoting if the Pix code copy event occurred within an iframe or in the mainframe. Always set to true. So acts as a counter for the number of times a likely Pix code is copied. chrome-kids-eng@google.com Logged whenever a page load for a supervised user had some or all of its content blocked. Either MainFrameBlocked is true, or NumBlockedIframes is non-zero. Whether the main frame for the navigation was blocked (in which case) none of the content of the site is loaded. The number of Iframes that were blocked by the supervised user filtering logic in this page load. Exponentially bucketed. zhouzj@google.com msramek@google.com Recorded when a user completes a HaTS survey. The answer the user has provided to survey question number 1. The answer the user has provided to survey question number 2. The answer the user has provided to survey question number 3. The answer the user has provided to survey question number 4. The answer the user has provided to survey question number 5. The answer the user has provided to survey question number 6. The answer the user has provided to survey question number 7. The answer the user has provided to survey question number 8. The answer the user has provided to survey question number 9. Boolean that records if a survey has been successfully completed. Survey ID associated with HaTS. This ID should be unique per HaTS survey and should be separate from any existing IDs. Boolean that records if a survey has been successfully loaded. mych@chromium.org rakina@chromium.org Records metrics of a fetch keepalive request with a specific set of "category" values in the request URL parameter. This is logged only once per request, at the end of the request handling in the browser process. https://docs.google.com/document/d/1TSbtp5I5Bc1pbAKrrEHfmZlAwgeh6AlgAH7GbDMNPRQ/edit?tab=t.0 An 8-bit integer that can be mapped to a specific value of a "category" parameter in the request URL. Deprecated as of 2025/05/22. An integer representing the error code of the request when it completes or fails. This comes from network::URLLoaderCompletionStatus.error_code. Deprecated as of 2025/05/22. An integer representing the error code of the request when it completes or fails. This comes from network::URLLoaderCompletionStatus.extended_error_code. An enum representing the end stage of the fetch keepalive request when the event is logged. The high bits of the unique number that connects a fetch keepalive request throughout different processes. The low bits of the unique number that connects a fetch keepalive request throughout different processes. Whether the Context is detached or not when this event is logged. An integer representing the error code of the request when it reaches the kLoaderCompleted stage. This comes from network::URLLoaderCompletionStatus.error_code. An integer representing the extended error code of the request when it reaches the kLoaderCompleted stage. This comes from network::URLLoaderCompletionStatus.extended_error_code. The number of redirects a fetch keepalive request loader has experienced so far when the event is logged. Note that the maximum value is 20, as specified at https://fetch.spec.whatwg.org/#http-redirect-fetch. The number of retries a fetch keepalive request loader has experienced so far when the event is logged. Note that the maximum value is 10. An enum representing the previous stage of the fetch keepalive request when the event is logged. When the request is abandoned from the "LoaderCreated" stage, this is also set to "LoaderCreated" stage. An integer representing the error code of the request when it reaches the kRequestfailed stage. This comes from network::URLLoaderCompletionStatus.error_code. An integer representing the extended error code of the request when it reaches the kRequestFailed stage. This comes from network::URLLoaderCompletionStatus.extended_error_code. An integer representing the error code of the request when it reaches the kRequestRetried stage. This comes from network::URLLoaderCompletionStatus.error_code. An integer representing the extended error code of the request when it reaches the kRequestRetried stage. This comes from network::URLLoaderCompletionStatus.extended_error_code. An enum representing the type of the fetch keepalive request. It tells whether this request is a fetch request, an attribution reporting request (obsolete), or a fetchLater request. The time difference (in milliseconds) between the "BrowserShutdown" stage and the "LoaderCreated" stage. Only set when the stage is reached. The time difference (in milliseconds) between when this event is logged and the "LoaderCreated" stage. The time difference (in milliseconds) between the "FirstRedirectReceived" stage and the "LoaderCreated" stage. Only set when the stage is reached. The time difference (in milliseconds) between the "LoaderCompleted" stage and the "LoaderCreated" stage. Only set when the stage is reached. The time difference (in milliseconds) between the "LoaderDisconnectedFromRenderer" stage and the "LoaderCreated" stage. Only set when the stage is reached. The time difference (in milliseconds) between the "RequestCancelledAfterTimeLimit" stage and the "LoaderCreated" stage. Only set when the stage is reached. The time difference (in milliseconds) between the "RequestCancelledByRenderer" stage and the "LoaderCreated" stage. Only set when the stage is reached. The time difference (in milliseconds) between the "RequestFailed" stage and the "LoaderCreated" stage. Only set when the stage is reached. The time difference (in milliseconds) between the "RequestRetried" stage and the "LoaderCreated" stage. Only set when the stage is reached. The time difference (in milliseconds) between the "RequestStarted" stage and the "LoaderCreated" stage. Only set when the stage is reached. The time difference (in milliseconds) between the "ResponseReceived" stage and the "LoaderCreated" stage. Only set when the stage is reached. The time difference (in milliseconds) between the "SecondRedirectReceived" stage and the "LoaderCreated" stage. Only set when the stage is reached. The time difference (in milliseconds) between the "SecondRedirectReceived" stage and the "LoaderCreated" stage. Only set when the stage is reached. mych@chromium.org rakina@chromium.org Records metrics to connect a pair of the following requests happening around the same time with the same "category" param value: (1) A navigation request with a URL param "category". (2) A fetch keepalive request with the same URL param "category". This is logged only once per such pair of requests, roughly at the time when the last of the two requests in a pair completes. This should be used together with "FetchKeepAliveRequest.WithCategory" event. https://docs.google.com/document/d/13P-jwVXFDrh6RBr-8H11GURKG89LziNlV1nAmlWAuB0/edit?tab=t.0 An 8-bit integer that can be mapped to a specific value of a "category" parameter in the request URL. The high bits of the unique number that connects a fetch keepalive request throughout different processes. The low bits of the unique number that connects a fetch keepalive request throughout different processes. A 64-bit integer that represents the navigation ID of the page navigation that has the same category param value as the fetch keepalive request with the Id.Low and Id.High logged in this event. mek@chromium.org rhalavatigit@chromium.org Logged when a website requests accessing FileSystem API. Marks a request for a Persistent FileSystem. Marks a request for a Temporary FileSystem. dslee@chromium.org memmott@chromium.org Quota usage of FileSystemObserver OS resource, collected when usage changes. Recorded when FileSystemAccessObserverQuotaManager is destroyed. High water mark of OS-specific observer resource usage. The unit used depends on OS. On inotify-based OS (Linux, Android), it is the count of inotify watches. On Mac, it is the count of FSEventsCreate calls. On Windows, it is the number of bytes used for buffer memory. This is rounded down to the nearest exponential bucket, with bucket spacing of 1.1. High water mark of OS-specific observer resource usage over quota limit in percentage, rounded down to the nearest integer. Whether FileSystemObserver usage change caused quota to be exceeded. yaoxia@chromium.org jkarlin@chromium.org Whenever the floc is computed, this event will be recorded at the commit time of the next page load. Caveat: if the browser is closed before the next page load occurs (in a new session), the event won't be recorded. The user's floc id. This ignores any general or per-context access permissions, and records the internal int64_t number directly. This metric may still be skipped if the floc was not calculated or has been invalidated. muyaoxu@google.com openscreen-eng@google.com The browser enters the fullscreen mode. The keyboard and pointer lock states. muyaoxu@google.com openscreen-eng@google.com The browser exits the fullscreen mode. Duration of the fullscreen session. phshah@chromium.org arc-gaming@google.com Logged when Game Controls' options menu button functions are triggered. The functions on the button options menu. phshah@chromium.org arc-gaming@google.com Logged when Game Controls edit-delete menu functions are triggered. The functions on the edit-delete menu. phshah@chromium.org arc-gaming@google.com Logged when Game Controls editing list functions are triggered. The functions on the editing list. phshah@chromium.org arc-gaming@google.com Log once whether the ARC game is played with Game Controls per game session. If it is played with Game Controls, record true immediately. If it is never played with Game Controls, record false when shutting down the game. Boolean indicating whether the ARC game is played with Game Controls. phshah@chromium.org arc-gaming@google.com Logged when Game Controls feature or hint is toggled on or off. Toggle the feature or mapping hint. Toggle the feature or mapping hint. Whether the function is toggled on or off. phshah@chromium.org arc-gaming@google.com Logged when Game Controls feature is toggled on or off. Game Controls feature is toggled on or off. phshah@chromium.org arc-gaming@google.com Logged when Game Controls mapping hint is toggled on or off. Where the mapping hint is toggled from. Game Controls mapping hint is toggled on or off. phshah@chromium.org arc-gaming@google.com Logged whether Game Controls is empty when entering Game Controls edit mode. Whether Game Controls is empty. phshah@chromium.org arc-gaming@google.com Logged when the game dashboard functions are triggered. Game dashboard function. phshah@chromium.org arc-gaming@google.com Logged where the recording starts from. Where is the recording started from, main menu or toolbar. phshah@chromium.org arc-gaming@google.com Logged where the screenshot is taken from. Where is the screenshot taken from, main menu or toolbar. phshah@chromium.org arc-gaming@google.com Logged when the game dashboard main menu is toggled on or off and how it is toggled. How the main menu is toggled. Whether the main menu is toggled on or off. phshah@chromium.org arc-gaming@google.com Logged when the game dashboard toolbar is expanded or collapsed. Whether the toolbar is expanded. phshah@chromium.org arc-gaming@google.com Logged when the game dashboard toolbar is snapped to a new location. Toolbar new snap location. phshah@chromium.org arc-gaming@google.com Logged when the game dashboard toolbar is toggled on or off. Whether the toolbar is toggled on or off. phshah@chromium.org arc-gaming@google.com Logged when the game dashboard welcome dialog notification is toggled on or off. Welcome dialog notification is toggled on or off. phshah@chromium.org arc-gaming@google.com An event that is logged when key mapping position is changed. Type about how the reposition is done. phshah@chromium.org arc-gaming@google.com An event that is logged when key mapping position is changed. Window state type when the reposition is done. phshah@chromium.org arc-gaming@google.com An event that is logged when button group position is changed. Type about how the reposition is done. phshah@chromium.org arc-gaming@google.com An event that is logged when menu entry position is changed. Window state type when the reposition is done. phshah@chromium.org arc-gaming@google.com An event for gaming input overlay that is logged when users enter into edit mode. Whether the customization option was used. phshah@chromium.org arc-gaming@google.com An event for gaming input overlay that is logged when the input overlay feature itself changes state. Whether the gaming input overlay feature is in use. phshah@chromium.org arc-gaming@google.com An event for gaming input overlay that is logged when the mapping hint state is changed. Whether the mapping hint for gaming input overlay is on. phshah@chromium.org arc-gaming@google.com An event that is logged when key mapping position is changed. Type about how the reposition is done. phshah@chromium.org arc-gaming@google.com An event that is logged when menu entry position is changed. Window state type when the reposition is done. ryansturm@chromium.org avi@chromium.org Logged when the user navigated to their home page or their default search engine. Recorded if the first URL or the final URL matches the pattern for Default Search or matches the user's home page URL. Recorded upon commit (even if the page was not foreground). If the navigation committed URL matches the template URL of the user's Default Search Engine. If the navigation committed URL matches the user's home page URL. Not recorded on Android because home page is managed outside of Chrome. If the navigation starting URL matches the template URL of the user's Default Search Engine. If the navigation starting URL (first URL in the redirect chain) matches the user's home page URL. Not recorded on Android because home page is managed outside of Chrome. sanaakbani@google.com iwells@google.com Recorded when an auto-opened Glic instance (such as for a PDF) is closed or dismissed by the user or system. The source_id represents the URL of the tab where auto-open was initially triggered. This metric is not recorded on tab switches and is only recorded when the last side panel associated with the auto-opened instance closes. The duration in milliseconds that the auto-opened Glic instance remained open. sanaakbani@google.com birnie@google.com Recorded when an unconsented user interacts with Glic during the onboarding funnel (opening Glic, viewing the FRE opt-in UI, accepting consent, or submitting a prompt). The source_id reflects the focused tab at the moment the event occurs, or NoURLSourceId if no tab is focused. The onboarding funnel step reached (NewUserOpen, FreOptInShown, FreOptInAccepted, FirstPromptSubmitted). The entry point used to invoke Glic. erikchen@chromium.org iwells@google.com Recorded when a response starts and was visible. The source id reflects the targeted tab, and can be an empty source ID if no tab is targeted. Whether the glic window was attached or detached. The source of the invocation of the glic window. The mode of the glic window. erikchen@chromium.org iwells@google.com Recorded when a glic window opens. The source id reflects the targeted tab, and can be an empty source ID if no tab is targeted. Whether the glic window was attached or detached. The source of the invocation of the glic window. dullweber@chromium.org arabm@google.com chrome-privacy-core@google.com Logged when a website requests a shared resource from Google Docs Offline Extension. Note that this is double counted for imports since it is done once in preload and once in actual load. Marks a resource request from a website. yuezhanggg@chromium.org ssid@chromium.org Collects information about member tabs that are in a tab group suggestion. Recorded after a suggestion is shown and user has reacted to the suggestion. An int ID corresponding to the suggestion. This will enable joining all tabs of the suggestion together. Enum of the type of the suggestion. This is decided by the type of heuristics used to trigger the suggestion. Boolean indicating whether this source is the active tab for the trigger, only recorded for 1 tab within all GroupSuggestionPromo.Suggestion events recorded with the same GroupSuggestionID. Int recording the number of times that the tab has been foregrounded in a predefined time window, currently 10 minutes. Int recording the number of seconds since the tab was last active. Exponentially bucketed using GetExponentialBucketMin with a spacing of 1.3 Enum of the type of user response. yaoxia@chromium.org shivanisha@chromium.org Records usage of `window.history.pushState` method, capturing details about user activation and ad-related information. Recorded when this method is invoked. True if the history API is invoked when an ad script is in the JavaScript stack, or if the invoking context is an ad iframe, as determined by Chrome's ad tagging mechanisms. (https://chromium.googlesource.com/chromium/src/+/lkgr/docs/ad_tagging.md) Whether the invoking context has seen user activation. (https://mustaqahmed.github.io/user-activation-v2) chrome-memories@google.com mcrouse@chromium.org Metrics that capture the iteractions with HistoryClusters, or Journeys, This will be recorded once per visit to the HistoryClusters UI and logged when the UI is closed or navigated away from. The initial state that describes how an interaction with the HistoryClusters UI was started. The number of times the user performed a query on the HistoryClusters UI during this interaction. This can include partial queries that occur due to the UI's query timing. The number of times the user toggled to the basic history page during this interaction with the HistoryClusters UI. shivanisha@chromium.org Logged when an entry in the back-forward list is marked to be skipped on subsequent back/forward button clicks as part of the history manipulation intervention. This is logged when the entry is navigated away from. arthursonzogni@chromium.org hajimehoshi@chromium.org Metrics recorded each time we commit a history navigation, which are needed to estimate benefits of back-forward cache. Except as otherwise noted, the metrics are recorded for history navigations, but not for same-document navigations. When navigating back to a page in the session history, this records what blocklisted features were used when the page was loaded and prevented this page from being stored in the back-forward cache. Multiple blocklisted features can be recorded for the same page as a binary mask. This entry records the features numbered from 0 to 63. When navigating back to a page in the session history, this records what blocklisted features were used when the page was loaded and prevented this page from being stored in the back-forward cache. Multiple blocklisted features can be recorded for the same page as a binary mask. This entry records the features numbered from 64 to 127. When navigating back to a page in the session history and the BackForwardCache wasn't used because of the related browsing instances, this records the reason why the browsing instance wasn't swapped. The number of the reasons passed to BackForwardCache::DisableForRenderFrameHost calls. Boolean whether the now-restored page uses AMP (see https://amp.dev). If the page wasn't restored from the back-forward cache, this metric will not be populated. This metric will never be true when BackForwardCache.IsServedFromBackForwardCache is false. Boolean whether the page was restored from the back-forward cache or not. Bitmask of the reasons why the page was not restored from the back-forward cache. See contents::BackForwardCacheMetrics::NotRestoredReason for the meaning of the individual bits. This entry records the features numbered from 0 to 63. Bitmask of the reasons why the page was not restored from the back-forward cache. See contents::BackForwardCacheMetrics::NotRestoredReason for the meaning of the individual bits. This entry records the features numbered from 64 to 127. Measures the maximum cumulative layout shift (bit.ly/cls-explainer) that has occurred in a session window, after the page is restored from the back-forward, before the first soft navigation. The gap between two consecutive shifts in a window is not bigger than 1000ms and the maximum window size is 5000ms. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. The number of distinct user interactions on a page after it is restored from the back-forward cache, before the first soft navigation. See definition of user interaction: https://web.dev/better-responsiveness-metric/#group-events-into-interactions An approximation of a high percentile of user interaction latency of a page after it is restored from the back-forward cache, in ms. For this metric, we measure the latency of user interactions by Responsiveness.UserInteraction.MaxEventDuration. Only interactions before the first soft navigation are considered. Bitmask of features numbered from 0 to 63 used by the cross origin subframes of the previous page load. See blink::SchedulingPolicy::Feature for the meaning for the individual features. Bitmask of features numbered from 64 to 127 used by the cross origin subframes of the previous page load. See blink::SchedulingPolicy::Feature for the meaning for the individual features. Deprecated as of 04/23. It is still reported, but most users should use PageLoad.LayoutInstability.MaxCumulativeShiftScore.AfterBackForwardCacheRestore.SessionWindow.Gap1000ms.Max5000ms. Measures the cumulative layout shift (bit.ly/3fQz29y) that has occurred during the session, after the page is restored from the back-forward cache. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. Measures First Input Delay, the duration between the hardware timestamp and the start of event processing on the main thread for the first meaningful input per navigation, after the page is restored from the back-forward cache. Measures the time duration between the page restore from back-forward cache and the first requestAnimationFrame time in milliseconds. For page loads that start in the foreground, measures the duration of time in milliseconds until one of the following events occurs: the load of the main resource fails, the page load is stopped, the tab hosting the page is closed, the render process hosting the page goes away, a new navigation which later commits is initiated in the same tab, or the tab hosting the page is backgrounded. This metric has a very long tail, for pages that spend a long period of time in the foreground. As such, use of the mean or tail values are not recommended. For history navigations and reloads, the source id of the previous navigation which loaded the page we're trying to navigate back to. Bitmask of features numbered from 0 to 63 used by the main frames of the previous page load. See blink::SchedulingPolicy::Feature for the meaning for the individual features. Bitmask of features numbered from 64 to 127 used by the main frames of the previous page load. See blink::SchedulingPolicy::Feature for the meaning for the individual features. Measures the maximum cumulative layout shift (bit.ly/cls-explainer) that has occurred in a session window, after the page is restored from the back-forward. The gap between two consecutive shifts in a window is not bigger than 1000ms and the maximum window size is 5000ms. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. Boolean for whether we navigated to the same navigation entry as the one which was last visible. It can be false when subframes or same-document navigations are present. For example, after navigating from http://foo to http://foo#bar and then http://bar and then going back by 2 entries will mean we went to http://foo while the last committed entry from this document was http://foo#bar. It's expected to be rare, but might be problematic for back-forward cache. Measures the time in milliseconds from navigation timing's navigation start to the time the first paint is performed, after the page is restored from the back-forward cache. The number of distinct user interactions on a page after it is restored from the back-forward cache. See definition of user interaction: https://web.dev/better-responsiveness-metric/#group-events-into-interactions The reason (as a |page_load_metrics::PageEndReason|) why the user navigated away from the page after it was restored from the back-forward cache. This includes navigations after which the page is stored in the back-forward cache again, as well as other page-ending events. Bitmask of features numbered from 0 to 63 used by the same origin subframes of the previous page load. See blink::SchedulingPolicy::Feature for the meaning for the individual features. Bitmask of features numbered from 64 to 127 used by the same origin subframes of the previous page load. See blink::SchedulingPolicy::Feature for the meaning for the individual features. Measures the time duration between the page restore from back-forward cache and the second requestAnimationFrame time in milliseconds. Number of soft navigations detected by heuristics. See https://github.com/WICG/soft-navigations Soft navigations are JS-driven same-document navigations that are using the history API or the new Navigation API, triggered by a user gesture and meaningfully modify the DOM, replacing the previous content with new one. Measures the time duration between the page restore from back-forward cache and the third requestAnimationFrame time in milliseconds. Time in milliseconds from the moment the current navigation stopped being active to the start of the current navigation. This is clamped to hours for values greater than 3 hours, to minutes for values greater than 3 minutes, to seconds for values greater than 5 seconds. An approximation of a high percentile of user interaction latency of a page after it is restored from the back-forward cache, in ms. For this metric, we measure the latency of user interactions by Responsiveness.UserInteraction.MaxEventDuration. The maximum value of user interaction latency of a page after it is restored from the back-forward, in ms. For this metric, we measure the latency of user interactions by Responsiveness.UserInteraction.MaxEventDuration. meacer@chromium.org jdeblasio@chromium.org Metrics recorded when an HTTPS-First Mode event happens. The reason why the navigation fell back to HTTP. An enum value representing the decision the user made when shown an interstitial (proceed, don't proceed, etc.). The enum is defined in |HttpsOnlyModeBlockingPage|. mych@google.com chrome-webium-product-eng@google.com Navigation timing metrics associated with the main frame navigation for Initial WebUI pages. The time relative to navigation start that the first HTTP request is sent for the main resource of a main frame navigation. The time relative to navigation start that the headers of the first HTTP response are received for the main resource of a main frame navigation. The time relative to navigation start that navigation commit message for a main frame navigation is received from a renderer process. The time relative to navigation start that navigation commit reply for a main frame navigation is sent to a renderer process. The time relative to navigation start that navigation commit message for a main frame navigation is sent to a renderer process. The time relative to navigation start that navigation for a main frame navigation did commit from a renderer process. mych@google.com chrome-webium-product-eng@google.com Core metrics associated with the page loads for Initial WebUI pages. Amount of CPU wall time over the duration of a page load, while the page is in the foreground. Measured in milliseconds. The day of the week when the page load was initiated. The time is measured in the local time zone according to the system clock. Only collected for page loads that start while on foreground. Measures the time in milliseconds from navigation timing's navigation start to the time the DOMContentLoaded event is fired, for main frame documents. Measures the time in milliseconds from navigation timing's navigation start to the time the load event is fired, for main frame documents. The hour of the day when the page load was initiated. The time is measured in the local time zone according to the system clock. Only collected for page loads that start while on foreground. A boolean indicating if any mainpage resource has a 'Cache-control: no-store' response header. Recorded only for main frame request. The |page_load_metrics::PageEndReason| for the main frame navigation of this page load. Replaced Navigation.PageEndReason2 on January 2021 to get rid of END_HIDDEN and use END_APP_BACKGROUND on cases where the metrics are flushed and the app enters the background. The |ui::PageTransition| for the main frame navigation of this page load. The |net::Error| code for this page load. Only logged for failed provisional navigations. For page loads that start in the foreground, measures the duration of time in milliseconds until one of the following events occurs: the load of the main resource fails, the page load is stopped, the tab hosting the page is closed, the render process hosting the page goes away, a new navigation which later commits is initiated in the same tab, or the tab hosting the page is backgrounded. This metric has a very long tail, for pages that spend a long period of time in the foreground. As such, use of the mean or tail values are not recommended. Measures the time in milliseconds from navigation timing's navigation start to the time the provisional load failed. Only measures provisional loads that failed in the foreground. The total amount of time the page spent in the foreground, in milliseconds. Note that this metric adds all foreground durations occurring for the page. For example: if the page starts foregrounded during 1 second, then goes to the background for a while, then is foregrounded again for 2 seconds, and then is unloaded, this metric reports 3 seconds. To see metrics about consecutive foreground sessions, see the PageForegroundSession event instead. This metric is recorded at the end of each page load. As usual with PageLoad metrics, we consider app backgrounding on Android to be the end of page load as well as the app could be evicted by Android anytime after that happens. Measures the time in milliseconds from navigation timing's navigation start to the time the first contentful paint was rendered by Blink and submitted to the compositor (but not yet presented on screen). Measures the time in milliseconds from navigation timing's navigation start to the time Lit DOM composition completes. Measures the time in milliseconds from navigation timing's navigation start to the time WebUI JS resources have evaluated. Measures the time in milliseconds from navigation timing's navigation start to the time loadTimeData initial state is parsed. Measures the time in milliseconds from navigation timing's navigation start to the time when the page first paints content. A contentful paint includes a paint of text, image, SVG, or canvas. Log of major changes: http://bit.ly/chrome-speed-metrics-changelog Measures the time in milliseconds from navigation timing's navigation start to the time the first paint is performed. Measures the time in milliseconds from navigation timing's navigation start to the time the parser started, for main frame documents. Records how the main frame's site instance was assigned to a renderer process. Set to 1 if the main resource was served from cache. mehrab@google.com Metrics recording a summary of user interactions with Autocorrect feature. The event is recorded whenever an autocorrect suggestion is shown or a user interacted with a suggestion to accept or reject it. A summary of interactions with Physical Keyboard Autocorrect suggestions for the purpose of identifying App Compatibility issues. A summary of interactions with Virtual Keyboard Autocorrect suggestions for the purpose of identifying App Compatibility issues. jiwan@chromium.org Recorded when an assistive action could be triggered according to the surrounding text. The type of the assistive suggestion which is triggered. jopalmer@google.com essential-inputs-team@google.com Recorded when a user uses the long press diacritics feature to insert diacritics via long press with a physical keyboard. Action taken by the user to interact with this feature. curtismcmullan@google.com Metrics recording a summary of critical user interactions with the Editor feature. The event is recorded whenever a user triggers a critical event in the Editor funnel (specifically; the UI is shown, a request is triggered, or a user inserts some suggested text). A summary of interactions with the Editor feature for the purpose of identifying common contexts where the feature is used. shend@chromium.org Metrics indicating what Input Method Editor (IME) operations are behaving incorrectly on an input field. This event is recorded whenever an IME operation is executed on the input field but the result does not match the IME's expectations (e.g. IME wants to insert 'a' but the input field inserted 'b' instead). The operation that behaved incorrectly. rayankans@chromium.org peter@chromium.org Recorded before resolving a call to navigator.getInstalledRelatedApps(). Always true. nicolasmacbeth@google.com bling-alchemy-eng@google.com Logged when the user chooses one of the calendar types on the ExperienceKit context menu. The type of calendar provider the user chose. nicolasmacbeth@google.com bling-alchemy-eng@google.com Logged when the user successfully adds an event to one of the calendar providers in ExperienceKit. The type of calendar provider the user chose. thegreenfrog@chromium.org michaeldo@chromium.org Logged when the FindInPage returns a user search request result. Starting from M111, this is only logged at the end of the Find session i.e. when the Find UI is dismissed. True if there were matches. ajuma@chromium.org alionadangla@chromium.org Logged when a user exited the fullscreen mode when browsing a web page in iOS. True if the user tapped on the top toolbar to exit and false if they exited it in any other way. amarjotgill@chromium.org bling-ai-foundations@google.com Recorded when a page is classified. The category ID that was the best match. The confidence score (0-100) of the best match. Confidence score (0-100) for Car Search. Confidence score (0-100) for Clothes Search. Confidence score (0-100) for Flight Search. Confidence score (0-100) for Hotel Search. Confidence score (0-100) for Job Search. Confidence score (0-100) for News Search. Confidence score (0-100) for None/Unknown. Confidence score (0-100) for Restaurant Search. Whether Annotated Page Content (text) was successfully extracted. Is this something to try on. Is this an item to view in your room or house. thegreenfrog@chromium.org Logged when the user adds a page to the Reading List. True if the page was added by the user tapping on the Messages prompt to save to Reading List. thegreenfrog@chromium.org Logged when the Distilibility score of the current page is returned and the Reading List Message is shown. Muliplied by 10X to get tenth digit granularity. 0.1 granularity will be recorded for scores between 0.5 and 1.5. Otherwise, the granularity will be 0.5. True if the user tapped on the Messages prompt to save the current page to Reading List. Distilibility score value indicating the "long readability" of the page. Distilibility score value indicating the "readability" of the page. rkgibson@google.com Logged when the user changes their zoom level on iOS. The content size category the user is currently using. The overall zoom level. This is based on a combination of the user zoom level and a multiplier from the user's content size category. The current zoom multiplier (percentage-based, with 100 -> 100%) the user has chosen. thegreenfrog@chromium.org huitingyu@google.com Logged when the current page has a valid parcel tracking number in the HTML of the initial page load. The parcel carrier for the parcel number detected. As a privacy measure, 80% of the time when the event occurs, a random enum value is assigned. fernandex@google.com qpubert@google.com Recorded for a subset of page loads (default: 1/1000 page loads). Records the latency of running the DOMDistiller distillation to evaluate its performance. Records duration in milliseconds from when the page distillation is triggered to when the JavaScript distillation code completes. fernandex@google.com qpubert@google.com Recorded for a subset of page loads (default: 1/1000 page loads). Records the result of running the DOMDistiller distillation to evaluate its performance. Records the page distillation result i.e. distillable or not distillable. fernandex@google.com qpubert@google.com Recorded for a subset of page loads (default: 1/1000 page loads) if the latency data is available. Records the latency of running the DOMDistiller heuristic to evaluate its performance. Records duration in milliseconds from when the heuristic JavaScript is executed to when all required feature values for the model are collected. fernandex@google.com qpubert@google.com Recorded for a subset of page loads (default: 1/1000 page loads). Describes the result of running the DOMDistiller heuristic to evaluate its performance. Records the page distillation heuritic result i.e. kMalformedResponse, kReaderModeEligible, kReaderModeNotEligibleContentOnly, kReaderModeNotEligibleContentLength, kReaderModeNotEligibleContentAndLength. fernandex@google.com qpubert@google.com Records when Reading Mode UI was successfully displayed to the user. Recorded for iOS interactions only. The access point that triggered the Reading Mode UI that is shown for the current page. olivierrobin@chromium.org Logged when WKWebView process is reported gone. Metrics will give information of the context in which some website present sad tabs. It will help debug some renderer issues either in Chrome or in WebKit. The number of WKWebView alive in the last 2 seconds. The number of WKWebView alive at the moment this renderer is gone. Whether the app/tab was in foregound or background. 0: tab in foreground, app in foreground, 1: tab in background, app in foreground, 2: tab in foreground, app in background, 3: tab in background, app in background. Whether a memory warning was received less than 5 seconds before the renderer process is gone. houssein@chromium.org npm@chromium.org Metrics stating whether the main frame of a page uses certain JavaScript frameworks. This event is recorded when the page is going to be unloaded or when the app enters the background on Android. The framework detection is done when the document is loaded (similar to when load event is fired). If a framework is used after this then it would not be detected. Note that a page can use multiple frameworks at the same time, hence why this reports each framework individually instead of a single enum. True if the page loaded in the main frame uses the Angular JavaScript framework. True if the page loaded in the main frame uses the Gatsby JavaScript framework. True if the page loaded in the main frame uses the MediaPipe JavaScript framework. True if the page loaded in the main frame uses the Next.js JavaScript framework. True if the page loaded in the main frame uses the Nuxt.js JavaScript framework. True if the page loaded in the main frame uses the Preact JavaScript framework. True if the page loaded in the main frame uses the React JavaScript framework. True if the page loaded in the main frame uses the Sapper JavaScript framework. True if the page loaded in the main frame uses the Svelte JavaScript framework. True if the page loaded in the main frame uses the TensorflowJS framework. True if the page loaded in the main frame uses the Transformers.js framework. True if the page loaded in the main frame uses the Vue JavaScript framework. True if the page loaded in the main frame uses the VuePress JavaScript framework. niharm@google.com lens-chrome@google.com Recorded once when the Lens Overlay feature is dismissed, sliced by page content type. Records whether the contextual searchbox was focused at least once in sessions in which it was shown. Whether the searchbox was focused in the session. Focused in session sliced by page content type. mercerd@google.com lens-chrome@google.com Recorded once when the Lens Overlay feature is dismissed. Records whether the contextual searchbox was shown in the session. The page content type the contextual searchbox was shown on. Whether the search box was shown. niharm@google.com lens-chrome@google.com Recorded once when the Lens Overlay feature is dismissed, sliced by page content type. Records whether at least one contextual query was issued in sessions in which the contextual searchbox was shown (includes zero prefix suggest queries and typed queries). The page content type is the intial page type when the lens overlay is invoked, not necessarily the same as the page type as when ZPS is shown. Query issued in session sliced by page content type. Whether a query was issued in the session. niharm@google.com lens-chrome@google.com Recorded once when the Lens Overlay feature is dismissed, sliced by page content type. Records whether any zero prefix contextual suggestions were shown in sessions in which the contextual searchbox was shown. The page content type is the intial page content type when the lens overlay is invoked, not necessarily the same as the page content type as when ZPS is shown. ZPS shown in session sliced by page content type. Whether ZPS was shown in the session. niharm@google.com lens-chrome@google.com Recorded once when the Lens Overlay feature is dismissed, sliced by page content type. Records whether at least one contextual zero prefix suggestion was used in sessions in which the contextual searchbox was shown. The page content type is the intial page content type when the lens overlay is invoked, not necessarily the same as the page content type as when ZPS suggestion was used. ZPS used in session sliced by page content type. Whether ZPS was used in the session. jdonnelly@google.com lens-chrome@google.com Recorded each time the user interactions with the Lens Overlay, enumerated by those actions. The user interaction with the Lens Overlay. jdonnelly@google.com lens-chrome@google.com Recorded once when the Lens Overlay feature is dismissed. Measures the invocation source of the session, the duration, and whether a search performed in the session. The number of tabs opened by the user during a Lens Overlay session. On iOS the Lens Overlay opens all non-lens URLs in new tabs. New unimodal searches are also opened in a new tab. Additionally the user can open tabs in the background through the web context menu. The type of document the Lens Overlay was invoked on. Indicates whether or not a search was performed at any point since the overlay was invoked. The UI entry point used to invoke the Lens Overlay feature. Measures the elapsed time between when the overlay was invoked and when it was dismissed. Measures the total time the Lens Overlay is in the foreground, between invocation and dismissal. Sum of all times that the overlay tab is returned to the foreground. jdonnelly@google.com lens-chrome@google.com Recorded once when the user first interacts with the overlay. Measures the elapsed time (ms) between when the overlay was invoked and when the user first interacts with the overlay. Nothing is logged when the user enters through the image context menu. Time to first interaction for all entry points. Time to first interaction when overlay invoked through the app menu. Time to first interaction when overlay invoked through the page content area context menu. Time to first interaction when overlay invoked through the selected text context menu. Time to first interaction when overlay invoked through the find in page button. The first interaction event type. Time to first interaction when overlay invoked through the omnibox Lens homework page action suggestion. Time to first interaction when overlay invoked through the omnibox. Time to first interaction when overlay invoked through the omnibox contextual suggestion. Time to first interaction when overlay invoked through the omnibox Lens page action suggestion. Time to first interaction when overlay invoked through the toolbar. fdoray@chromium.org chrome-catan@google.com Counts loads of each type that occurred on a top-level document. Recorded when a top-level document is unloaded. This can be used to diagnose changes to the "Total pageloads" shown on stability dashboard. Exponential bucketing is applied to all counts. The bucketing function preserves exact count up until 20. Number of main frame same-document loads that occurred while the page was hidden. Number of main frame same-document loads that occurred while the page was visible. Number of sub frame different-document loads that occurred while the page was hidden. Number of sub frame different-document loads that occurred while the page was visible. Number of sub frame same-document loads that occurred while the page was hidden. Number of sub frame same-document loads that occurred while the page was visible. sophiechang@chromium.org tbansal@chromium.org Metrics that reflect the accuracy of the predictions used by the Loading Predictor. This event will be recorded once per page load on DocumentOnLoadCompletedInPrimaryMainFrame(). The number of subresource origin preconnects that were initiated by the Loading Predictor for the navigation and were actually used by the page load, excluding the main frame URL's origin. If more than 100 subresource origins were preconnected to and were actually used by the page load, this will be capped at 100. The number of subresource prefetches that were initiated by the Loading Predictor for the navigation and were actually used by the page load. If more than 100 subresources were prefetched and were actually used by the page load, this will be capped at 100. The number of subresource origins that were correctly predicted by the local prediction database. This will be recorded if there is a prediction available in the local prediction database and does not necessarily mean that the prediction was used. If more than 100 subresource origins were correctly predicted, this will be capped at 100. The number of subresource origins that were predicted locally. This will be recorded if there is a prediction available in the local prediction database and does not necessarily mean that the prediction was used. If more than 100 subresource origins were predicted, this will be capped at 100. The time (in milliseconds) that elapsed between navigation start and the first subresource prefetch that was initiated based on these predictions. The time (in milliseconds) that elapsed between navigation start and the commit finished. The time (in milliseconds) that elapsed between navigation start and the optimization guide prediction arriving for the navigation. The number of predicted subresource origins that were only used by the page for low-priority subresource loads before it unloaded. This will be recorded if the Optimization Guide was consulted for the page load and the Optimization Guide had a prediction for it. If more than 100 subresource origins were correctly predicted, this will be capped at 100. The number of predicted subresources that were only used by the page for low-priority subresource loads before it unloaded. This will be recorded if the Optimization Guide was consulted for the page load and the Optimization Guide had a prediction for it. If more than 100 subresources were predicted, this will be capped at 100. The number of predicted cross-origin subresources that were only used by the page for low-priority subresource loads before it unloaded. This will be recorded if the Optimization Guide was consulted for the page load and the Optimization Guide had a prediction for it. If more than 100 subresources were predicted, this will be capped at 100. The number of subresource origins that were correctly predicted by the Optimization Guide. This only counts origins of high-priority subresources that are used before the page finishes loading. This will be recorded if the Optimization Guide was consulted for the page load and the Optimization Guide had a prediction for it. If more than 100 subresource origins were correctly predicted, this will be capped at 100. The number of subresources that were correctly predicted by the Optimization Guide. This only counts high-priority subresources used before the page finishes loading. This will be recorded if the Optimization Guide was consulted for the page load and the Optimization Guide had a prediction for it. If more than 100 subresources were predicted, this will be capped at 100. The number of cross-origin subresources that were correctly predicted by the Optimization Guide. This only counts high-priority subresources used before the page finishes loading. This will be recorded if the Optimization Guide was consulted for the page load and the Optimization Guide had a prediction for it. If more than 100 subresources were predicted, this will be capped at 100. The decision coming from the Optimization Guide for whether there was a prediction available for the Page Load. This will be recorded if the Optimization Guide was consulted for the page load. The number of subresource origins that were predicted by the Optimization Guide. This will be recorded if the Optimization Guide was consulted for the page load and the Optimization Guide had a prediction for it. If more than 100 subresource origins were predicted, this will be capped at 100. The number of subresources that were predicted by the Optimization Guide. This will be recorded if the Optimization Guide was consulted for the page load and the Optimization Guide had a prediction for it. If more than 100 subresources were predicted, this will be capped at 100. The number of cross-origin subresources that were predicted by the Optimization Guide. This will be recorded if the Optimization Guide was consulted for the page load and the Optimization Guide had a prediction for it. If more than 100 subresources were predicted, this will be capped at 100. The number of subresource origins that were initiated by the Loading Predictor and were preconnected to, excluding the main frame URL's origin. If more than 100 subresource origins were preconnected to, this will be capped at 100. The number of subresource prefetches that were initiated by the Loading Predictor for the navigation. If more than 100 subresources were prefetched, this will be capped at 100. yigu@chromium.org tbansal@chromium.org web-identity-eng@chromium.org Records the info about login detected for sites based on different heuristics. Recorded once per page load when OAuth login was detected on page load. Note: the first version of LoginDetection recorded all pageload types including password logins and no logins etc. An enum for representing the different ways of user login to a site. At this moment only OAuth logins are recorded in UKM. yigu@chromium.org tbansal@chromium.org web-identity-eng@chromium.org Records the OAuth provider info about login detected for sites based on different heuristics. Recorded once per page load when OAuth login was detected on page load. Note: the first version of LoginDetection recorded all pageload types including password logins and no logins etc. An enum for representing the different ways of user login to a site. At this moment only OAuth logins are recorded in UKM. meacer@chromium.org cthomp@chromium.org Metrics recorded when a navigated URL's domain name is visually similar to a popular domain or a domain that the user engaged with. An enum value representing the type of the match (popular domain, engaged domain, etc.). The enum is defined in |LookalikeUrlBlockingPage|. A boolean representing whether or not the warning was triggered by the first URL in the chain, or whether the warning was triggered by the final (committed) URL. An enum value representing the decision the user made when shown an interstitial (ignored, suggestion_accepted, etc.). The enum is defined in |LookalikeUrlBlockingPage|. thegreenfrog@chromium.org Metric associated with a main frame navigation. Currently implemented only on iOS. Set to 1 if the navigation successfully committed. nidhijaju@chromium.org blink-network-stack@google.com Event recorded on navigation commit when a page load is content-encoded with zstd and is a main frame navigation. Set to 1 if the main frame served zstd-compressed content. bkeen@google.com media-dev@chromium.org Recorded when browser-initiated automatic picture-in-picture is triggered. This can be a dry run (the feature is enabled, but the action is not taken), or it can result in entering picture-in-picture. This event allows analysis of the feature impact. True if the event was a dry run, meaning the feature was enabled but the action was not taken. False if the browser actually entered picture-in-picture. bkeen@google.com media-dev@chromium.org Recorded when the user either selects an option from the Automatic picture-in-picture permission prompt, or when the user manually closes the picture-in-picture window without responding to the prompt, or when we decide to not show the prompt at all. When the prompt is shown, this records their selection (or "Ignored" if they did not select an option). When the prompt is not shown, records why it was not shown. Note that although "Not shown because incognito" is part of the MediaAutoPictureInPicturePromptResult enum, UKMs are not collected in incognito. Prompt result for entering picture in picture automatically due to browser initiated auto picture in picture. Prompt result for entering picture in picture automatically due to "MediaPlayback". Prompt result for entering picture in picture automatically due to "VideoConferencing". dalecurtis@chromium.org media-dev@chromium.org Event recorded when there is an attempt to autoplay (ie. no user gesture). It will be recorded regardless of the result of this attempt. Whether the element had an audio track when autoplay was attempted. Whether the document has a high media engagement. Whether the element was muted when autoplay was attempted. Source of the autoplay attempt: 0 for attribute; 1 for play(). Whether a user gesture was required per autoplay rules at the time of attempt. By definition there is no user gesture on the stack when the attempt is registered. Reflects the current status of user gesture/activation. This is a bit field with the following values: - 0b0001 if there is a user gesture on the stack; - 0b0010 if there was a user gesture on the page (ie. was activated); - 0b0100 if there was a user gesture propagated after a navigation. Whether the element had a video track when autoplay was attempted. dalecurtis@chromium.org media-dev@chromium.org Records the AudioContext autoplay information. A boolean indicating if start() was called for any source node associated to the AudioContext. Status of this AudioContext when the autoplay policy applies. It will match the values from AudioContext::AutoplayStatus. How the AudioContext was unlocked if it was. It will match the values from AudioContext::AutoplayUnlockType unless the AudioContext was never unlocked in which case it will be equal to -1. dalecurtis@chromium.org media-dev@chromium.org Event recorded when there is an attempt to unmute an media that was autoplaying following the rules of autoplay muted. 0 means that the unmute failed because it happened without a user gesture. 1 means that it succeeded because it had a user gesture. Similar to "Source" in "Media.Autoplay.Attempt" with the addition of a value for both sources being used: 0 for attribute; 1 for play(); 2 for both attempted. dalecurtis@chromium.org media-dev@chromium.org Core metrics associated with a media playback through WebMediaPlayerImpl. Reported at the time of WatchTimeRecorder destruction or render process termination; whichever comes first. There will be multiple entries for a given playback; one for every PlaybackProperties change as well as one for foreground and background. Records may be aggregated based on PlayerID and correlated with the Media.WebMediaPlayerState event. media::AudioCodec enum value. Can be kUnknownAudioCodec even when HasAudio is true because we don't always know the codec. media::AudioCodecProfile enum value. Can be AudioCodecProfile::kUnknown even when HasAudio is true because we don't always know the codec profile. Enumeration of audio decoder implementations, zero if none or unknown (Cast, HLS, etc). See AudioDecoderType (media/base/decoder.h) for expected values. media::EncryptionScheme enum value (media/base/encryption_scheme.h). Can be kUnencrypted if IsEME is false or if the audio track is unencrypted. Boolean value indicating whether the initial playback was initiated via autoplay. Multilpe chunk records of the same player should have the same value. Autoplay is defined as a playback starting before there was a user activation on the page or without a user initiated same-domain navigation. Integer count of the number of times a previous rebuffer succeeded and playback continued. Sum in milliseconds of all completed rebuffering events. Only reported if a rebuffering completion occurred. Duration in milliseconds, rounded to the most significant digit, of the media being played. May be missing if the duration is unknown or infinite. Boolean value indicating the presence of audio. Boolean value indicating the presence of video. Boolean value indicating if this event is for a background playback. Boolean value indicating if this event is for an EME playback. Boolean value indicating if this event is for an MSE playback. If false it means this was a SRC playback. Boolean value indicating if this event is for a muted playback. Flag indicating whether the report comes from the top frame or some inner frame. For privacy, metrics from inner frames are recorded with the top frame's origin, so this flag helps separate top frame vs. embedded playbacks. media::PipelineStatusCodes enum value (media/base/pipeline_status.h). Always 0 if the playback succeeded; all other values indicate the playback ended in an error. Average number of milliseconds between rebuffering events. Only reported if a rebuffering event occurred. Computed by dividing total watch time by the number of rebuffering events. Enum value indicating the type of MediaStream if the playback is from a MediaStream, or kNonMediaStream otherwise. ID which corresponds to a given WebMediaPlayerImpl instance. May be linked with Media.WebMediaPlayerState events to understand playback more deeply. Integer count of the number of times playback experienced rebuffering. Type of media::Renderer used for the playback. In most cases we do not support dynamically changing RendererType during one playback. media::VideoCodec enum value. Can be kUnknownVideoCodec even when HasVideo is true because we don't always know the codec. media::VideoCodecProfile enum value. Can be VIDEO_CODEC_PROFILE_UNKNOWN even when HasVideo is true because we don't always know the codec profile. Enumeration of video decoder implementations, zero if none or unknown (Cast, HLS, etc). See VideoDecoderType (media/base/decoder.h) for expected values. media::EncryptionScheme enum value (media/base/encryption_scheme.h). Can be kUnencrypted if IsEME is false or if the video track is unencrypted. Integer count of the video frames decoded in this record. Integer count of the video frames dropped in this record. Should not exceed VideoFramesDecoded. Integer value indicating the natural height of the playback. Integer value indicating the natural width of the playback. Watch time in milliseconds for this playback. See Media.WatchTime UMA metrics. Watch time on AC power in milliseconds for this playback. See Media.WatchTime UMA metrics. Watch time for automatic picture in picture. Only reported for the "media playback" automatic picture in picture reason, regardless of the picture in picture window type: document vs. video. Watch time on battery power in milliseconds for this playback. See Media.WatchTime UMA metrics. Watch time for fullscreen in milliseconds for this playback. See Media.WatchTime UMA metrics. Only reported for foreground playbacks. Watch time for inline display in milliseconds for this playback. See Media.WatchTime UMA metrics. Only reported for foreground playbacks. Watch time for picture in picture display in milliseconds for this playback. See Media.WatchTime UMA metrics. Only reported for foreground playbacks. Watch time with no controls in milliseconds for this playback. See Media.WatchTime UMA metrics. Only reported for foreground playbacks. Watch time with native controls in milliseconds for this playback. See Media.WatchTime UMA metrics. Only reported for foreground playbacks. xhwang@chromium.org media-dev@chromium.org Event recorded when an Encrypted Media Extensions (EME) API promise is rejected. In most cases this indicates a failure. The EME API for which the promise is rejected. The key system associated with the current CDM. A CDM specific code providing more details about the cause of rejection. Whether hardware secure codecs are required for the current CDM. frs@chromium.org vpasupathy@chromium.org media-dev@chromium.org Metrics reported by the CDM to the browser, collected in an utility process. This metric is reported every time on the destruction of the CdmAdapter object. Some fields of this metric can be left unset if the CDM does not report them. The certificate serial number used by the main url given by the CDM server certificate. Different Key systems will report the serial number in different ways. The number of blocks that can bypass the decoder present in an encrypted playback. The number of video frames that returned an error result when running decoder check1 in an encrypted playback. The number of video frames that returned a successful result when running decoder check1 in an encrypted playback. The number of video frames that returned a warning result when running decoder check1 in an encrypted playback. Whether or not the Key System is reporting true for bool 1. This is a general metric that is mapped to specific Key System data. The time in seconds since the unix epoch for the date of the creation of time 1. This is a general metric that is mapped to specific Key System data. The time in seconds since the unix epoch for the date of the creation of time 2. This is a general metric that is mapped to specific Key System data. The time in seconds since the unix epoch for the date of the creation of time 3. This is a general metric that is mapped to specific Key System data. The SDK version of the license server of the URL that communicates with the CDM. The number of OnMessage events for the CDM. The number of Update events for the CDM The number of video frames in an encrypted playback. The type of initialization data used in the CDM session. vpasupathy@chromium.org media-dev@chromium.org Event recorded whenever CDM reports changed associated with keys for MediaKeySession. An integer corresponding to a system code for a singular key supplied by the CDM. xhwang@chromium.org media-dev@chromium.org Event recorded when createMediaKeys() is called as part of Encrypted Media Extensions (EME) API. Whether the frame is an ad frame. Whether the frame is Cross-Origin to the main frame. Whether the frame is the main frame. The key system associated with this call. frs@chromium.org media-dev@chromium.org Event recorded when the EME API GetStatusForPolicy promise is resolved. Whether the frame is an ad frame. This is not set if the frame is null. The key system associated with the current CDM. The minimum HDCP version requested for EME API GetStatusForPolicy. Whether hardware secure codecs are required for the current CDM. xhwang@chromium.org media-dev@chromium.org Event recorded when RequestMediaKeySystemAccess() is called as part of Encrypted Media Extensions (EME) API. Whether the frame is an ad frame. Whether the frame is Cross-Origin to the main frame. Whether called by MediaCapabilities or not. Whether the frame is the main frame. The key system passed in requestMediaKeySystemAccess() call. Whether there are any "videoCapabilities". Whether there are any "videoCapabilities" with empty robustness. Whether there are any "videoCapabilities" with robustness being "HW_SECURE_ALL". jrummell@chromium.org media-dev@chromium.org Event recorded when an Encrypted Media Extensions (EME) API is called. The EME API that is called. For EME API calls done on MediaKeySession, whether the session is persistent or not. The key system associated with the current CDM. Whether hardware secure codecs are required for the current CDM. beccahughes@chromium.org media-dev@chromium.org The Media Engagement index stores the number of significant media playbacks per origin and the number of visits. From that we calculate a Media Engagement Score. To tweak the scoring function we are logging the total number of significant media playbacks, the total number of visits, the calculated engagement score and the new number of significant media playbacks that occurred this visit. Whether the Media Engagement Service considers the score to be high (we are using a two threshold approach so there is one threshold to be considered high and another one to lose that status to reduce jitter). The calculated Media Engagement score for the current origin. The score is calculated by dividing the number of significant media playbacks by the number of visits. If the number of visits is below 5 then the score will be zero. This score is taken from MediaEngagementService. The total number of significant media playbacks on this origin. The number of unique audio/video players on a page that was audible (made sound) during a visit. The number of unique audio/video players on a page that was audible (made sound) and considered significant (played for at least 7 seconds) during a visit. The total number of visits to this origin. beccahughes@chromium.org media-dev@chromium.org The Media Engagement index stores the number of significant media playbacks per origin and the number of audible players. From that we calculate a Media Engagement Score. Media with a short playback length is ignored so we are logging any time the player is ignored with the length in msec. This will allow us to identify whether sites are being penalized or there is abuse and allow us to tweak the length considered "short". beccahughes@chromium.org media-dev@chromium.org Media Feeds are out of band feeds of media recommendations that are fetched by the browser. This records when we discover a feed that has been placed in a meta tag on a website. Whether the site has a media feed. steimel@chromium.org media-dev@chromium.org Records when a user presses a button in the Global Media Controls. Records no more than 100 actions per source per session. Records which button was pressed (e.g. play, pause, next track, picture-in-picture). The action of the button that was pressed (e.g. play, pause, next track, picture-in-picture). liberato@chromium.org A record of some predicted value vs. an observation. If the prediction is for a regression-style task (e.g., "predict a number" rather than "predict an enum value"), then the predicted and observed values are the numeric values, scaled by default to be in the integer range 0-100. The exact scaling used depends on the particular regression problem. Please see the LearningTask structure for the task of interest to find out how it maps values into the [0, 100] output range. A record is created for every prediction separately. For example, in the MediaCapabilities experiments, each playback generates a record. The predicted / observed values are the ratio of dropped to decoded frames, expressed as a percentage scaled to 0-100. The learning task to which this record applies. This is a persistent hash of the |name| field in the LearningTask. Presumably, you'll want to include only records of a particular task when analyzing results. Observed value for this sample, scaled into the range [0, 100]. The exact scaling factor depends on the particular task. MediaCapabilities, for example, reports 100 * percentage_of_dropped_frames in this field. Predicted value for this sample, scaled into the range [0, 100]. The exact scaling factor depends on the particular task. MediaCapabilities, for example, reports 100 * percentage_of_dropped_frames in this field. Number of training examples that were used to train this model. Total weight of all examples that were used to train this model. steimel@chromium.org media-dev@chromium.org Event recorded when a website tries to play audio but is muted by the sound content setting. Enum value giving the reason the site was muted. Defined as |SoundContentSettingObserver::MuteReason|. chcunningham@chromium.org media-dev@chromium.org A record of decoding performance metrics from a video playback with the given stream characteristics (profile, resolution, fps). Also includes what Media Capabilities API would claim for such a stream as a means of assessing the API's accuracy. Boolean signaling whether MediaCapabilities would classify streams with these characteristics as power efficient prior to considering this latest record. MediaCapabilities prediction is accurate when this value matches RecordIsPowerEfficient. Boolean signaling whether MediaCapabilities would classify streams with these characteristics as smooth prior to considering this latest record. MediaCapabilities prediction is accurate when this value matches RecordIsSmooth. Integer count of past video frames decoded for the given video type. Integer count of past video frames dropped for the given video type. Integer count of past video frames power efficient for the given video type. Boolean signaling whether MediaCapabilities would classify this isolated playback record as power efficient. Boolean signaling whether MediaCapabilities would classify this isolated playback record as smooth. Integer count of the video frames decoded in this record. Integer count of the video frames dropped in this record. Should not exceed VideoFramesDecoded. Integer count of the video frames decoded via power efficient means in this record. Should not exceed VideoFramesDecoded. media::VideoCodecProfile enum value. Can be VIDEO_CODEC_PROFILE_UNKNOWN if we don't know the video codec. Name of the KeySystem used during for an encrypted (EME) playback. Will be unset when EME is not used. Boolean indicating whether content decryption module (CDM) was configured to use hardware secure decoding. May be true or false for EME playbacks. Will be unset for non-EME playbacks. Integer representing video frames per second. This is the cadence of video frames as described by stream timestamps multiplied by HTMLMediaELement.playbackRate. Flag indicating whether the record comes from the top frame or some inner frame. For privacy, metrics from inner frames are recorded with the top frame's origin, so this flag helps separate top frame vs. embedded playbacks. Integer representing height of video natural size. Integer representing width of video natural size. ID which corresponds to a given WebMediaPlayerImpl instance. May be linked with Media.WebMediaPlayerState events to understand playback more deeply. hiroh@chromium.org media-dev@chromium.org A record of encoding usage metrics in chrome, for example, for cast mirroring and webcodecs. This is recorded in the end of each video encoding session. Integer value indicating the height of the encoded stream. This is bucketed per 100 and capped by 8200. Boolean value indicating whether the encoder implementation is hardware accelerated one or not. Integer value indicating the number of the successfully encoded frames. This is bucketed per 100. media::VideoCodecProfile enum value. The codec profile configured to encode the video. media::EncoderStatus::Codes enum value. The first error status code on error, otherwise media::EncoderStatus::Codes::kOk. Integer value indicating the SVC configuration of the encoding. The value corresponds to media::SVCScalabilityMode. Integer value indicating the video encoder use case. The value corresponds to media:::mojom::VideoEncoderUseCase. Integer value indicating the width of the encoded stream. This is bucketed per 100 and capped by 8200. hongchan@chromium.org mjwilson@chromium.org webaudio-dev@chromium.org Records the AudioContext audible time information. Audible time in milliseconds for this event. Indicates whether the event is fired from main frame. hongchan@chromium.org mjwilson@chromium.org webaudio-dev@chromium.org Records the creation of a ScriptProcessorNode to monitor usage of this deprecated API. ScriptProcessorNode has been deprecated and developers are strongly encouraged to use AudioWorklet instead. This data helps gauge migration progress and the impact if it were to be removed. Fired and set to true when a ScriptProcessorNode is constructed. dalecurtis@chromium.org media-dev@chromium.org Final state of WebMediaPlayerImpl instance. Records only immutable playback properties. Contains a PlaybackID which links to Media.BasicPlayback events. Encryption type of the audio stream such as whether audio is encrypted and if so whether it has clear lead. media::container_names::MediaContainerName enum value. Only recorded for !IsMSE src=URL playbacks, this is the container of the media being played back; E.g., mp4, avi, mp3, etc. Type of media::Demuxer used for the playback. DemuxerType does not change dynamically during one playback, and is only determined after any HLS fallback might happen. media::PipelineStatus enum value. Always 0 if the playback succeeded; all other values indicate the playback ended in an error. Boolean value indicating whether an EME playback has been blocked waiting for the CDM to be set or waiting for decryption key. For an non-EME playback, this is always reported as false. Boolean value indicating if this event is for an EME playback. Note: EME can be attached anytime during the lifecycle of a WebMediaPlayerImpl, but once attached can't be removed. Boolean value indicating whether the EME playback is using hardware secure pipeline. For clear playback, this will always be false. Boolean value indicating if this event is for an MSE playback. If false it means this was a SRC playback. Flag indicating whether the report comes from the top frame or some inner frame. For privacy, metrics from inner frames are recorded with the top frame's origin, so this flag helps separate top frame vs. embedded playbacks. For EME playback, this is the key system used. For clear playback, this will always be the default value 0 (kUnknownKeySystemForUkm). ID which corresponds to a given WebMediaPlayerImpl instance. May be linked with Media.BasicPlayback events to understand a playback more deeply. Type of media::Renderer used for the playback. In most cases we do not support dynamically changing RendererType during one playback. Time in milliseconds from when WebMediaPlayerImpl starts loading until the first video frame has been shown. Time in milliseconds from when WebMediaPlayerImpl starts loading until metadata is known. Time in milliseconds from when WebMediaPlayerImpl starts loading until it has buffered enough to start playback. media::mojom::MediaURLScheme enum value. Only recorded for !IsMSE src=URL playbacks, this is the scheme of that URL; E.g., http, https, filesystem, etc. Encryption type of the video stream such as whether video is encrypted and if so whether it has clear lead. The visibility ratio of the video element at the time playback started, as a percentage from 0 to 100. 0 means that the element is not visible due to factors like being off-screen, clipped, occluded by other elements, or styled as invisible (e.g., opacity:0). 100 means that the element is fully visible. This value is currently set only for encrypted media using hardware secure codecs, and if the kEncryptedMediaOcclusionTracking flag is enabled. rwkeane@chromium.org openscreen-eng@google.com Records that the a resource has been loaded from the Media Router Component Extension. A boolean signaling that the event has occurred; only records true values. muyaoxu@google.com openscreen-eng@google.com Recorded whenever user starts a site-initiated mirroring session via Media Router. Whether audio capture is allowed in the cast session. If allowed, audio playback will happen on the receiver side. If not allowed, audio playback will happen on the sender side. takumif@chromium.org openscreen-eng@google.com Recorded whenever user starts a tab mirroring session via Media Router. The audio playback state of the WebContents at the start of the tab mirroring session. mfoltz@chromium.org openscreen-eng@google.com Events recorded when a page requests access device media, i.e. by starting screen capture or opening a webcam or microphone, and the result of that request. The result of a page's request to open a video capture device (e.g., webcam). The suffix 3 was added when the MediaStreamRequestResult enum was updated with new entries. erikchen@chromium.org ssid@chromium.org Metrics associated with memory consumption, in MB. Measure of memory consumed by Array Buffer. Measure of memory consumed by Oilpan. Measure of memory consumed by live allocations made from Oilpan. Measure of memory consumed by GL command buffer. Measure of memory consumed by Discardable memory service. Measure of memory used by download service. Measure of memory consumed by Key Value Store databases of extensions. Measure of memory used by font platform and shape caches in renderer process. Measure of GPU memory used by Chrome. Approximate measure of memory consumed by History service. Indicates whether the tab is visible or not at the time of metric collection. Measure of memory consumed by java heap on Android. Measure of memory consumed by unaccounted level databases. Measure of memory allocated by malloc, that is not classified by other metrics. Measure of total memory used by objects allocated using malloc. The number of ad subframes that the associated renderer owns. The number of arraybuffer contents that the associated renderer owns. The number of detached ScriptState that the associated renderer owns. The number of documents that the associated renderer owns. The number of extensions that are served from the associated renderer process. The number of frames that the associated renderer owns. The number of layout objects that the associated renderer owns. The number of mojo handles stored in HandleTable. The number of nodes that the associated renderer owns. The number of web media players in the associated renderer. Measure of memory used due to URL indexing and autocomplete suggestions. Measure of memory allocated by PartitionAlloc allocator. Measure of total memory used by objects allocated using PartitionAlloc. Measure of memory used by Array Buffer partition in PartitionAlloc. Measure of memory used by Buffer partition in PartitionAlloc. Measure of memory used by Fast Malloc partition in PartitionAlloc. Measure of memory used by Layout partition in PartitionAlloc. Measure of total private memory consumed by process. Measure of private swap memory consumed by a process. Available on Linux and Android. Type of process (e.g. browser, renderer, GPU --- see services/resource_coordinator/public/mojom/memory_instrumentation/memory_instrumentation.mojom) of associated metrics. Size of process' working set. Measure of total shared memory consumed by process. Measure of memory used due to web storage APIs in browser process. Measure of memory used by in-memory blob file API in browser process. Measure of memory used due to IndexedDB API in browser process. Measure of memory used due to Local Storage API in browser process. Measure of memory used due to Session Storage API in browser process. Measure of memory used by Skia. Measure of memory used by Skia Glyph Cache. Measure of memory used by Skia Resource Cache. Measure of memory used by all sqlite databases. Measure of memory used by Sync storage. Approximate measure of memory used by Tab restore service. The time in seconds since the Tab navigated. Only emitted for tabs that are not sharing a process. The time in seconds since the Tab changed visibility. Only emitted for tabs that are not sharing a process. Measure of total private memory consumed by all processes. Measure of total shared memory consumed by all processes. Measure of memory used by Android UI bitmaps. Process uptime. Measure of memory consumed by V8. Measure of memory consumed by live objects in V8. Measure of memory consumed by the main isolate of V8. Measure of memory consumed by live objects in the main isolate of V8. Measure of memory consumed by global handles in V8. Measure of memory consumed by used global handles in V8. Measure of memory consumed by the main heap of V8. Measure of memory consumed by live objects in the main heap of V8. Measure of memory consumed by the code large object space of the main heap of V8. Measure of memory consumed by live objects in the code large object space of the main heap of V8. Measure of memory consumed by the code space of the main heap of V8. Measure of memory consumed by live objects in the code space of the main heap of V8. Measure of memory consumed by the large object space of the main heap of V8. Measure of memory consumed by live objects in the large object space of the main heap of V8. Measure of memory consumed by the map space of the main heap of V8. Measure of memory consumed by live objects in the map space of the main heap of V8. Measure of memory consumed by the new large object space of the main heap of V8. Measure of memory consumed by live objects in the new large object space of the main heap of V8. Measure of memory consumed by the new space of the main heap of V8. Measure of memory consumed by live objects in the new space of the main heap of V8. Measure of memory consumed by the old space of the main heap of V8. Measure of memory consumed by live objects in the old space of the main heap of V8. Measure of memory consumed by the read-only space of the main heap of V8. Measure of memory consumed by live objects in the read-only space of the main heap of V8. Measure of memory consumed by the shared large object space of the main heap of V8. Measured in bytes. Measure of memory consumed by live objects in the shared large object space of the main heap of V8. Measured in bytes. Measure of memory consumed by the shared space of the main heap of V8. Measured in bytes. Measure of memory consumed by live objects in the shared space of the main heap of V8. Measured in bytes. Measure of memory allocated by malloc in the main isolate of V8. Measure of memory consumed by worker isolates of V8. Measure of memory consumed by live objects in worker isolates of V8. Measure of memory consumed by all resources in Blink Web Cache. Measure of memory consumed by CSS Stylesheet resources in Blink Web Cache. Measure of memory consumed by data URLs of Image resources in Blink Web Cache. Measure of memory consumed by Font resources in Blink Web Cache. Measure of memory consumed by Image resources in Blink Web Cache. Measure of memory consumed by other resources in Blink Web Cache. Measure of memory consumed by Script resources in Blink Web Cache. Measure of memory consumed by V8 code cache held by Script resources in Blink Web Cache. Measure of memory consumed by XSL Stylesheet resources in Blink Web Cache. Measure of total memory used by media/webmediaplayer/audio. Measure of total memory used by media/webmediaplayer/data_source. Measure of total memory used by media/webmediaplayer/demuxer. Measure of total memory used by media/webmediaplayer/video. Measure of the backing memory consumed by the WebNN service. tommckee@chromium.org Measure of memory used by the processes that host an instance of a tab. Intended as a high-level metric for analyzing effective memory use of individual sites. Measure of private memory, in MB, consumed by the render process dedicated to hosting the main frame. Undefined if the main frame is hosted by a render process used by other tabs. If this is undefined, TabPMF will be undefined. Number of render processes that were blocked from contributing to SubFrameProcessPMF.Total. A process can have its contribution blocked if the process has responsibilities outside the scope of the relevant tab. If this is non-zero, TabPMF will be undefined. Number of render processes that contributed to SubFrameProcessPMF.Total. Measure of total private memory, in MB, consumed by the render processes dedicated to hosting sub-frames of the tab. Note that, if a render process hosts frames of other tabs, it isn't considered to be 'dedicated' and so it doesn't contribute to this total. Measure of private memory, in MB, consumed by all render processes dedicated to hosting some part of a tab. Undefined if there are processes responsible for hosting parts of this tab while hosting parts of other tabs. carlosil@chromium.org Status and Network error or HTTP response code for a resource request that was autoupgraded to HTTPS as part of the mixed content autoupgrade experiment. The HTTP response or network error code for an autoupgraded request. An enum with 0 representing started, 1 failed, and 2 response received. kumagi@google.com gaul@google.com Whether the page allows the user to zoom in/out. Percentage of small font text area in total text area. Percentage of pixels of text and images horizontally outside the viewport, relative to the frame width. Whether the width of the viewport specified as device-width or not. Specified hardcoded viewport width in CSS pixels. Specified initial viewport scaling multiplied by 10. 1 means 0.1, 100 means 10. [1-100]. kumagi@google.com gaul@google.com The distance from the center position of tapped target to display's edge is less than 5mm. Whether the page is manually 100%+ zoomed when tapped. Whether the center of the tapped tap target is close less than 5mm to other tap target's edge. Whether the tapped tap target is small as both width and height is less than 7mm. Whether the page is manually 20%+ zoomed when tapped. sayman@chromium.org User pressed 'Share' in the app menu. The intent most recently used to share was invoked directly. See also MobileMenu.Share. A boolean signaling that the event has occurred; only records true values. frechette@chromium.org User pressed 'Find in page' in the app menu. A boolean signaling that the event has occurred (typically only records true values). sayman@chromium.org User pressed 'Share' in the app menu. A picker was shown. See also MobileMenu.DirectShare. A boolean signaling that the event has occurred; only records true values. jihadghanna@google.com chrome-multistep-filter-eng@google.com Recorded when a Multistep Filter suggestion application session is concluded. The session starts when a suggestion is accepted. If the application is unsuccessful, the session ends immediately. If successful, the session continues to track post-application user engagement, ending when the user navigates away or closes the tab. This event records the final application outcome (represented by SuggestionApplicationResult enum) and any post-application engagement metrics. The overall technical filter application outcome (AllFiltersApplied, AbandonedBeforeVerification, or various Failure reasons) after a user accepts a Multistep Filter suggestion. The number of filter facets successfully applied (clamped to multistep_filter::kMultistepFilterMaxFacetsShownUkmClampingLimit). Logged only if outcome is AllFiltersApplied. The user's engagement after the filters were applied (e.g. further navigation, tab close, omnibox navigation) distinguished by whether it happened within the multistep_filter::kMultistepFilterPostApplicationSessionDuration window. The user's retention history state at the time the suggestion was shown (must match the UI session). A transient, randomly generated identifier used to correlate the UI session event (MultistepFilter.UiSession) with the subsequent landing page application session event (MultistepFilter.ApplicationSession). A bitmask representing the types of filter facets in the accepted suggestion. Each bit N corresponds to a filter facet type with enum value N defined in MultistepFilterFacetType enum (e.g. Bit 1 is kAgeChild (1), Bit 11 is kDateCheckin (11), Bit 0 is unused). This must match the UI session. Time from the user accepting the suggestion to when the filters are successfully applied on the landing page (in milliseconds, exponentially bucketed). Logged only if outcome is AllApplied. The task type of the applied suggestion (must match the UI session). jihadghanna@google.com chrome-multistep-filter-eng@google.com Recorded when a Multistep Filter suggestion UI session is flushed (concluded). This event captures the full lifecycle of the UI session, including whether the suggestion was accepted, dismissed, or ignored, and the associated latencies and age of the suggestion. Measures the time elapsed (in minutes, clamped to multistep_filter::kMultistepFilterSessionDuration) from when the potential suggestion was extracted to when the user accepted it. Logged only if the suggestion was accepted. True if the suggestion was shown on the same eTLD+1 as the page from which it was extracted. Time from the end of the navigation that triggered the suggestion generation to when the user accepted it (in milliseconds, exponentially bucketed). Logged only if accepted. Time from the end of the navigation that triggered the suggestion generation to when it was first shown (in milliseconds, exponentially bucketed). The number of filter facets which were part of the suggestion which was shown (clamped to multistep_filter::kMultistepFilterMaxFacetsShownUkmClampingLimit). True if the suggestion cue was reopened from the Omnibox. The user's retention history state at the time the suggestion was shown. A transient, randomly generated identifier used to correlate the UI session event (MultistepFilter.UiSession) with the subsequent landing page application session event (MultistepFilter.ApplicationSession). Measures the time elapsed (in minutes, clamped to multistep_filter::kMultistepFilterSessionDuration) from when the potential suggestion was extracted to when it was first shown to the user. A bitmask representing the types of filter facets in the suggestion shown to the user. Each bit N corresponds to a filter facet type with enum value N defined in MultistepFilterFacetType enum (e.g. Bit 1 is kAgeChild (1), Bit 11 is kDateCheckin (11), Bit 0 is unused). Time from when the suggestion was first shown to when the user accepted it (in milliseconds, exponentially bucketed). Logged only if accepted. The task type of the suggestion (e.g. Flight, Accommodation, etc). The user's final decision on the suggestion (Accepted, Dismissed, Settings Opened, Ignored). annasato@chromium.org chrome-navigation@google.com Records the number of subsequent duplicate navigation requests that were ignored in favor of an ongoing navigation. This is caused by IgnoreDuplicateNavs. The number of duplicate navigations that were ignored. ukm::GetExponentialBucketMinForCounts1000() is applied. suzukikeita@chromium.org chrome-loading@google.com Metrics recorded each time a navigation from Google Search is abandoned (e.g. the navigation got cancelled, or the user backgrounded the app while the page load is not finished yet, etc. See AbandonedPageLoadMetricsObserver for more details). The reason the navigation/page load gets abandoned. The time when the navigation/page load is abandoned, relative to the last milestone. The time when the navigation/page load is abandoned, relative to navigation start. The last milestone before the navigation is abandoned. The `net::Error` code for this page load. Only logged for failed provisional navigations (i.e. abandoned before commit). The extended error code for the net error for this page load, which represents the detailed context of why the net error happened. Only logged for failed provisional navigations. If we've experienced backgrounding while the navigation/page load is running before this abandonment event happened, this will have the time when the backgrounding happened, relative to navigation start. If we've experienced hiding while the navigation/page load is running before this abandonment event happened, this will have the time when the hiding happened, relative to navigation start. The time when the final renderer process for the navigation is initialized, relative to navigation start. suzukikeita@chromium.org chrome-loading@google.com Records the navigation milestone timing information. This is triggered either when we are navigating away from the page (i.e. `OnComplete` of the `PageLoadMetricsObserver`) or when the navigation is being abandoned. For non-terminal abandonments such as tab hidden or app backgrounded, we record this field every time the abandonment happens so that we have an entry recorded even if app / tab is killed. Note that there may be more than one entry per navigation if non-terminal abandonment is involved, as a terminal abandonment may happen after recording and also non-terminal abandonment may happen multiple times within a navigation, and we may pass succeeding milestones after the last record. An 8-bit integer that can be manually mapped to a specific URL param "category=" value. Note that only the number is logged. The time when the CommitNavigation IPC for the navigation is received in the renderer process, relative to navigation start. The time when the CommitNavigation IPC for the navigation is sent from the browser process, relative to navigation start. The time when the DidCommit IPC for the navigation is received, relative to navigation start. The timestamp of when the "DOMContentLoaded" event is fired for the page load, relative to navigation start. The timestamp of the first contentful paint for the page load, relative to navigation start. The time when the first network request that got redirected for the navigation started, relative to navigation start. Whether the first redirect response is received yet. The delay in milliseconds caused by connection setup for the first HTTP request which is sent for the main resource of a main frame navigation. The delay in milliseconds caused by domain lookup for the first HTTP request which is sent for the main resource of a main frame navigation. The delay in milliseconds caused by SSL handshake for the first HTTP request which is sent for the main resource of a main frame navigation. Determines whether the navigation is committed or not. The timestamp of the largest contentful paint for the page load, relative to navigation start. The time when the URLLoader for the navigation started, relative to navigation start. The timestamp of when the "load" event is fired for the page load, relative to navigation start. The time when the final network request that didn't got redirected for the navigation started, relative to navigation start. Whether the non-redirect/final response is received yet. The time when the parsing of the HTTP response body started, relative to navigation start. The total number of redirects that happened within the navigation. If the navigation is abandoned before commit, we record the number of redirects up until that point. The time when the second network request that got redirected for the navigation started, relative to navigation start. Whether the second redirect response is received yet. meacer@google.com chrome-trusty-transport@google.com Recorded upon main frame navigations to hostnames that contain characters that are treated differently between IDNA 2003 and IDNA 2008, aka deviation characters. Ignores reload navigations. The deviation character in the URLs eTLD+1. Chrome's URL parsing code normalizes deviation characters so the characters don't appear in the recorded URL. This is recorded to make the deviation character identification simpler. miketaylr@chromium.org katabolism-finch@google.com Recorded when the navigation receives a response from the networking layer. Only recorded for main frames (i.e. top-level frames). A boolean value representing whether the navigation included the processing of an ACCEPT_CH ALPS frame. The duration (in milliseconds) for how long it took from the time the navigation loader is created to the time the navigation loader receives the response from the networking layer. koretadaniel@chromium.org chrome-desktop-ui-waterloo@google.com Recorded immediately after a successful navigation completes if the user explicitly navigates to a domain via the Omnibox. Recorded as 1 if the explicitly typed search engine destination matches the user's currently active Default Search Engine (DSE), and 0 if it differs. ryansturm@chromium.org Metrics that describe an anchor element's data. This event is emitted at most 10 times per page load, once for each of the top n URLs on a page, and the order of the events is randomized. If there is no nth anchor element, no value is returned. The index of the clicked-on anchor element in the tracked list. The anchor element's href URL path is hashed, then bucketed using 10 different buckets. The bucket ({0, 1, ..., 9}) the hash falls into is logged in this field. 1 if the anchor element contains an image. The bucketed font size of the anchor element, according to its computed style. 1 corresponds to a pixel font size in [0, 10), 2 corresponds to a fonts size in [10, 18) and 3 is logged when the font size is 18 pixels or larger. 1 if the anchor element has an immediate text sibling, 0 otherwise. 1 if the anchor's computed style's font-weight is greater than 500, 0 otherwise. 1 if the anchor element is in an iframe. 1 if the anchor element's target URL and the document's source URL only differ by one number, and the target URL increments that number in the source URL by 1. The time in ms between navigation start and the moment this entry was logged. Exponentially bucketed using GetExponentialBucketMin a value of 1.3. The number of slashes in the anchor element's href URL path. Truncated at 5. The number of characters in the anchor element's href URL path. Buckets of size 10, truncated at 100. The percent of the total document area that this anchor element's clickable area occupies. How far down the document the top of the anchor element is, expressed as a ratio with the total document height. 1 if the anchor element's target URL and the document's source URL are the same origin. ryansturm@chromium.org Sent on an in-page link click. An integer (0-indexed), corresponding to the index of that anchor element in the list of top ten anchor elements on that page, sorted by their navigation scores. If an anchor element is clicked that is not in NavigationPredictorAnchorElementMetrics, -1 is returned. 1 if the href of the anchor element at the time the click is handled is the same as the href when the element was created. 0 if the href changed. The time in milliseconds between the moment an entry for the anchor element was logged to UKM and the moment the user clicked on it. Exponentially bucketed, 1.3 factor. ryansturm@chromium.org Metrics that describe aggregate information about the links within a page. Some anchor elements present on a page may not be represented in these metrics, if they are outside the viewport or if their placement in the page is after the 40th anchor element. This event is emitted at most once per page load. Unless otherwise specified, each of the aggregate metrics is bucketted with exponential buckets of size 1.3. The median vertical link location in the viewport, as a ratio of the distance from the top of the link to the visible top of the page. Bucketted with a linear bucket size of 10. The total number of links in the page that are images. The total number of links in the page in any iframe. The total number of links in the page that point to the same origin as the current page's origin. The total number of links within a page. The total number of links on the page, where each link only differs from the current page's URL by a number, and that number is incremented by 1 in the link. The total clickable space in the entire page, expressed as a ratio of the visible clickable region to the entire viewport region. The height of the viewport, in pixels. The width of the viewport, in pixels. ryansturm@chromium.org domenic@chromium.org isaboori@google.com This event records required data to assess and fine-tune the preloading on-hover and on pointer down heuristics, and is submitted once per mouse over, mouse out, mouse down and mouse up events for randomly selected anchor elements. Duration that the user was hovering over the anchor element and it did not result in a navigation to that link. Exponentially bucketed using GetExponentialBucketMin with a spacing of 1.3. Duration that the user was hovering over the anchor element and it resulted in a navigation to that link. Exponentially bucketed using GetExponentialBucketMin with a spacing of 1.3. Time since the user pressed the left/middle mouse button over the anchor element and that did not result in a navigation to that link. Exponentially bucketed using GetExponentialBucketMin with a spacing of 1.3. Time since the user presses the left/middle mouse button over the anchor element and that resulted in a navigation to that link. Exponentially bucketed using GetExponentialBucketMin with a spacing of 1.3. ryansturm@chromium.org isaboori@google.com Metrics that describe user interaction with an anchor element. This event is emitted once per randomly selected anchor elements. The index of the anchor element in NavigationPredictor's |top_urls_|. Vertical distance from the center of the anchor element to the most recent pointerdown that started a scroll, as a percentage of screen height. This value is between -100 and 100 (inclusive), and is negative when the pointerdown was below the anchor's center (positive otherwise). This will only be set for anchor elements that are in the viewport. Linearly bucketed with a bucket size of 10. How many times the anchor element entered the viewport. Exponentially bucketed using GetExponentialBucketMin a value of 1.3. 1 if anchor element is in viewport, otherwise 0. 1 if the pointer is hovering over the anchor element, otherwise 0. Maximum time between the moment the anchor element entered the viewport and the moment it left the viewport in milliseconds. Exponentially bucketed using GetExponentialBucketMin a value of 1.3. Maximum dwelltime of pointer hovering over the anchor element in milliseconds. Exponentially bucketed using GetExponentialBucketMin a value of 1.3. Mouse acceleration when the OnHover event triggered. Exponentially bucketed using GetExponentialBucketMin a value of 1.3. Mouse velocity when the OnHover event triggered. Exponentially bucketed using GetExponentialBucketMin a value of 1.3. How many times the pointer was hovering over the anchor element. Exponentially bucketed using GetExponentialBucketMin a value of 1.3. Vertical position of the center of the anchor element (y-coordinate of the center), as a percentage of the viewport's height. This value is between 0 and 100 (inclusive). This will only be set for anchor elements that are in the viewport. Linearly bucketed with a bucket size of 10. chikamune@chromium.org hayato@chromium.org chrome-loading@google.com Metrics to measure the main frame navigation timings by tracking NavigationRequest in the browser process. This is recorded at the frequency of 0.001. Measures the duration from BeginNavigation to LoaderStart, in milliseconds. Measures the duration from FetchStart to RequestStart, in milliseconds. Measures the duration from LoaderStart to FetchStart, in milliseconds. Measures the duration from LoaderStart to ReceiveResponse, in milliseconds. Measures the duration from NavigationStart to BeginNavigation, in milliseconds. Measures the duration from ReceiveResponse to CommitNavigation, in milliseconds. ryansturm@chromium.org chrome-brapp-loading@google.com Information recorded about a navigation throttle when a navigation is deferred at any stage of the navigation. Only recorded for main frame navigations that were deferred and later resumed. The duration from the start of the navigation deferral until the throttle resumes the navigation recorded in milliseconds. An enum representing each event that can cause deferral method (WillFailRequest, WillProcessResponse, WillRedirectRequest, WillStartRequest). See NavigationThrottleRunner::Event for enum details. A (MD5) hash of the logging name string which represents a unique navigation throttle class (subclass of NavigationThrottle). See NavigationThrottle::GetNameForLogging(). hayato@chromium.org chrome-loading@google.com Metrics to measure the navigation timings. These are recorded for 1 out of 1,000 navigations. These metrics are based on the equivalent UMA metrics defined under Navigation.Timeline.{DurationFromTo}.Duration. The start and end of beforeunload phase 1 in the initiating renderer process, if that phase runs in a given navigation. When beforeunload phase 1 is finished to when the NavigationRequest is created. The start and end of beforeunload phase 2 in the browser process, if that phase runs in a given navigation. When beforeunload phase 2 is finished to when NavigationRequest::BeginNavigation() is called. When NavigationRequest::BeginNavigation() is called to start the navigation, to when the CommitNavigation message is sent. This is only logged for navigations that don't go through the network stack, such as about:blank navigations or browser-initiated same-document navigations. When NavigationRequest::BeginNavigation() is called to when the NavigationURLLoader starts. Represents the browser process initialization work between beforeunload processing and starting the network request. When the CommitNavigation IPC is sent to the renderer process to when its reply is received by RenderFrameHostImpl::DidCommitNavigation. When the browser process receives the DidCommitNavigation IPC to when the browser process finishes processing that IPC, making the navigation committed. When fetch of the document via an HTTP(S) request starts, to when the final response headers are received in the network service. This can include redirects. Reports how much time from beforeunload phases 1 and 2 was correctly excluded from the Navigation.Timeline.TotalExcludingBeforeUnload.Duration metric, so that beforeunload handlers and dialogs would not be counted as part of the navigation. Recorded at the point when a navigation finishes committing, i.e. finishes processing the DidCommitNavigation message. A corresponding event is also emitted for tracing. Reports how much time navigation guardrail metrics like FCP are incorrectly excluding from the navigation duration, due to collecting start times after beforeunload phase 1 finishes and possibly adjusting them to after beforeunload phase 2 finishes. This excluded time represents actual time the user is waiting that is not accounted for in current metrics. Recorded at the point when a navigation finishes committing, i.e. finishes processing the DidCommitNavigation message. A corresponding event is also emitted for tracing. The duration in milliseconds from a) the OS-level timestamp of the user input event that started a navigation to b) when the navigation starts. This metric is recorded only for cross-document navigations where a user interaction timestamp is available. The duration in milliseconds from a) the OS-level timestamp of the user input event that started a navigation to b) when the navigation finishes committing. This metric is recorded only for cross-document navigations where a user interaction timestamp is available. The duration in milliseconds from a) the OS-level timestamp of the user input event that started a navigation to b) when the navigation finishes, excluding the time spent running beforeunload handlers/dialogs. This metric is recorded only for cross-document navigations where a user interaction timestamp is available. When the NavigationURLLoader starts, to when the browser is ready to fetch the document using an HTTP(S) request. When the NavigationURLLoader starts to when NavigationRequest::OnResponseStarted() is called. When the NavigationRequest is created to when beforeunload phase 2 is started in the browser process. Includes time to pick a target RenderFrameHost, SiteInstance, and process. When the NavigationRequest is created to when NavigationRequest::BeginNavigation() is called. Logged only when no beforeunload handler processing occurs (otherwise, see NavigationRequestToBeforeUnloadPhase2Duration and BeforeUnloadPhase2ToBeginNavigationDuration). Includes time to pick a target RenderFrameHost, SiteInstance, and proces. When the final response headers are received in the network service to when NavigationRequest::OnResponseStarted() is called in the browser process. When NavigationRequest::OnResponseStarted() is called to when the CommitNavigation IPC is sent to the renderer process. When the renderer process receives a CommitNavigation IPC to when the renderer process sends a DidCommitNavigation reply. The time when the renderer process was created. The time when the renderer process was launched. When the navigation starts, which could be in the renderer process, to when beforeunload phase 1 starts in that process. When the navigation starts, which could be in the renderer process, to when the NavigationRequest is created. When the navigation starts in the renderer process, to when the renderer process begins a synchronous renderer commit. When the navigation starts to when the browser process finishes processing the DidCommitNavigation IPC, at which point the navigation is committed. Includes the time spent running any beforeunload handlers/dialogs. When the navigation starts to when the browser process finishes processing the DidCommitNavigation IPC, at which point the navigation is committed. Excludes the time spent running any beforeunload handlers/dialogs. nhiroki@chromium.org chrome-loading@google.com Metrics associated with main frame navigation. The time relative to navigation start that a callback for the navigation loader is last invoked for the main resource of a main frame navigation. The time relative to navigation start that the final HTTP request is sent for the main resource of a main frame navigation. The time relative to navigation start that the headers of the final HTTP response are received for the main resource of a main frame navigation. The time relative to navigation start that a callback for the navigation loader is first invoked for the main resource of a main frame navigation. The time relative to navigation start that the first HTTP request is sent for the main resource of a main frame navigation. The time relative to navigation start that the headers of the first HTTP response are received for the main resource of a main frame navigation. The time relative to navigation start that navigation commit message for a main frame navigation is sent to a renderer process. nidhijaju@chromium.org chrome-loading@google.com Recorded per resource request when parsing for a data URL is completed. If the data URL is for an image resource. The time taken (ms) to parse the data URL. This includes base64 decoding and building the response. The length of the data URL, indicating the size of the resource. sophiechang@chromium.org tiborg@chromium.org Metrics related to a History Clusters module shown on the New Tab Page. Recorded for each History Clusters module shown on a 1P New Tab Page load, which can be multiple for a page load. Cluster is of a boosted category. Cluster is of the most frequently seen category. Cluster is of the most frequently used category. Whether the user disabled the module. Whether the user dismissed the specific journey shown in this module. Whether the user engaged with the module. Whether the user dismissed with positive intent the cluster shown in this module. The layout type that was shown to the user for this module. Minutes since the shown cluster's most recent visit, exponentially bucketed using GetExponentialBucketMin with a spacing of 1.3. The number of times the most frequently seen category by a user occurred in a fixed time period. The number of times the most frequently used category by a user occurred in a fixed time period. Number of abandoned carts associated with the cluster. Number of categories associated with the cluster. Number of times the cluster was seen by the user in the last 24 hours. Number of times the cluster was used by the user in the last 24 hours. Number of total visits in the same cluster including ones that are not necessarily shown in the module. Number of unique hosts represented in the cluster. Number of visits with an image to show. tbansal@chromium.org Metrics related to NoStatePrefetch that are recorded using the same UKM IDs as PageLoad. Final status of the nostate prefetch if one was recently attempted for either the committed URL of this navigation or for any URL in the redirect chain for the main frame resource for this navigation. page. Recorded as enum PrerenderFinalStatus in //tools/metrics/histograms/enums.xml. Triggering origin of the nostate prefetch if one was recently attempted for either the committed URL of this navigation or for any URL in the redirect chain for the main frame resource for this navigation. Recorded as enum PrerenderOrigin in //tools/metrics/histograms/enums.xml. Records the time (in milliseconds) from the start of the nostate prefetch to the time of commit of this navigation. Metric is recorded down to the nearest power of 2. Recorded only if a nostate prefetch was recently attempted for either the committed URL of this navigation or for any URL in the redirect chain for the main frame resource for this navigation. peter@chromium.org A notification event corresponds with the entire lifespan of a single notification, and will be logged to UKM when the notification is deleted from the notification database. Enum for how the notification was closed. {user, developer, unknown} Boolean value representing whether a tag is present and a previously shown notification has been replaced. For UKM logging, we treat a replacement notification as a new notification. Boolean value indicating whether the user accessed the Chrome provided settings from the notification. Boolean value representing if a badge URL was included. Boolean value representing if an icon URL was included. Boolean value representing if an image URL was included. Boolean value representing whether default notification indicators (sound, vibration, light) should be played again if the notification is replacing an older notification. Boolean value representing if a tag is present. Boolean value representing whether default notification indicators (sound, vibration, light) should be suppressed. Number of clicks on developer provided action buttons. Number of developer specified actions that can be taken. Number of clicks, excluding action buttons. I.e. clicks on the notification itself. Boolean value representing whether the notification should remain onscreen indefinitely, rather than being auto-minimized to the notification center (if allowed by platform). The length of time, in ms, between when the notification was triggered and when it was closed. Time, in ms, between when the notification is shown and the first click. Time, in ms, between when the notification is shown and the last click. nyquist@chromium.org ssid@chromium.org User pressed 'Share' while viewing edit URL suggestions which shares the current URL. A boolean signaling that the event has occurred; only records true values. mahmadi@google.com chrome-desktop-search@google.com Recorded when the user uses the omnibox to search or navigate. The type of page the user was on when they used the omnibox. Corresponds to the PageClassification enum in //third_party/metrics_proto/omnibox_event.proto The provider type the selected suggestion came from. Corresponds to the ProviderType enum in //third_party/metrics_proto/omnibox_event.proto. The type of the suggestion the user selected in the omnibox. Corresponds to the ResultType enum in //third_party/metrics_proto/omnibox_event.proto. A value that indicates if the user searched or navigated from the omnibox. Corresponds to the ClientSummarizedResultType enum in //components/omnibox/browser/omnibox_metrics_provider.h. The index of the suggestion that the user selected in the omnibox popup. The amount of time, in milliseconds, since the user last focused the omnibox until they searched or navigated from the omnibox. Bucketed into exponential intervals with a standard bucket_spacing of 2.0. The number of characters the user typed, excluding autocompletion. The amount of time, in milliseconds, since the user first began modifying the text in the omnibox until they searched or navigated from the omnibox. Bucketed into exponential intervals with a standard bucket_spacing of 2.0. Whether at least one zero-prefix Contextual Search suggestion was shown in the Omnibox session. Whether at least one zero-prefix Lens Action suggestion was shown in the Omnibox session. Whether at least one zero-prefix Search suggestion was shown in the Omnibox session. Whether at least one zero-prefix URL suggestion was shown in the Omnibox session. meacer@chromium.org Recorded when a form is submitted on an HTTPS page. Used to understand any regression introduced by updated omnibox security indicators. Placeholder metric, always true. sophiechang@chromium.org Optimization guide metrics associated with a page load. It is recorded when the optimization guide was consulted in making a decision for the page load. See //components/optimization_guide/. An enum value representing the status of a fetch attempt for the page load and what it contains (i.e. hosts or URLs). Captured at the start of the navigation. The duration (in milliseconds) for how long it took from the request to go out to when it was usable on the client. Only recorded if a fetch was initiated for the navigation. Will be recorded as INT64_MAX if the request did not complete successfully. A bitmask of the optimization types that were registered at the start of the page load where each flipped bit corresponds to its int value in |optimization_guide::proto::OptimizationType|. chrome-intelligence-core@google.com khushalsagar@chromium.org abigailklein@chromium.org gklassen@chromium.org Metrics related to page content extraction. Only recorded when depth limit is exceeded. Indicates that the node depth limit was exceeded while building AnnotatedPageContent. chrome-intelligence-core@google.com rajendrant@chromium.org sophiechang@chromium.org khushalsagar@chromium.org Metrics on the page content extracted from all frames in the page. The extraction can happen multiple times during the pageload to re-compute the changes, and each extraction will record the metrics. Page content extraction latency. The time taken to extract across all frames in the page. The mode used for page content extraction. Number of nodes in the page content. Whether the page content extraction was on the critical path. Total size of the extracted page content. Number of words in the web page. The searchable text (visible and invisible folded text) in the page is broken down by whitespace to get the word count. chrome-intelligence-core@google.com rajendrant@chromium.org sophiechang@chromium.org Metrics on the PDF pages. Recorded only for PDF pages, if a PDF renderer is attached to the main frame. Number of pages in the PDF. chrome-intelligence-core@google.com linnan@chromium.org Metrics related to the page settling and stability detection during page content extraction. Recorded once per committed primary main frame navigation when the page first becomes stable and ready for extraction. Whether the navigation was a same-document navigation. The duration in milliseconds from when a main frame navigation commits until the page is considered stable and ready for extraction. chrome-intelligence-core@google.com mcrouse@chromium.org Metrics on the content annotations of the page, one record per page load commit and has an annotation calculated on it. The metrics are recorded in a RAPPOR-style with low-entropy so only aggregate analysis can be performed. Records a low entropy representation of the education category score from the output of a page classifier model. The raw score is between 0 and 1, which is scaled to between 0 and 100, then placed into one of 16 linearly spaced buckets by dividing the score by the bucket size and taking the floor. The resulting bucket value has noise added with a probability of .5 where the noise is randomly applied to each individual bit. Records a low entropy representation of the shopping category score from the output of a page classifier model. The raw score is between 0 and 1, which is scaled to between 0 and 100, then placed into one of 16 linearly spaced buckets by dividing the score by the bucket size and taking the floor. The resulting bucket value has noise added with a probability of .5 where the noise is randomly applied to each individual bit. Records a low entropy respresentation of the visibility score output by a content annotations model. The low entropy value is calculated using a RAPPOR-style calculation to enable analysis only on aggregated results. The raw score is between 0 and 1, scaled to between 0 and 100. The score is then placed into one of 16 linearly spaced buckets by dividing the score by the bucket size and taking the floor. The resulting bucket value has noise added with a probability of .5 where the noise is randomly applied to each individual bit. uthakore@chromium.org invernizzi@chromium.org Metrics that describe the domain of pages that successfully commit. One event is generated per page load that commits. Currently associated with LocalNetworkRequests UKM metric collection. An enum value representing the type of domain of the loaded page. The enum is defined in |LocalNetworkRequestsPageLoadMetricsObserver|. Possible values are 0 for unknown (should never be logged), 1 for pages with public domains, 2 for pages with private domains (IP addresses that are part of a reserved IP space not including localhost), or 4 for localhost. sullivan@chromium.org Total time in foreground in milliseconds, recorded each time the page is backgrounded. May be recorded multiple times for a single page visit. Total time in foreground in milliseconds, recorded each time the page is backgrounded. sum(Input Delay), recorded each time the page is backgrounded. The Input Delay refers to the duration between the hardware timestamp and the start of event processing on the main thread for the meaningful input. In ms. dullweber@chromium.org chrome-privacy-core@google.com Recorded when an action on the Page Info Bubble (aka Origin Info Bubble or WebsiteSettings Bubble) is taken. May be recorded multiple times for a single page visit. Records an enum value representing which action on the Page Info Bubble was taken. chikamune@chromium.org chrome-loading@google.com Metrics for the previous page's PageLifecycle events on the new page's commit. This event is saved on the new page that is about to commit. Measures the total duration in milliseconds to run event handlers for the previous page's lifecycle events (e.g., pagehide, visibilitychange) upon the new page's commit. sullivan@chromium.org chrome-metrics-team@google.com Core metrics associated with web page loads. Amount of CPU wall time over the duration of a page load, while the page is in the foreground. Measured in milliseconds. The day of the week when the page load was initiated. The time is measured in the local time zone according to the system clock. Only collected for page loads that start while on foreground. The duration in milliseconds from the actual navigation start to the DOMContentLoadedEventFired event, excluding the time spent waiting for user interaction in beforeunload dialogs. This includes the time spent running beforeunload JavaScript handlers. Measures the time in milliseconds from navigation timing's navigation start to the time the DOMContentLoaded event is fired, for main frame documents. Measures the time in milliseconds from navigation timing's navigation start to the time the load event is fired, for main frame documents. The duration in milliseconds from the parse start to the DOMContentLoadedEventFired event. The duration since the last visit to this URL in seconds, if the user has visited the URL before. Recorded as -1 if the user has not visited the URL before, or if the History service is unavailable or slow to respond. Any duration that exceeds 30 days will be recorded as 30 days, so in practice, if this duration indicates 30 days, it can be anything from 30 to the maximum duration that local history is stored. Measures the time in milliseconds from when the OS level input event occurred which triggered a navigation, until the navigation was started. Only set if the code can trace back to the original input event. Some known cases where it is missing: crbug.com/889223 crbug.com/893162 crbug.com/897837 crbug.com/897840 Measures the cumulative layout shift (bit.ly/lsm-explainer) that has occurred on the page (including all subframes) until the first time the page moves from the foreground to the background. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. Stable since M98. Log of major changes: http://bit.ly/chrome-speed-metrics-changelog. This metric is trying to record the layout shift values which are not able to be recorded in LayoutInstability.CumulativeShiftScore for some reasons. Measures the maximum cumulative layout shift (bit.ly/cls-explainer) that has occurred in a session window until the first time the page moves from the foreground to the background. The gap between two consecutive shifts in a window is not bigger than 1000ms and the maximum window size is 5000ms. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. This metric is trying to record the layout shift values which are not able to be recorded in LayoutInstability.MaxCumulativeShiftScore.SessionWindow.Gap1000ms.Max5000ms for some reasons. True if this navigation was initiated by user input. For renderer navigations, only true if the code can trace back to the original input event. Some known cases where it is missing: crbug.com/889223 crbug.com/897837 crbug.com/897840 Re-adding experimental version of TotalForegroundDuration to allow for an analysis of the now obsolete field. This will help inform a discrepency found when switching to the PageTiming.TotalForegroundDuration field. The hour of the day when the page load was initiated. The time is measured in the local time zone according to the system clock. Only collected for page loads that start while on foreground. Measures First Input Delay, the duration between the hardware timestamp and the start of event processing on the main thread for the first meaningful input per navigation. See https://goo.gl/tr1oTZ for a detailed explanation. In ms. Log of major changes: http://bit.ly/chrome-speed-metrics-changelog The duration between navigation start and the hardware timestamp of the first meaningful input per navigation. See https://goo.gl/tr1oTZ for a detailed explanation. In ms. Measures the time, in milliseconds, from when a user starts scrolling until the result shows up on screen if the scroll was the first scroll since navigation. The duration between navigation start and the timestamp of the first scroll per navigation. In ms. This is rounded down to the nearest exponential bucket. The one-based offset of the interaction in the list of all interactions in this page load which was the INP (https://web.dev/inp). The time (in milliseconds from navigation start) of the interaction which was the INP (https://web.dev/inp). The number of distinct user interactions. See definition of user interaction: https://web.dev/better-responsiveness-metric/#group-events-into-interactions An approximation of a high percentile of user interaction latency of a page load, in ms. For this metric, we measure the latency of user interactions by Responsiveness.UserInteraction.MaxEventDuration. An approximation of a high percentile of user interaction latency of a page load, in ms. For this metric, we measure the latency of user interactions by Responsiveness.UserInteraction.MaxEventDuration. We record it when a page is moved from the foreground to the background for the first time. This metric is trying to record the INP values which are not able to be recorded in InteractiveTiming.UserInteractionLatency.HighPercentile2.MaxEventDuration for some reasons. The maximum value of user interaction latency of a page load, in ms. For this metric, we measure the latency of user interactions by Responsiveness.UserInteraction.MaxEventDuration. Each type of interaction has its own budget. The one-based offset of the interaction in the list of all interactions in this page load which was the INP before the first soft navigation (https://web.dev/inp). The time (in milliseconds from navigation starte) of the interaction which was the INP before the first soft navigation (https://web.dev/inp). The number of distinct user interactions before any soft navigation happens. It includes the user interaction that triggers the 1st soft navigtion. See definition of user interaction: https://web.dev/better-responsiveness-metric/#group-events-into-interactions An approximation of a high percentile of user interaction latency of a page load before any soft navigation happens, in ms. For this metric, we measure the latency of user interactions by Responsiveness.UserInteraction.MaxEventDuration. arthursonzogni@chromium.org Whether the navigation resulted in the main frame being hosted in a different renderer process than the current document or not. Marked true if the page was already bookmarked when the navigation committed. Marked true if the page was already part of a tab group when the navigation committed. Roughly: measures if the page was newly bookmarked. Specifically: this is marked true if the page was not bookmarked when the navigation committed, and is bookmarked at the end of the page lifetime. Note, this means this boolean is mutually exclusive with IsExistingBookmark. Also note, this means that a page that is bookmarked, and then sebsequently un-bookmarked is counted as false. Marked true if the page has been explicitly added (by the user) to the list of custom links displayed in the NTP. Links added to the NTP by History TopSites don't count for this. Always false on Android, because Android does not have NTP custom links. Roughly: measures if the page was added to a tab group. Specifically: this is marked true if the page was not part of a tab group when the navigation committed, and is part of a tab group at the end of the page lifetime. Note, this means this boolean is mutually exclusive with IsExistingPartOfTabGroup. Also note, this means that a page that is added, and then sebsequently removed from a tab group is counted as false. Marked true if the page was eligible for a scoped search when the navigation committed. Deprecated as of 04/23. It is still reported, but most users should use PageLoad.LayoutInstability.MaxCumulativeShiftScore.SessionWindow.Gap1000ms.Max5000ms. Measures the cumulative layout shift (bit.ly/lsm-explainer) that has occurred on the page (including all subframes) during the session. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. Stable since M79; previous versions are experimental and subject to fluctuation between releases. Log of major changes: http://bit.ly/chrome-speed-metrics-changelog Measures the cumulative layout shift (bit.ly/lsm-explainer) that has occurred on the page (including all subframes) during the session, before any user input or scroll (in any frame). This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. Measures the cumulative layout shift (bit.ly/lsm-explainer) that has occurred in the main frame during the session. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. Measures the cumulative layout shift (bit.ly/lsm-explainer) that has occurred in the main frame during the session, before any user input or scroll. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. Measures the maximum cumulative layout shift (bit.ly/cls-explainer) that has occurred in a session window. The gap between two consecutive shifts in a window is not bigger than 1000ms and the maximum window size is 5000ms. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. Measures the maximum cumulative layout shift (bit.ly/cls-explainer) that has occurred in a session window before any soft navigation happens. The gap between two consecutive shifts in a window is not bigger than 1000ms and the maximum window size is 5000ms. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. arthursonzogni@chromium.org An id for the main frame's document associated with the current navigation. It doesn't change for same-document navigations and when navigating back to the same document. Unique during the browsing session. The duration between the start of the connection establishment to the end of the connection establishment to the origin of the main frame request. In ms. The duration between the start of the DNS resolution to the end for the origin of the main frame request. In ms. HTTP Protocol scheme of the connection using which the main frame response was fetched. Recorded as HttpProtocolScheme enum in tools/metrics/histograms/enum.xml. nidhijaju@chromium.org chrome-loading@google.com The duration from the start of the navigation to the start of the connection establishment to the origin of the main frame request. In ms. The duration from the start of the navigation to the start of the response headers being received, in ms. The start of the response headers include early hints. This is TTFB in CruX. The duration from the start of the navigation to the start of the main frame request. This includes the time to process any redirects on the main frame request. In ms. The number of main frame redirects that occurred before commit. Unset if there were no redirects. A boolean indicating if any mainpage request in the redirect chain or committed request had cookies set on it. Only set when a query result was received from the cookie manager before event recording. A boolean indicating if any mainpage resource has a 'Cache-control: no-store' response header. Recorded only for main frame request. The duration from when the request was started to the time when its response headers were received. Recorded only for the main frame request. In ms. The duration from when the request was started to the time when its request headers were sent. Recorded only for the main frame request. In ms. The duration from when the request headers were sent to when its response headers were received. Recorded only for the main frame request. In ms. True if the socket was reused when establishing connection to the origin of the main frame request. page_load_metrics::NavigationHandleUserData::InitiatorLocation enum value. This field is used for identifying the types of prerender trigger. The |page_load_metrics::PageEndReason| for the main frame navigation of this page load. Replaced Navigation.PageEndReason2 on January 2021 to get rid of END_HIDDEN and use END_APP_BACKGROUND on cases where the metrics are flushed and the app enters the background. The |ui::PageTransition| for the main frame navigation of this page load. arthursonzogni@chromium.org Difference between the last committed navigation entry index before and after the commit of this navigation. The duration in milliseconds from the actual navigation start to when the navigation commit IPC is sent to the renderer, excluding the time spent waiting for user interaction in beforeunload dialogs. This includes the time spent running beforeunload JavaScript handlers. The duration in milliseconds from the actual navigation start to the navigation start (when beforeunload handlers have finished), excluding the time spent waiting for user interaction in beforeunload dialogs. This includes the time spent running beforeunload JavaScript handlers. The duration in milliseconds from the user interaction that started the navigation to the actual navigation start. The duration in milliseconds from the user interaction that started the navigation to the navigation start, excluding the time spent in beforeunload handlers. The duration in milliseconds from when the navigation commit IPC is sent to the renderer to when the parser starts. The duration in milliseconds from the navigation start to when the navigation commit IPC is sent to the renderer. The total number of body bytes loaded from cache for all resources on the page. This includes the http and memory cache. This is rounded down to the nearest exponential bucket (with a bucket spacing factor of 1.3). tbansal@chromium.org The downstream throughput (in kilobits per second) for this page load, at the time the navigation for this page load was initiated. The |SystemProfileProto::Network::EffectiveConnectionType| at the time the navigation for this page load was initiated. The |net::Error| code for this page load. Only logged for failed provisional navigations. The HTTP response code (e.g., 200, 404, etc.) associated with the main frame page load. The HTTP round trip time estimate for this page load, at the time the navigation for this page load was initiated. The total number of bytes loaded from the network for images on the page. This is rounded down to the nearest exponential bucket (with a bucket spacing factor of 1.3). On mobile this only records bytes up until the page/app is backgrounded. The total number of bytes loaded from the network for images on the subframes of the page. This is rounded down to the nearest exponential bucket (with a bucket spacing factor of 1.3). On mobile this only records bytes up until the page/app is backgrounded. The sum of decoded body lengths of all JavaScript resources that completed loading on the page. This is rounded down to the nearest exponential bucket (with a bucket spacing factor of 1.3). On mobile this only records bytes up until the page/app is backgrounded. The max of decoded body lengths of all JavaScript resources that completed loading on the page. This is rounded down to the nearest exponential bucket (with a bucket spacing factor of 1.3). On mobile this only records bytes up until the page/app is backgrounded. The total number of bytes loaded from the network for media (video, audio) on the page. This is rounded down to the nearest exponential bucket (with a bucket spacing factor of 1.3). On mobile this only records bytes up until the page/app is backgrounded. The total number of bytes loaded from the network for all resources on the page, including incomplete resources and headers. This is rounded down to the nearest exponential bucket (with a bucket spacing factor of 1.3). On mobile this only records bytes up until the page/app is backgrounded. The transport round trip time estimate for this page load, at the time the navigation for this page load was initiated. Marked true if the user copied the page's URL from the omnibox at any time for this page load. For page loads that start in the foreground, measures the duration of time in milliseconds until one of the following events occurs: the load of the main resource fails, the page load is stopped, the tab hosting the page is closed, the render process hosting the page goes away, a new navigation which later commits is initiated in the same tab, or the tab hosting the page is backgrounded. This metric has a very long tail, for pages that spend a long period of time in the foreground. As such, use of the mean or tail values are not recommended. Measures the time in milliseconds from navigation timing's navigation start to the time the provisional load failed. Only measures provisional loads that failed in the foreground. The total amount of time the page spent in the foreground, in milliseconds. Note that this metric adds all foreground durations occurring for the page. For example: if the page starts foregrounded during 1 second, then goes to the background for a while, then is foregrounded again for 2 seconds, and then is unloaded, this metric reports 3 seconds. To see metrics about consecutive foreground sessions, see the PageForegroundSession event instead. This metric is recorded at the end of each page load. As usual with PageLoad metrics, we consider app backgrounding on Android to be the end of page load as well as the app could be evicted by Android anytime after that happens. When the page UserTiming mark API is called with the standard 'mark_fully_loaded' name to mark that the page is fully loaded. If the mark is called more than once, the last is used. When the page UserTiming mark API is called with the standard 'mark_fully_visible' name to mark that the page is fully visible. If the mark is called more than once, the last is used. When the page UserTiming mark API is called with the standard 'mark_interactive' name to mark that the page is interactive. If the mark is called more than once, the last is used. Determines the final status for this page load. That is, records whether the page was never foregrounded, was foregrounded but did not reach FCP, or did reach FCP. The metric is recorded at the end of each page load. As usual with PageLoad metrics, we consider app backgrounding on Android to be the end of page load as well as the app could be evicted by Android anytime after that happens. The duration in milliseconds from the actual navigation start to the first contentful paint, excluding the time spent waiting for user interaction in beforeunload dialogs. This includes the time spent running beforeunload JavaScript handlers. The duration in milliseconds from the actual navigation start to the largest contentful paint, excluding the time spent waiting for user interaction in beforeunload dialogs. This includes the time spent running beforeunload JavaScript handlers. Measures the ratio of image file size (in bits) to displayed pixels for the LCP candidate. Bucketed into roughly exponential intervals to account for the wide range of encountered values (less than 1:100000 to greater than 10000:1). This metric is recorded whenever the LCP candidate is an image, and omitted when it is text. Measures the time in milliseconds from navigation timing's navigation start to the time when the image LCP element is requested. If the image LCP element is from a subframe, the difference in the navigation start of the main frame and that of the subframe is accounted for. This metric is not recorded when the LCP element is text or video. A bool that is set if the LCP element is an image that is cross-origin. This metric is not recorded when the LCP element is not an image, or the LCP element is not in the main frame. Measures the time in milliseconds from navigation timing's navigation start to the time when the LCP element(image) finishes loading. For animated images, this is when the image resource finishes loading completely, not the first frame time. See web.dev/optimize-lcp for more details. If the image LCP element is from a subframe, the difference in the navaigation start of the main frame and that of the subframe is accounted for. This metric is not recorded when the LCP element is text or video. Measures the time in milliseconds from navigation timing's navigation start to the time when the LCP element(image) starts loading. See web.dev/optimize-lcp for more details. If the image LCP element is from a subframe, the difference in the navaigation start of the main frame and that of the subframe is accounted for. This metric is not recorded when the LCP element is text or video. Resource loading priority used for the LCP candidate. This metric is recorded whenever the LCP candidate is an image, and omitted when it is text or video. Indicates that type of the LargestContentfulPaint candidate. This value is a packed set of flags, defined in third_party/blink/public/common/performance/largest_contentful_paint_type.h Measures the time in milliseconds from navigation timing's navigation start to the time when the page first paints content. A contentful paint includes a paint of text, image, SVG, or canvas. Log of major changes: http://bit.ly/chrome-speed-metrics-changelog Measures the time in milliseconds from navigation timing's navigation start to the time the first paint is performed. Measures the time in milliseconds from navigation timing's navigation start to the time when the page first paints the largest content (text or image) within viewport, across all frames. See http://bit.ly/fcp_plus_plus for more details. Log of major changes: http://bit.ly/chrome-speed-metrics-changelog. Version 2 differs from the first version in that it includes content that has been removed from the page. Similar to PaintTiming.NavigationToLargestContentfulPaint2, this measures the time in milliseconds from the navigation start of the top-level frame to the time when the page first paints the largest content (text or image) in the cross-site subframe which is the top candidate of largest contentful paint. This metric is computed separately from the page-level LCP, it's recorded even in page loads where the page-level LCP is not that cross-site LCP. Measures the time in milliseconds from navigation timing's navigation start to the time when the page first paints the largest content (text or image) within viewport, in the main frame. Includes content that has been removed from the page. See http://bit.ly/fcp_plus_plus for more details. Measures the time in milliseconds from navigation timing's navigation start to the time when the page first paints the largest content (text or image) within viewport, across all frames. Recorded at the point when a page which has been opened in the foreground is moved to the background for the first time. See http://bit.ly/fcp_plus_plus for more details. Log of major changes: http://bit.ly/chrome-speed-metrics-changelog. This metric is trying to record the LCP values which are not able to be recorded in PaintTiming.NavigationToLargestContentfulPaint2 for some reasons. The duration in milliseconds from the parse start to the first contentful paint. The duration in milliseconds from the parse start to the largest contentful paint. Similar to PaintTiming.NavigationToLargestContentfulPaint2, except that it is recorded before the first soft navigation. The duration in milliseconds from the actual navigation start to the parse start, excluding the time spent waiting for user interaction in beforeunload dialogs. This includes the time spent running beforeunload JavaScript handlers. Measures the time in milliseconds from navigation timing's navigation start to the time the parser started, for main frame documents. Records whether the refresh rate is throttled (capped to 30Hz) by Power Saver Mode for this Page Load, when the navigation starts. It is possible but very unlikely for this to change mid navigation, for instance due to a change by the user in settings or the battery being recharged above 20%. mcrouse@chromium.org tbansal@chromium.org Measures the Site Engagement Score for the committed URL rounded to the nearest 10. Records how the main frame's site instance was assigned to a renderer process. Number of soft navigations detected by heuristics. See https://github.com/WICG/soft-navigations Soft navigations are JS-driven same-document navigations that are using the history API or the new Navigation API, triggered by a user gesture and meaningfully modify the DOM, replacing the previous content with new one. dullweber@chromium.org Records whether third party cookie blocking was enabled for a site. Only recorded when third party cookies are bocked in general. Set to 1 if the main resource was served from cache. Records whether the Page Load is the result of the page having previously been discarded because of performance interventions or critical memory pressure. ryankalla@google.com potassium-anti-fraud-team@google.com Records when a frame from a known CAPTCHA provider is activated by the client. This could mean the user clicked or tapped on the frame. CAPTCHA frames are determined by matching the frame's src url against a list of known CAPTCHA frame url patterns. The list of known CAPTCHA url patterns is served via the component updater. This metric is split by whether an AI agent was active on the tab at the time of the frame activation. AI agent activity is determined by ActorKeyedService::IsActiveOnTab(), which is used by the Glic Agent. Records the CAPTCHA provider for the frame. ryankalla@google.com potassium-anti-fraud-team@google.com Records when a sub-frame is loaded from a known CAPTCHA provider. CAPTCHA frames are determined by matching the frame's src url against a list of known CAPTCHA frame url patterns. The list of known CAPTCHA url patterns is served via the component updater. This metric is split by whether an AI agent was active on the tab at the time of the frame load. AI agent activity is determined by ActorKeyedService::IsActiveOnTab(), which is used by the Glic Agent. Records the CAPTCHA provider for the frame. sullivan@chromium.org mushan@chromium.org Recorded for page loads that were navigated to via Google Search. Google Search mode used by the user. This corresponds to the tab (e.g. web result, image results, video results, etc.) the user is on. This information is extracted from the tbm query parameter on the Search URL. npm@chromium.org speed-metrics-dev@chromium.org Debugging metrics for the PageLoad event that do not require the same amount of sampling. Records whether the page's largest contentful paint, whose timestamp is measured by PageLoad.PaintTiming.NavigationToLargestContentfulPaint2, comes from text or image. This value is sampled and not on the PageLoad event, so it is not always recorded whenever PageLoad.PaintTiming.NavigationToLargestContentfulPaint2 is recorded. Records the state of the largest contentful paint algorithm at the time in which the browser attempts to record PageLoad.PaintTiming.NavigationToLargestContentfulPaint2. This state depends on whether there was a largest contentful paint candidate or not and the foreground state of the page. This value is sampled and not on the PageLoad event, so it is not always recorded whenever the browser attempts to record PageLoad.PaintTiming.NavigationToLargestContentfulPaint2. wanderview@chromium.org asamidoi@chromium.org Recorded for page loads controlled by a service worker. horo@chromium.org Recorded for page loads which main resource is an inner response of a signed exchange. Set to 1 if the main resource was served from Google's cache server. yaoxia@chromium.org jkarlin@chromium.org Metrics about the initiator context of a page load (e.g. whether certain condition was true and/or what triggered the page load). Recorded when a page commits. The metrics are specific to ad related conditions or purposes. Whether the page was opened due to a click on an ad. Whether the navigation occurred due to a user action. This includes omnibox navigations as well as renderer initiated navigations with user activation. battre@chromium.org Metrics about websites that contain password forms and the user's interactions with them. No events are created for pages that don't contain password forms. Records whether Chrome was sufficiently aware of forms during various stages of saving a credential on the site, or whether saving was aborted due to absence of the corresponding form manager. Values correspond to the enum PasswordManagerMetricsRecorder::FormManagerAvailable. Records if the user interacts with the password manager in a way that cannot (reliably) be attributed to a specific PasswordFormManager. Values correspond to the enum PasswordManagerMetricsRecorder::PageLevelUserAction. Records a provisional save failure in case the password manager cannot offer to save a credential. Recorded values correspond to the enum PasswordManagerMetricsRecorder::ProvisionalSaveFailure. Records a 1 for every page on which a user has modified a password text field - regardless of how many password fields a page contains or the user modifies. gklassen@chromium.org sullivan@chromium.org Records the paid content state of the page. This event is only recorded for pages that have paid content metadata. The paid content state of the page. ckitagawa@chromium.org fredmello@chromium.org mahmoudi@chromium.org vollick@chromium.org Metrics related to the performance of capturing Paint Previews. The amount of time spent painting in Blink to record a Paint Preview of the main frame. The compressed on-disk size in KB of the captured artifacts. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). selya@chromium.org dylancutler@google.com Recorded when a page has a partitioned cookie present and the cookie's name is not receive-cookie-deprecation. Records '1' if one or more partitioned cookies are present. This metric is only logged for sites that have partitioned cookies. battre@chromium.org chrome-password-manager-metrics-alerts@google.com Metrics about password forms on websites and the user's interactions with them. A separate event is generated for each password form discovered on a site. Records the correctness of a confirmation password field classification produced on the page load. Is recorded for successfully submitted password forms containing a confirmation password field. The classification is considered correct iff values in the new and in the confirmation password fields match. Recorded on the corresponding form manager destruction. Records the correctness of a current password field classification produced on the page load. Is recorded for successfully submitted password forms containing a current password field, for which the user had saved passwords. The classification is considered correct, if the field contained a previously saved password at the moment of submisison, and incorrect, if any other fields in the form contained a previosuly saved password. Recorded on the corresponding form manager destruction. Records the correctness of a new password field classification produced on the page load. Is recorded for successfully submitted password forms containing a new password field. The correctness is estimated based on previosuly saved credentials (if the field value was saved before, it's not a new password), and on whether the confirmation password field is present and validates form classification. Recorded on the corresponding form manager destruction. Records the correctness of a username field classification produced on the page load. Is recorded for successfully submitted password forms containing a username field, for which the user had saved usernames. The classification is considered correct, if the field contained a previously saved username at the moment of submisison, and incorrect, if any other fields in the form contained a previosuly saved username. Recorded on the corresponding form manager destruction. Records a low entropy hash of the form signature in order to be able to distinguish two forms on the same site. Records a bitmask of password form changes after the first time this form is seen by Password Manager. Recorded once per a form when the form is removed from DOM. Bit 0 - changes in number of fields, bit 1 - changes in fields renderer ids, bit 2 - changes in autocomplete attributes, bit 3 - changes in form control types, bit 4 - changes in field names. Records whether the PasswordAutofillAgent in the renderer manages to fill credentials as instructed by the browser or records a failure reason otherwise. Only the outcome of the first attempt to fill is recorded. This is replicated as PasswordManager.FirstFillingResultInRenderer in UMA. This metric records why the browser instructs the renderer not to fill the credentials on page load but to wait for the user to confirm the credential to be filled. Only the value of the first decision to fill is recorded. This is replicated as PasswordManager.FirstWaitForUsernameReason in UMA. Records '1' if the user has generated a password on this form. Records '1' if the user has modified a generated a password on this form. Records an entry if (and only if) a popup was shown. Records for each password form (and HTTP auth), whether the password manager has filled/provided credentials. Recorded values correspond to the enum ManagerAutofillEvent. Records for each successful submission of a password form, the degree to which the user has been assisted by the password manager. This is replicated as PasswordManager.FillingAssistance in UMA. Records for each successful submission of a single username form, the degree to which the user has been assisted by the password manager. This is replicated as PasswordManager.FillingAssistanceForSingleUsername in UMA. Records comparison of username and password fields picked during parsing when the form is filled and when the form is saved. Recorded only if the form is submitted successfully, on corresponding form manager destruction. Records the readonly status encoded with parsing success after parsing for filling. The |value| is constructed as follows: The least significant bit says whether parsing succeeded (1) or not (0). The rest, shifted by one bit to the right is the FormDataParser::ReadonlyPasswordFields representation of the readonly status. Records the readonly status encoded with parsing success after parsing for creating pending credentials. The |value| is constructed the same way as for ReadonlyWhenFilling. Records how the user interacted with a saving prompt. Records, for each password form seen by the password manager, whether the user was shown a prompt that asked for permission to save new credentials. In case a store() via the Credential Management API, a virtual password form is created, for which this metric is recorded. Records the trigger of each password save bubble that was shown to the user to ask for permission to save new credentials. Records if a manual fallback for saving option was shown to the user (on Android this is recorded, even though the UI does not exist at the moment). This is a bit mask. 1 means a prompt was shown, 2 means the prompt was due to a generated password, 4 means the prompt was due to an update. Records the event that led the password manager to believe the submission was successful. Metrics is recorded only when the password is saved. Records whether a submission of a password form has been observered. The values 0 and 1 correspond to false and true respectively. Records the outcome of a password form submission. Records the outcome of a password form submission for forms on which users generated passwords. Records the presumed type (e.g. signup, login, change password) of a submitted password form. This is replicated as PasswordManager.SubmittedFormType2 in UMA. Records how the user interacted with an updating prompt. Records, for each password form seen by the password manager, whether the user was shown a prompt that asked for permission to update existing credentials. Records the trigger of each password update bubble that was shown to the user to ask for permission to update existing credentials. Counts how many times the user corrected a username in a password form. This is only populated if events happened. Counts how many times the user edited a username in a password save/update bubble. This is only populated if events happened. Counts how many times the user selected a different password in a save/update bubble. This is only populated if events happened. Counts how many times the user triggered the manual fallback for password saving. This is only populated if events happened. otara@google.com vsemeniuk@google.com Metrics related to the outcome of the password change flow. Recorded when the flow is stopped. Records the state at which the password change flow was stopped, either successful, failed or interrupted. This mirrors CoarseFinalPasswordChangeStatus in UMA. otara@google.com vsemeniuk@google.com Metrics related to the recovery flow of the password change feature. Records users enter and exit the recovery flow. Records the start and the end of the password recovery flow. We will consider the start to be an interaction with a "Trouble signing in" button in the password autofill dropdown or a proactive recovery popup that we show if we detect a failed log-in attempt. The end of the flow is considered to be an updated primary password (the backup password becomes the primary password). This happens after the user accepts the "update password" bubble This mirrors PasswordChangeRecoveryFlow in UMA. fiorellab@google.com vsemeniuk@google.com Metrics related to password change submission flow. Recorded when password are saved by password change submission trajectory. Records whether or not a login is successful using either the primary or the backup password previously saved through password change submission flow. It is recorded at the moment of detecting a succesful or failed log in. This mirrors LogInWithPasswordChangeSubmission in UMA. kazinova@google.com chrome-password-manager-metrics-alerts@google.com Metrics for clientside password form classifications. Recorded when the field types are received by Password Manager after the model finishes execution, iff there is data available for the below metrics. Records if the predictions made by Password manager clientside classification model are empty (= contain only NO_SERVER_DATA values). Records whether a form is an OTP form, a normal multi-use password form, or a combination of both. ioanap@chromium.org kazinova@google.com Logged on Android for the first page that a Chrome Custom Tab loads. Recorded when the primary page changes away from the first loaded one or when the WebContents is destroyed. True if there was a password form detected on the page. jkeitel@google.com chrome-duplex@google.com Metrics related to leak warnings shown on login and password reset. Leak warning dialogs in Chrome aim to inform users about leaked credentials and advise to change them. We collect type of the dialog that is shown and the response of the user to it. Records the response to the password leak dialog. Records the type of the password leak dialog. kazinova@google.com chrome-password-manager-metrics-alerts@google.com Metrics for newly saved credentials, recorded at the moment of saving. Whether the password has an empty username. Whether the password was generated by Chrome or typed manually by the user. Obsolete: The metric is not recorded in M131+. However this entry is kept to gather UKM data from pre-M131 clients. kazinova@google.com chrome-password-manager-metrics-alerts@google.com Metrics related to password form parsing. Recorded when the form is parsed for filling (so parsing for saving on every keystroke does not lead to metric recording) by Password Manager (so parsing by other interested components does not lead to metric recording). Recorded iff there is data for any of the metrics below (e.g. when the parsed form contains either current password fields, or fields irrelevant to password flow). Reports whether the field predicted as password field in a parsed form is masked (has HTML type "password") or not. Recorded iff form contains a field with PASSWORD type prediction. Reports whether any of the fields predicted to be unrelated to password flows is masked (has HTML type "password") or not. Recorded iff the form contains fields with UNKNOWN_TYPE prediction. fiorellab@google.com vsemeniuk@google.com Metrics reported after the change password form submission in the password change flow. Records the change password form submission result (success or the error type). This is recorded when receiving a response from the optimization guide service to a request made with the representation of the web page where the submission happened. This mirrors PasswordManager.PasswordChangeSubmissionOutcome in UMA. brunobraga@google.com chrome-duplex@google.com Metrics related to password change. The Password Checkup in Chrome Settings aims to help users find their leaked passwords. After getting list of compromised passwords, user can trigger password change. Records what type of password change the user triggers. kazinova@google.com vsemeniuk@google.com Records password filling for IOS. The metric records whether the password filling is successful upon suggestion click. A boolean indicating success or failure. kazinova@google.com chrome-password-manager-metrics-alerts@google.com Metrics for saved passwords, recorded at the moment of saving. Whether the password was generated by Chrome or typed manually by the user. kazinova@google.com chrome-password-manager-metrics-alerts@google.com Metrics related to the usage and support of .well-known/change-password. The .well-known/change-password standard is defined to redirect to the sites change password form. We collect whether a site supports the standard or not. https://wicg.github.io/change-password-url/ Records if a site supports .well-known/change-password and if a url override was used or not. smcgruer@chromium.org chrome-payments-eng@google.com Records checkout events as well as completion status for the invoked payment app when its origin is shown inside the payment handler modal window. This metric uses the specific PAYMENT_APP_ID ukm_source_id which is generated from the app's origin. How the Payment Request ended. This metric will only get recorded if a non-autofill payment app gets invoked to handle the request. The 2.0 version of Events. Bitfield representing the events that occurred during the Payment Request. This metric will only get recorded if a non-autofill payment app gets invoked to handle the request. Values defined in the Event2 enum of components/payments/core/journey_logger.h. smcgruer@chromium.org chrome-payments-eng@google.com How the Payment Request ended. Values defined in the CompletionStatus enum of components/payments/core/journey_logger.h. The 2.0 version of Events. Bitfield representing the events that occurred during the Payment Request. Values defined in the Event2 enum of components/payments/core/journey_logger.h. The reason why the browser window size check failed and the Payment Request or Payment Handler dialog was rejected. raymes@chromium.org Event recorded when a flash instance connects to the broker, resulting in a channel being opened to the broker process. nrosenthal@chromium.org zuoaoyuan@chromium.org This event is associated with the LongAnimationFrame Web Platform API. The metrics are recorded when a long animation frame is detected and reported to the page. For more details see https://crbug.com/1392685. A bitfield representation of third party technologies who have contributed at least one event/callback script that blocked the long animation frame for more than 5ms. The bitwise fields are defined by ThirdPartyTechnology enum in third_party/blink_renderer/core/timing/third_party_script_info.h. A bitfield representation of third party technologies who have contributed at least one script execution block that blocked the long animation frame for more than 5ms. The bitwise fields are defined by ThirdPartyTechnology enum in third_party/blink_renderer/core/timing/third_party_script_info.h. True if there was a pause (sync XHR, dialog) during the long frame. A positive or negative duration in milliseconds, representing the time between the desired time for rendering the frame and rendering it in practice. Positive means delayed, negative means deferred. The duration in milliseconds, representing the time the LoAF was blocking. The total duration of long-script JS compilation, in milliseconds. The total duration of long-script JS execution, not including compilation, in milliseconds. The total duration spent in event listeners (e.g. onclick). The total duration spent in promise resolve/reject tasks. The total duration spent in script blocks (script element, import) The total duration spent in user callbacks (e.g. requestAnimationFrame). The total duration of forced style and layout in long scripts. The total duration of the style and layout lifecycle phase, not including forced style and layout. The duration of the long animation frame in milliseconds. Mirrors the value reported in PerformanceEntry.duration. hpayer@chromium.org mlippautz@chromium.org ulan@chromium.org This event is associated with the LongTask Web Platform API. The metrics are recorded when a long task is detected and reported to the page. For more details see the design doc linked in https://crbug.com/1173527. The duration of the long task in microseconds. Mirrors the value reported in PerformanceEntry.duration. The total time in microseconds that is V8 spent in main-thread V8, including JS-time, but without time spent in callbacks to blink. The total time in microseconds that V8 spent collecting garbage within the long task. The total time in microseconds that V8 spent in the atomic pause of the full garbage collector within the long task. The total time in microseconds that V8 spent in the incremental pause of the full garbage collector within the long task. The total time in microseconds that V8 spent in the young garbage collector within the long task. The duration in millseconds between the document navigation start and the long task start. Mirrors the value reported in PerformanceEntry.startTime. hpayer@chromium.org mlippautz@chromium.org ulan@chromium.org This event is associated with the performance.measureUserAgentSpecificMemory API. The metrics are recorded when the memory measurement is reported to the page. All values represent memory consumption in KB. The total DOM memory measured in KB. The total JavaScript memory measured in KB. The total JavaScript memory in the DedicatedWorker scope measured in KB. The total shared memory measured in KB. hpayer@chromium.org mlippautz@chromium.org ulan@chromium.org This event is associated with the performance.memory API. The metrics are recorded when the memory measurement is reported to the page. All values represent memory consumption in KB. The JavaScript memory (usedJSHeapSize) measured in KB. fdoray@chromium.org chrome-catan@google.com Properties of a tab involved in determining its freezing eligibility. Recorded for each tab every minute (subsampled). True if the page was audible over the last minute True if the page was mirrored over the last minute True if the page captured audio, video, a window or the display over the last minute True if the the page was connected to a device (USB, Bluetooth, Hid or Serial) over the last minute Highest CPU usage for a group of same-origin frames on the page over any past 1 minute period during which no opt-out reason was applicable, in percentage of a core (100 = 100% of a core used for 1 minute). Exponential bucketing is applied. Highest CPU usage for a group of same-origin frames on the page over the last 1 minute period, in percentage of a core (100 = 100% of a core used for 1 minute). Exponential bucketing is applied. True if the page held an IndexedDB lock blocking another page over the last minute True if the page held a Web Lock over the last minute True if the page was in the loading state over the last minute True if the page had the notification permission over the last minute True if the page was opted-out via origin trial over the last minute True if the page was recently audible over the last minute True if the page was recently visible over the last minute True if the page was visible over the last minute True if the page used WebRTC over the last minute agale@chromium.org anthonyvd@chromium.org joenotcharles@google.com Metrics detailing the estimated resources used by pages. The resources of a renderer process are allocated to all pages hosted in that process using the algorithm in performance_manager::ProcessMetricsDecorator. An event is collected for each open page every 2 minutes, which is the same frequency as the PerformanceMonitor.AverageCPU10.* histograms. Whether the page was in the foreground, background, or mixed during the last 2 minutes. The algorithm used to estimate the resource usage of the page. If this is missing, the Legacy algorithm was used. The estimated private footprint of the page, in kilobytes. The estimated CPU usage of the page over the last 2 minutes, with 10,000 as 100% usage of 1 full core. (The same scale as the PerformanceMonitor.AverageCPU10.* histograms.) The estimated resident set size of the page, in kilobytes. The sum of the estimated CPU usage of all pages over the last 2 minutes, with 10,000 as 100% usage of 1 full core. (The same scale as the PerformanceMonitor.AverageCPU10.* histograms.) RecentCPUUsage can be compared to this to find how the page compares to other open tabs. nator@chromium.org rayankans@chromium.org peter@chromium.org A website can register to periodically update itself in the background using a service worker. Updates can be done in response to the periodicsync event. Logged when a periodicsync event completes and returns a status. We collect information about retries of this event. Integer value of the maximum number of allowed retries. Integer value of the number of retries / sync events dispatched. The enum value of the completion status, defined in blink::ServiceWorkerStatusCode. nator@chromium.org rayankans@chromium.org peter@chromium.org A website can register to periodically update itself in the background using a service worker. Updates can be done in response to the periodicsync event after a periodic Background Sync registration has been created. Logged when a periodic Background Sync registration is created. We collect information about this registration request. Boolean for whether a registration with this tag has already been created. Integer value of the minimum interval, in milliseconds, of periodicsync events requested by the origin for this registration. This number is exponentially bucketed and uses the UKM GetExponentialBucketMin method with a value of 2.0 for spacing. permissions-dev@chromium.org Metrics for a user's permission actions, logged when a permission prompt is resolved or when a permission is revoked. Grouped (microphone+camera) permission prompts are reported separately. An enum of type PermissionAction. One of GRANTED, DENIED, DISMISSED or IGNORED. An enum of type GeolocationPromptType (see https://source.chromium.org/search?q=class:%5CbGeolocationPromptType%5Cb). This is only emitted for geolocation permission prompts if approximate geolocation is enabled. In that case, it encodes the variant of the geolocation prompt type and it is either 0 (kApproximateOrPrecise), 1 (kApproximateOnly) or 2 (kUpgradeToPrecise). An enum of type PermissionRequestGestureType, whether the prompt was initiated by a user gesture. An enum of type GeolocationAccuracy (see https://source.chromium.org/search?q=class:%5CbGeolocationAccuracy%5Cb). This is only emitted for geolocation permission prompts with precise/approximate options (corresponding to ContentSettingsType::GEOLOCATION_WITH_OPTIONS). In that case, it is either 0 (kPrecise) or 1 (kApproximate) and represents the preselected accuracy option in the prompt. An enum of type ElementAnchoredBubbleVariant. The screen that prompts the user for the OS screen, if applicable. An enum of type ElementAnchoredBubbleVariant. The screen that directs the user to the OS Settings, if applicable. Version of the AI model used for the permission relevance prediction. An enum of type PermissionAutoRevocationHistory, representing whether the permission has been automatically revoked in the past at least once. An enum of type PermissionRequestRelevance describing the permission request relevance returned by a genAI and/or TFLite on-device model. The type of permission, encoded through the mapping defined by `kHistogramValue` in https://source.chromium.org/chromium/chromium/src/+/main:components/content_settings/core/browser/content_settings_uma_util.cc;l=22;bpv=1;bpt=1. Beware that this is not the same as the ContentSettingsType enum (casted to int). An enum of type ContentSettingsType. An enum of type PermissionUiSelector::PredictionGrantLikelihood describing the likelihood returned by the Web Permission Predictions Service, if the service was successfully queried. A boolean that records whether the response by the permissions prediction service was held back. The number of dismissed prompts for the given (origin, permission) pair, capped at 10 dismissals. The number of ignored prompts for the given (origin, permission) pair, capped at 10 ignores. An enum of type PermissionPromptDisposition. The UI used to display a permission prompt, if any. An enum of type PermissionPromptDispositionReason. The reason why the particular prompt disposition was used, if any. An enum of type UkmPermissionPromptOptions (see https://source.chromium.org/search?q=class:%5CbUkmPermissionPromptOptions%5Cb). This is only emitted if the permission prompt supports additional prompt options, which currently is the case only for geolocation permission prompts when the ApproximateGeolocationPermission is enabled. In that case, this is either 1 (APPROXIMATE_LOCATION) or 2 (PRECISE_LOCATION) and specifies whether the prompt selection was "approximate" or "precise" at the time when the action was taken. A bitmask consisting of flags defined in enum type AdaptiveTriggers. A bit is set if the corresponding adaptive triggering condition has been satisfied at least once in the past for the user. This metric is not recorded for users who don't have the activation dry run enabled. An enum of type ElementAnchoredBubbleVariant. The site level screen shown to the user, if any. An enum of type PermissionSourceUI. The UI surface for this action. Total number of prompts shown to the user in the last 28 days. The count is bucketed. The bucket minimums being 20, 15, 12, 10, 9, 8, 7, 6, 5, 4 and 0. The correct bucket is the first one that is smaller or equal to the prompt count. What percentage of prompts from the last 28 days did the user deny. Aggregated across all origins, all PermissionTypes and all PromptDisposition. Rounded to the nearest 10%. What percentage of prompts from the last 28 days did the user dismiss. Aggregated across all origins, all PermissionTypes and all PromptDisposition. Rounded to the nearest 10%. What percentage of prompts from the last 28 days did the user grant. Aggregated across all origins, all PermissionTypes and all PromptDisposition. Rounded to the nearest 10%. What percentage of prompts from the last 28 days did the user ignore. Aggregated across all origins, all PermissionTypes and all PromptDisposition. Rounded to the nearest 10%. Total number of prompts shown to the user in the last 28 days for the currently requested permission type. The count is bucketed. The bucket minimums being 20, 15, 12, 10, 9, 8, 7, 6, 5, 4 and 0. The correct bucket is the first one that is smaller or equal to the prompt count. What percentage of prompts from the last 28 days did the user deny. Aggregated across all origins, the current requested PermissionType and all PromptDisposition. Rounded to the nearest 10%. What percentage of prompts from the last 28 days did the user dismiss. Aggregated across all origins, the current requested PermissionType and all PromptDisposition. Rounded to the nearest 10%. What percentage of prompts from the last 28 days did the user grant. Aggregated across all origins, the current requested PermissionType and all PromptDisposition. Rounded to the nearest 10%. What percentage of prompts from the last 28 days did the user ignore. Aggregated across all origins, the current requested PermissionType and all PromptDisposition. Rounded to the nearest 10%. Total number of loud UI prompts shown to the user in the last 28 days. The count is bucketed. The bucket minimums being 20, 15, 12, 10, 9, 8, 7, 6, 5 ,4 and 0. The correct bucket is the first one that is smaller or equal to the prompt count. What percentage of prompts from the last 28 days did the user deny. Aggregated across all origins, all PermissionTypes and loud UI PromptDispositions only. Rounded to the nearest 10%. What percentage of prompts from the last 28 days did the user dismiss. Aggregated across all origins, all PermissionTypes and loud UI PromptDispositions only. Rounded to the nearest 10%. What percentage of prompts from the last 28 days did the user grant. Aggregated across all origins, all PermissionTypes and loud UI PromptDispositions only. Rounded to the nearest 10%. What percentage of prompts from the last 28 days did the user ignore. Aggregated across all origins, all PermissionTypes and loud UI PromptDispositions only. Rounded to the nearest 10%. Total number of loud UI prompts shown to the user in the last 28 days for the currently requested permission type. The count is bucketed. The bucket minimums being 20, 15, 12, 10, 9, 8, 7, 6, 5, 4 and 0. The correct bucket is the first one that is smaller or equal to the prompt count. What percentage of prompts from the last 28 days did the user deny. Aggregated across all origins, currently requested PermissionType and loud UI PromptDispositions only. Rounded to the nearest 10%. What percentage of prompts from the last 28 days did the user dismiss. Aggregated across all origins, currently requested PermissionType and loud UI PromptDispositions only. Rounded to the nearest 10%. What percentage of prompts from the last 28 days did the user grant. Aggregated across all origins, currently requested PermissionType and loud UI PromptDispositions only. Rounded to the nearest 10%. What percentage of prompts from the last 28 days did the user ignore. Aggregated across all origins, currently requested PermissionType and loud UI PromptDispositions only. Rounded to the nearest 10%. Time (in milliseconds) taken by the user to make a decision on a permission prompt after it was displayed, if at all. Ignoring a prompt doesn't count as a decision. hkamila@chromium.org andypaicu@chromium.org permissions-dev@chromium.org Metrics for a user's permission actions for permission prompts triggered by the user clicking on the Embedded Permission Element, logged when a permission prompt is resolved. An enum of type ElementAnchoredBubbleAction. One of GRANTED, GRANTED_ONCE, OK, DENIED, DISMISSED, DISMISSED_SCRIM, SYSTEM_SETTINGS or IGNORED. Represents the permission type that the prompt is for. This can sometimes differ from PermissionType when the PermissionType is Multiple, as some screens are designed to only be shown for one permission at a time. This is encoded via RequestTypeForUma in https://source.chromium.org/chromium/chromium/src/+/main:components/permissions/permission_uma_util.h;bpv=1;bpt=1;l=73. Beware that this is not the same as the ContentSettingsType enum (casted to int). Represents how many screens have already been shown to the user prior to the current one as part of the permission prompt. Also serves as an order number. Should never be bigger than the number of screen variants. It is recorded as int64_t (upper bound 10). Represents the permission type that current screen is for. This is encoded via RequestTypeForUma in https://source.chromium.org/chromium/chromium/src/+/main:components/permissions/permission_uma_util.h;bpv=1;bpt=1;l=73. Beware that this is not the same as the ContentSettingsType enum (casted to int). An enum of type ElementAnchoredBubbleVariant. The screen that the user has currently taken action on. jkarlin@chromium.org chrome-ads-team@google.com Collects once-per-page statistics on usage of privacy sensitive permission policy APIs to better understand if restricting access to these APIs in some situations would result in an acceptable breaking change or not. If a privacy-sensitive labeled feature backed by a permission policy was requested and granted, whether ad-script was in the stack at the time of grant. This is only recorded once per page load. engedy@chromium.org ravjit@chromium.org permissions-dev@chromium.org Metrics for usage of capabilities gated behind permission prompts. One event is recorded per permission type for each page load that uses that permission type. The type of permission used, encoded through the mapping defined by `kHistogramValue` in https://source.chromium.org/chromium/chromium/src/+/main:components/content_settings/core/browser/content_settings_uma_util.cc;l=22;bpv=1;bpt=1. Beware that this is not the same as the ContentSettingsType enum (casted to int). skrakowi@chromium.org chrome-counter-abuse-alerts@google.com Recorded when the notification content detection model is ran on notification contents, which happens on a request to display a persistent notification. A boolean signaling whether the user had previously decided to always allow suspicious notifications from the URL. A boolean signaling whether the URL is on the high confidence allowlist or not. The site engagement level of the site when the user receives a notification from them. Set to 0 even if the origin is not tracked by the Site Engagement Service. The 'suspicious' score given to notification contents that came from the URL. Ranges from 0 to 100. tommycli@chromium.org Recorded when a Flash object instance is created. jkeitel@google.com vizcay@google.com An email was submitted in a form for which the user had previously seen a plus address suggestions. More precisely, this event is recorded whenever a user submits a form that contains a form field for which a plus address suggestion was shown and the form field contains either a plus address or the current profile's GAIA email. Boolean value representing whether commerce heuristics suggest that the main frame URL belongs to a checkout or a cart page. The number of fields in the browser form. This uses exponential bucketing with a spacing of 1.15. The number of fields in the renderer form. This uses exponential bucketing with a spacing of 1.15. Boolean value representing whether the active profile is managed. Boolean value representing whether the submitted email was newly created for this submission. The password manager classification of the form focused by the user when showing the plus address suggestion. The (bucketed) number of plus addresses the user had when shown the plus address suggestion. Boolean value representing whether the submitted email was a plus address. The context in which the plus address suggestion (filling or creation) was shown in the Autofill UI. Boolean value representing whether the user was offered a suggestion to create a new plus address. csharrison@chromium.org jkarlin@chromium.org A popup window was closed. The time (in ms, rounded down to the nearest power of 2) a popup WebContents is visible / foregrounded, until it is closed. Keyed by the popup opener's URL (not the URL of the popup). Integer value representing how many mouse down, touch start user, or raw key down interactions the popup had. This is capped at 50. Integer value representing how many gesture scroll begin user interactions the popup had. This is capped at 50. Integer value representing how many user interactions the popup had. This is capped at 100. See IsUserInteractionInputType in web_contents_impl for what an interaction is. Currently it is a mouse, scroll begin, touch start, or raw key down event. The number of redirects taken by the pop-up's initial navigation. This does not include javascript navigations initiated by the opening document for the pop-up. The site's safe browsing status if the pop-up's navigation commits. Boolean value representing whether the popup was opened via some trusted event (e.g. context menu, ctrl-click, etc). Boolean value to represent whether the popup was closed by user gesture. The UI window open disposition used when creating the web contents of the pop-up. Set by the browser process when the pop up tracker is created. justinmiron@google.com Page level pop-up statistics. Whether a user has, explicitly, allowed all popups on the last committed url of a page. The user's site level popup content setting is queried when a page that opened a popup is destroyed. Only emitted when at least one popup has been opened and the last committed page url is valid. lfg@chromium.org A portal was activated. A portal activation enables a seamless transition between different pages. See the portals explainer for more details: https://github.com/WICG/portals/blob/master/README.md. When this page load happened inside a Portal, this metrics records the time in milliseconds between the portal activation and the first paint after activation. Measures the time between LCP and the portal activation. Uses exponential bucketing with a spacing of 2, with the units in milliseconds. Indicates whether the portal activation happened before or after the largest contentful paint. fdoray@chromium.org chrome-catan@google.com Metrics to measure power consumption in key usage scenarios, as well as the time spent in each scenario. These metrics are reported for a fixed interval of time, the duration of this interval is recorded in the IntervalDurationSeconds metric. The events aren't tied to a specific URL, a timer independent of any load event runs in Chrome and report these events at regular intervals. All the metrics associated to this event are for the last interval. The URL attached to the event is the one that has been visible for the longest time during the interval, the exact visibility duration for this URL during the interval is reported in the URLVisibilityTimeSeconds metric. e.g., at T-0 an interval ends and this event is emitted, a new interval then starts. During the interval the users spends 75% of his time watching a fullscreen video on https://foo.com with 10 tabs opened in background. Just before the end of the interval the user closes all the tabs but one. An event will be emitted for https://foo.com at T+X with the following metrics: IntervalDurationSeconds: X in seconds; URLVisibilityTimeSeconds: 0.75 * X in seconds; FullscreenVideoSingleMonitorSeconds: 0.75 * X in seconds; MaxTabCount: 10; TabClosed: 9; ... (not all metrics are listed here because it'd be too long, but in practice all the metrics are reported for each interval). This is reported on Mac and Windows. See go/chrome_power_use_per_scenario for more details. Battery discharge mode describing whether BatteryDischargeRate could be reported or not, and why. Recorded as enum BatteryDischargeMode in //tools/metrics/histograms/enums.xml. Battery discharge rate per minute, with 1/10000 of full charge resolution, example: - Battery capacity = 4000 mAh; - Battery charge at the beginning of the interval: 3900 mAh; - Battery charge at the end of the interval: 3700 mAh; - Discharge proportion: (3900-3700) / 4000 = 0.05 - Reported value: 500. Indicates if the browser is shutting down when this event is reported. The CPU time spent executing code, across all Chrome processes in the interval. This can exceed 100% in multi-thread processes running on multi-core systems. Whether or not the device has been sleeping at any point during the interval. Time spent playing video in fullscreen, single-monitor. In seconds. This includes video for any tabs during the interval. Sum of idle-wakeups for all Chrome processes (Mac-only). The interval duration, in seconds. The maximum number of tabs that existed at the same time during the interval. The maximum number of windows that have been visible at the same time. The visibility time for the origin of the URL associated with this event, in seconds. This can differ from URLVisibilityTimeSeconds if there's multiple tabs loaded with the same URL that are visible during the interval. E.g. if there's one tab loaded with foo.com/a visible for 1 second during the interval and one loaded with foo.com/b visible for 2 seconds the report will have URLVisibilityTimeSeconds = 2 seconds and OriginVisibilityTimeSeconds = 3 seconds. This can also exceed the interval duration (e.g. 2 visibles tab with loaded the same origin during the entire interval). Sum of package exits from all Chrome processes (Mac-only). The time spent playing audio in at least one tab, in seconds. The number of tabs that have been closed during the interval. The time spent playing video in at least one visible tab, in seconds. The time since the last interaction with the browser when this event gets recorded, in seconds. The time since the last interaction with the system when this event gets recorded, in seconds. The time spent with at least one opened WebRTC connection, in seconds. Number of main frame different-document navigations. Elapsed time since Chrome has started, in seconds. The visibility time for the URL associated with this event, in seconds. Number of user interaction (scroll, click or typing) during the interval. The time spent capturing video in at least one tab, in seconds. curranmax@chromium.org ryansturm@chromium.org spelchat@chromium.org Each event corresponds to a single resource, either mainframe or subresource, that was prefetched during an Isolated Prerender. Logged one or many times on a Google Search Result Page for eligible users. The bucketed number of bytes that transited the network to prefetch this resource. Only set when the resource was prefetched. The number of milliseconds that the prefetch took from start to completion. Only set when the resource was prefetched. How ISP filtering interacted with the mainframe resource if it was navigated to. Only set when LinkClicked is true. Whether the mainframe was navigated to. Always set when the event is recorded, even on subresources which take the value from their corresponding mainframe. The location in the original navigation prediction of where the mainframe url was. Always set when the event is recorded, even on subresources which take the value from their corresponding mainframe. The number of milliseconds between the start of the navigation and the start of the prefetch request. Only set when the resource was attempted to be prefetched. The type of resource that was prefetched, either mainframe or subresource. Always set when the event is recorded. The outcome of the prefetch eligibility check and possible attempt to prefetch the resource and reuse it. See |PrefetchProxyPrefetchStatus| for a mapping of enum values and their meanings. Always set when the event is recorded. domenic@chromium.org chrome-instant-ecosystem@google.com This event is logged for each preloading attempt i.e., for preloading features like preresolve, preconnect, prerender, prefetch, and others, as long as the user navigated to another page after the preloading attempt occurred. It is associated with the Source ID of the page user navigates to, i.e., the page the user visits *after* a browser feature or the previous page attempted to preload. The page the user navigates to may or may not be the same as the URL that was attempted to be preloaded. If they are the same, AccurateTriggering will be true; otherwise, it will be false. As such, there can be many Preloading.Attempts for a single UKM source, of which only one will have AccurateTriggering set to true. This event is logged when the user navigates onward from that successor page, or that sucessor page's WebContents is being destroyed. This means that if a UKM source that attempts a preload is not followed by another navigation to a successor page, e.g. because the user closed their browser, then the Preloading.Attempt event is not recorded: only the Preloading.Attempt.PreviousPrimaryPage is recorded. On the other hand, in cases where there is no initiating page, e.g. because the preload is initiated via Chrome UI, then only Preloading.Attempt is recorded, and not Preloading.Attempt.PreviousPrimaryPage. Note that this event is sampled based on the PreloadingConfig feature configuration (in addition to the normal UKM sampling mechanism). The sampling rate for each event (which may vary for different preloading features) is logged in the SamplingLikelihood field. Whether the preloading operation was triggered accurately for the predicted URL (regardless of the eligibility, failures, etc.). Set to true if the user navigated to the predicted URL. Eligibility status of preloading operation for the prediction. Values are from the Chrome enum content::PreloadingEligibility. Reason why the preloading operation failed. Values are from the enum specific to the preloading type. The outcome of the holdback check. This is not part of eligibility status to make clarify that this check needs to happen after we are done verifying the eligibility of a preloading attempt. In general, eligibility checks can be reordered, but the holdback check always needs to come after verifying that the preloading attempt was eligible. Records how many times are we checking if a Service Worker is registered for an origin (fast check) or for a certain path of the origin (slow check) to decide if we are proceeding with a prefetch or not. Enum values would be: OriginOnly and Path. OriginOnly value describes the case when there are no Service Workers registered for the origin and we only need to run the fast check. Path value describes the case when there are Service Workers registered for the origin and we need to do a more expensive path check to find out if the prefetch URL falls under a registered Service Worker. Records how long (in microseconds) does the slow check for a Service Worker being registered for a certain path take when deciding if we are proceeding with a prefetch or not. We expect this check to take up to 2 ms. Exponentially bucketed using GetExponentialBucketMin with a value of 1.15. The type of preloading predictor which made the prediction. Values are from the Chrome enum content::PreloadingPredictor or ChromePreloadingPredictor. The type of preloading operation used for the preloading attempt. Values are from the Chrome enum content::PreloadingType. The amount of time the attempt took to reach a state in which it is ready to be used for the next navigation, in milliseconds. Note: currently only Prefetch, Prerender and NoStatePrefetch have a ready state. This field is only meaningful to these three types of preloading. The sampling rate of this particular event multiplied by 1,000,000 (because we cannot log floating point values in UKM). A value of 1,000,000 means that the event had a 100% chance of being sampled, whereas a value of one means that the event had a one-in-a-millon chance of being sampled. The sampling rate for preloading attempts is set via field trial through the PreloadingConfig feature. The logged value just represents the configuration used for that preloading attempt. The eagerness of the speculation rule that triggered this preload. This is only set if the Preloading Attempt is triggered by a speculation rule. Values are from the Chrome enum blink::mojom::SpeculationEagerness. The amount of time the attempt took to reach the next navigation, in milliseconds. This can happen multiple times, but only the first is recorded. Specifies whether the preloading attempt triggered and was used for the next navigation. Values are from the Chrome enum content::PreloadingTriggeringOutcome. domenic@chromium.org chrome-instant-ecosystem@google.com This event is logged for each preloading attempt i.e., for preloading features like preresolve, preconnect, prerender, prefetch, and others, as long as there is a triggering page. The difference from Preloading.Attempt is that it is associated with the Source ID of the primary triggering page instead of the navigated-to page. This allows you to find all preloading attempts initiated by a given UKM source, in order to, e.g., filter by URL. This event is logged when the user navigates onward from that initiating page, or that initiating page's WebContents is being destroyed. This means that in cases where there is no initiating page, e.g. because the preload is initiated via Chrome UI, then the Preloading.Attempt.PreviousPrimaryPage UKM is not recorded; only the Preloading.Attempt UKM is recorded. On the other hand, if a preload attempt is not followed by a navigation, e.g. because the user closed their browser, then only Preloading.Attempt.PreviousPrimaryPage is recorded, and not Preloading.Attempt. Note that some UKM sources that record Preloading.Attempt.PreviousPrimaryPage may not appear as normal page loads. For example, the New Tab Page. Note that this event is sampled based on the PreloadingConfig feature configuration (in addition to the normal UKM sampling mechanism). The sampling rate for each event (which may vary for different preloading features) is logged in the SamplingLikelihood field. Whether the preloading operation was triggered accurately (regardless of the eligibility, holdback, failures, etc.). Set to true if the user navigated to the predicted URL. Eligibility status of preloading operation for the prediction. Reason why the preloading operation failed. The outcome of the holdback check. This is not part of eligibility status to make clarify that this check needs to happen after we are done verifying the eligibility of a preloading attempt. In general, eligibility checks can be reordered, but the holdback check always needs to come after verifying that the preloading attempt was eligible. Records how many times are we checking if a Service Worker is registered for an origin (fast check) or for a certain path of the origin (slow check) to decide if we are proceeding with a prefetch or not. Enum values would be: OriginOnly and Path. OriginOnly value describes the case when there are no Service Workers registered for the origin and we only need to run the fast check. Path value describes the case when there are Service Workers registered for the origin and we need to do a more expensive path check to find out if the prefetch URL falls under a registered Service Worker. Records how long (in microseconds) does the slow check for a Service Worker being registered for a certain path take when deciding if we are proceeding with a prefetch or not. We expect this check to take up to 2 ms. Exponentially bucketed using GetExponentialBucketMin with a value of 1.15. The type of predictor that made the prediction. Values are from either content::PreloadingPredictor or ChromePreloadingPredictor. The type of preloading operation used for the preloading attempt. The amount of time the attempt took to reach a state in which it is ready to be used for the next navigation, in milliseconds. Note: currently only Prefetch, Prerender and NoStatePrefetch have a ready state. This field is only meaningful to these three types of preloading. The sampling rate of this particular event multiplied by 1,000,000 (because we cannot log floating point values in UKM). A value of 1,000,000 means that the event had a 100% chance of being sampled, whereas a value of one means that the event had a one-in-a-millon chance of being sampled. The sampling rate for preloading attempts is set via field trial through the PreloadingConfig feature. The logged value just represents the configuration used for that preloading attempt. The eagerness of the speculation rule that triggered this preload. This is only set if the Preloading Attempt is triggered by a speculation rule. Values are from the Chrome enum blink::mojom::SpeculationEagerness. The amount of time the attempt took to reach the next navigation, in milliseconds. This can happen multiple times, but only the first is recorded. Specifies whether the preloading triggered and was used for the next navigation. isaboori@google.com mcnee@chromium.org Metrics that describe an anchor element, the user's interaction therewith, and whether a prediction of a navigation to the target destination would be accurate. This event is emitted on navigation. The number of events emitted at a time depends on the number of predictions the model makes, subject to sampling. See also NavigationPredictorAnchorElementMetrics and NavigationPredictorUserInteractions. 1 if the anchor element contains an image. The bucketed font size of the anchor element, according to its computed style. 1 corresponds to a pixel font size in [0, 10), 2 corresponds to a fonts size in [10, 18) and 3 is logged when the font size is 18 pixels or larger. 1 if the anchor element has an immediate text sibling, 0 otherwise. Dwell time of pointer hovering over the anchor element in milliseconds. Exponentially bucketed using GetExponentialBucketMin with a value of 1.3. 1 if a navigation to the anchor's destination occurs. 1 if the anchor's computed style's font-weight is greater than 500, 0 otherwise. 1 if the anchor element is in an iframe. 1 if the anchor element's target URL and the document's source URL only differ by one number, and the target URL increments that number in the source URL by 1. The time in ms between navigation start and the moment the anchor was first seen. Exponentially bucketed using GetExponentialBucketMin with a value of 1.3. The number of slashes in the anchor element's href URL path. Truncated at 5. The number of characters in the anchor element's href URL path. Buckets of size 10, truncated at 100. The percent of the total document area that this anchor element's clickable area occupies. How far down the document the top of the anchor element is, expressed as a ratio with the total document height. How many times the pointer was hovering over the anchor element. Exponentially bucketed using GetExponentialBucketMin a value of 1.3. 1 if the anchor element's target URL and the document's source URL are the same host. Indicates whether the number of predictions for the associated page was over a limit and needed to be downsampled. All events for a given page logged at the same time have the same SamplingAmount. The value is the exponentially bucketed value of (1,000,000 - (1,000,000 * sampling_rate)). The multiplication by 1,000,000 is because we cannot log floating point values in UKM, and we have larger values indicate more sampling so that a lack of sampling can be unambiguous by having its own bucket. For example, if the number of predictions was under the limit and no sampling was done, then the value would be 0. If the number of predictions was double the limit, the value would be 500,000. domenic@chromium.org chrome-instant-ecosystem@google.com This event is logged for each preloading prediction and metrics related to the preloading predictors like OmniboxDirectURLInput, Speculation Rules and others when the navigation is finished or when the WebContents is being destroyed. Associated with the Source ID of the page user navigates to. Contains information about the type of the predictor, its confidence value and whether the prediction ended up being correct. Whether the user navigated to the predicted URL. The confidence percentage of predictor's correct prediction for preloading. The type of preloading predictor that made the prediction. Values are from content::PreloadingPredictor or ChromePreloadingPredictor. Indicates whether the number of preloading predictions for the associated page was over a limit and needed to be downsampled. All events for a given page logged at the same time have the same SamplingAmount. The value is the exponentially bucketed value of (1,000,000 - (1,000,000 * sampling_rate)). The multiplication by 1,000,000 is because we cannot log floating point values in UKM, and we have larger values indicate more sampling so that a lack of sampling can be unambiguous by having its own bucket. For example, if the number of predictions was under the limit and no sampling was done, then the value would be 0. If the number of predictions was double the limit, the value would be 500,000. The amount of time the prediction took to reach the next navigation, in milliseconds. This can happen multiple times, but only the first is recorded. domenic@chromium.org chrome-instant-ecosystem@google.com This is the same as Preloading.Prediction but UKM is associated with the WebContents primary triggering page (if exists) and logged when the user navigates away or when the WebContents is being destroyed. Associated with the Source ID of the primary page where the preloading prediction is triggered (user navigates from). All the metrics data are similar to Preloading.Prediction and please see the comments for more details. Whether the user navigated to the predicted URL. The confidence percentage of predictor's correct prediction for preloading. The type of preloading predictor which made the preloading prediction. Values are from content::PreloadingPredictor or ChromePreloadingPredictor. Indicates whether the number of preloading predictions for the associated page was over a limit and needed to be downsampled. All events for a given page logged at the same time have the same SamplingAmount. The value is the exponentially bucketed value of (1,000,000 - (1,000,000 * sampling_rate)). The multiplication by 1,000,000 is because we cannot log floating point values in UKM, and we have larger values indicate more sampling so that a lack of sampling can be unambiguous by having its own bucket. For example, if the number of predictions was under the limit and no sampling was done, then the value would be 0. If the number of predictions was double the limit, the value would be 500,000. The amount of time the prediction took to reach the next navigation, in milliseconds. This can happen multiple times, but only the first is recorded. nhiroki@chromium.org chrome-prerendering@google.com Metrics related to Prerender2 (https://docs.google.com/document/d/1P2VKCLpmnNm_cRAjUeE-bqLL0bslL_zKqiNeCzNom_w/edit?usp=sharing). A status of a prerendering attempt. Recorded as enum PrerenderHostFinalStatus in //tools/metrics/histograms/enums.xml. This is replicated as Prerender.Experimental.PrerenderHostFinalStatus.* in UMA. The duration between the hardware timestamp and the start of event processing on the main thread for the first meaningful input. Recorded on the prerender activation. The number of distinct user interactions on a prerendered page after it is successfully activated. See definition of user interaction: https://web.dev/better-responsiveness-metric/#group-events-into-interactions An approximation of a high percentile of user interaction latency of a prerendered page after it is successfully activated, in ms. For this metric, we measure the latency of user interactions by Responsiveness.UserInteraction.MaxEventDuration. The maximum value of user interaction latency of a prerendered page after it is successfully activated, in ms. For this metric, we measure the latency of user interactions by Responsiveness.UserInteraction.MaxEventDuration. The number of distinct user interactions before any soft navigation happens. It includes the user interaction that triggers the 1st soft navigation. See definition of user interaction: https://web.dev/better-responsiveness-metric/#group-events-into-interactions An approximation of a high percentile of user interaction latency of a page load before any soft navigation happens, in ms. For this metric, we measure the latency of user interactions by Responsiveness.UserInteraction.MaxEventDuration. Measures the maximum cumulative layout shift (bit.ly/cls-explainer) that has occurred on the prerendered page that is successfully activated in a session window. The gap between two consecutive shifts in a window is not bigger than 1000ms and the maximum window size is 5000ms. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. Measures the maximum cumulative layout shift (bit.ly/cls-explainer) that has occurred in a session window before any soft navigation happens. The gap between two consecutive shifts in a window is not bigger than 1000ms and the maximum window size is 5000ms. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. A boolean indicating if any prerender activation page load mainpage resource has a 'Cache-control: no-store' response header. Recorded only for main frame request. page_load_metrics::NavigationHandleUserData::InitiatorLocation enum value. This field is used for identifying the types of prerender trigger. The |ui::PageTransition| for the main frame navigation that this activated prerendering navigation fulfilled. Number of soft navigations detected by heuristics. See https://github.com/WICG/soft-navigations Soft navigations are JS-driven same-document navigations that are using the history API or the new Navigation API, triggered by a user gesture and meaningfully modify the DOM, replacing the previous content with new one. The duration in milliseconds from time of activation navigation start to the time when the page first paints content. A contentful paint includes a paint of text, image, SVG, or canvas. The duration in milliseconds from time of activation navigation start to the time when the page first paints the largest content (text or image) within viewport, across all frames. The duration in milliseconds from the start of initial prerendering navigation to the start of activation navigation. This corresponds to the activationStart attribute of PerformanceNavigationTiming. The duration in milliseconds from time of activation navigation start to the time when the page first paints the largest content (text or image) within viewport, across all frames, before any soft navigation happens. A boolean that is set to true if the page triggered prerendering. Unlike other PrerenderPageLoad metrics, this is recorded not for prerendered page loads, but for page loads that initiated prerender. A boolean that is set to true if the page was prerendered and later activated. Cancelled prerender page loads are not logged. muyaoxu@google.com openscreen-eng@google.com Records when users start a Presentation and receive a response from the receiver. Android only. The following failures are not recorded because they don't represent an actual error: (1) PresentationErrorType::PRESENTATION_REQUEST_CANCELLED This error is created when users close the dialog without starting a Presentation. (2) PresentationErrorType::NO_PRESENTATION_FOUND This error might happen when a website implementing CastSDK try to join a route every time the page loads. Records whether the users' request to start a Presentation with a "cast:" presentation URL succeed. Records whether the users' request to start a Presentation through RemotePlayback succeed. yfriedman@chromium.org hartmanng@chromium.org pkotwicz@chromium.org PWA usage metrics recorded during a user's interaction with the site. Launch source of the WebAPK (tap from homescreen, WebShareTarget, deeplink, etc). Recorded every time a non-installed "WebAPKable" site is visited. juanmojica@google.com benwgold@google.com Metric recorded when a context menu item is selected. This is currently only recorded for Lens context menu items. The menu item selected by the user on the context menu. selya@google.com cfredric@chromium.org Records invocations of document.requestStorageAccessFor and reasons the request was approved or rejected. Recorded for every call to the function, regardless of outcome. The result of the requestStorageAccessFor call. selya@google.com cfredric@chromium.org Tracks when a cookie is included in a network request due to the ExemptionReason::kTopLevelStorageAccess. This state can only occur when a request is cors-enabled and made from the top-level frame. Reads from document.cookies (and requests that are not cors-enabled) are not included because the value is always 0 and those are not recorded. Cookies that have been written by document.cookies or network responses may be included if subsequent request meets the requirments for kTopLevelStorageAccess. Number of cookies where ExemptionReason::kTopLevelStorageAccess is the reason a cookie is included in a given request. The value is rounded down to the nearest exponential bucket with factor of 2.0. mmocny@chromium.org sullivan@chromium.org speed-metrics-dev@chromium.org Metrics that are associated with user interactions and recorded when users interact with web pages. Currently we are interested in keyboard, tap or click. A user interaction often encompasses multiple input events. For keyboard interactions, we only care about keydown and keyup. For tap or click we only measure pointerdown, pointerup and click. The type of the user interaction that triggered input event sequence, not the input event type. It can be keyboard, tap, click. The longest input event duration within the user interaction. The unit of duration is ms. For example, if a user pressed a key, we can get a keydown and a keyup. If the duration(keydown) > duration(keyup), the value of this metric will be duration(keydown). npm@chromium.org tdresser@chromium.org Responsiveness metrics recorded once every 5 seconds. skrakowi@chromium.org chrome-counter-abuse-alerts@google.com Records that the user encountered a Safe Browsing interstitial and whether they proceeded past the interstitial, bypassing the warning. A boolean signaling that the user bypassed the interstitial; only records true values. This event gets logged right when a user clicks to proceed on the Safe Browsing interstitial UI. A boolean signaling that the interstitial was shown to the user; only records true values. This event gets logged right when the Safe Browsing interstitial is shown to the user. skrakowi@chromium.org chrome-counter-abuse-alerts@google.com Recorded when the user takes a Safety Hub action after an abusive notification is auto-revoked. The type of action that the user has taken after being shown an abusive notification auto-revocation in Safety Hub. The reason for auto-revocation. olesiamarukhnno@google.com antoniosartori@chromium.org Recorded when an user interaction was detected for a site that was previously put in a potential revocation list for disruptive notification permissions or for a site for which notification permission was revoked. The metrics will be recorded on either page visit or a notification click. Average daily volume of notifications sent by the site, measured over the last 7 days. The number of days between the website was put in the potential revocation list and the detected increase in site engagement. New site engagement after user interaction with the site. Site engagement score when the site marked for revocation. The reason why the revocation was deemed a false positive: either a web site visit or a notification click. The revocation state (either proposed or actually revoked). olesiamarukhnno@google.com antoniosartori@chromium.org Recorded when enough user interactations were detected to be deemed a false positive revocation for a site that was previously put in a potential revocation list for disruptive notification permissions or for a site for which notification permission was revoked. The metrics will be recorded on either page visit or a notification click. A site engagement score has to increase by a minimum amount, set in feature params. A false positive revocation can only be reported after a min cooldown period has passed, set in feature params. Average daily volume of notifications sent by the site, measured over the last 7 days. The number of days between the website was put in the potential revocation list and the detected increase in site engagement. New site engagement after user interaction with the site. The notification clicks count, observered while the site is in the proposed revocation list. Site engagement score when the site marked for revocation. The page visits count, observered while the site is in the proposed revocation list. The revocation state (either proposed or actually revoked). olesiamarukhnno@google.com antoniosartori@chromium.org Recorded when Safety Hub would propose to revoke a disruptive notification permissions for a site. The check is performed every time a notification from the website is displayed. Notifications are considered disruptives when sent by websites with both very low site engagement index and high volume of notifications. After a revocation is proposed, this metric will not be reported anymore for the website for 30 days. After metrics were collected, during the full launch the permission will be revoked on the next Safety Check run. Average daily volume of notifications sent by the site, measured over the last 7 days. Site engagement for the disruptive site. olesiamarukhnno@google.com antoniosartori@chromium.org Recorded when a user has granted a notification permission for a site where notifications where revoked due to disruptive notifications. Recorded when user is visiting a site and has allowed notification permission for a site in the revoke list. Average daily volume of notifications sent by the site, measured over the last 7 days. The number of days between the website was put in the potential revocation list and the detected increase in site engagement. New site engagement after user interaction with the site. Site engagement score when the site marked for revocation. chrome-intelligence-core@google.com tbansal@chromium.org Metrics on the salient image of the page, one record per page load commit and has an annotation calculated on it. Records whether salient image was available or not for this page, the source of the salient image and why it was unavailable. bingler@chromium.org bingler@chromium.org bingler@chromium.org A collection of metrics which are recorded for cookies whose inclusion would be changed if the redirect chain was considered in their same-site calculation. This change in inclusion is because the same-site context would "downgrade" from a more permissive context to a less permissive context, one which may not allow a cookie's inclusion. Records the age of the cookies, in minutes, for cookies which are being read (on a requset or via JS). The value rounded down to the exponential bucket with factor of 2.0. Records the SameSite value of the cookie for cookies which are being read (on a request or via JS). Records the SameSite value of the cookie for cookies which are being written (on a response or via JS). qinmin@chromium.org shaktisahu@chromium.org A collection of metrics which are recorded for navigations inside saved tab groups. The metrics are recorded when a URL is synced from another client and Chrome finishes the navigation to that URL. A boolean indicating whether the current navigation is triggered by a POST request. Records how many times the same URL, including the current navigation, is loaded in the current tab in the past 2 minutes. Records how many times the same URL, without ref and query parameters, is loaded in the current tab in the past 2 minutes. The current navigation is also included in this count. Records the type of the tab group the navigation belongs to, such as shared or synced. See |tab_groups::SavedTabGroupTypes| for enum elements. A boolean indicating whether the current navigation involves any redirection. pdyson@chromium.org Collects metrics periodically and on screen brightness change. Percentage of battery, rounded down to a multiple of 5%, i.e. [0, 5%) is mapped to 0, [5%, 10%) is mapped to 5 etc. Screen percentage brightness. between 0 and 100. An enum representing the of the week that the data was logged, defined in |ash::power::ml::ScreenBrightnessEvent::Features::ActivityData:: DayOfWeek|. An enum representing the mode of the device, defined in |ash::power::ml::ScreenBrightnessEvent::Features::EnvData:DeviceMode| The hour of the day when the data is logged. Hours since midnight in the local time zone. Boolean value to represent whether the auto click is currently enabled. Boolean value to represent whether the braille display is currently enabled. Boolean value to represent whether the caret highlight is currently enabled. Boolean value to represent whether the cursor highlight is currently enabled. Boolean value to represent whether the focus highlight is currently enabled. Boolean value to represent whether the high contrast is currently enabled. Boolean value to represent whether the large cursor is currently enabled. Boolean value to represent whether the full screen magnifier is currently enabled. Boolean value to represent whether the mono audio is currently enabled. Boolean value to represent whether select to speak is currently enabled. Boolean value to represent whether the spoken feedback is currently enabled. Boolean value to represent whether Switch Access is currently enabled. Boolean value to represent whether video is currently playing. Boolean value to represent whether the virtual keyboard is currently enabled. Time between the last activity and when the data was logged. A duration timestamp in seconds. The temperature percentage of night light screen color adjustment when night light is on. Ranges from 0 to 100. The number of keyboard events in the last hour. The number of mouse events in the last hour. The number of stylus events in the last hour. The number of touch events in the last hour. Boolean value to represent whether the device is currently on battery power. The brightness of the screen prior to the time at which the current event was logged. The reason that the event is logged. Values are enumerated in |ash::power::ml::ScreenBrightnessEvent::Event::Reason|. How long the user has been active. A duration timestamp in seconds. The ID of this event in the current session. Time in seconds since last time an event was logged. jdeblasio@chromium.org enamelites@google.com beggs@google.com Collected when any safety tip heuristic triggers, regardless of whether the Safety Tip UI is shown. How the user interacted with the UI, if applicable. The final reported Safety Tip status. Whether the sensitive keywords heuristic triggered. Whether the lookalike keywords heuristic triggered. Whether the server-side blocklist heuristic triggered. Whether the user had already ignored a safety tip warning for this page in the past, meaning no tip was shown. cthomp@chromium.org enamelites@google.com Site engagement score behavior for a page with a given Security Level. The final SecurityLevel of the page before it is closed or another navigation occurs. Based on the enum security_state::SecurityLevel. The initial SecurityLevel of the page when the navigation commits. Based on the enum security_state::SecurityLevel. The Safety Tip status of the page when the navigation commits, after the reputation check finishes. The change in Site Engagement score for the page before it is closed or another navigation occurs compared to the score from when the navigation committed. Score is between 0.0 and 100.0; delta is between -100.0 and 100.0. Rounded to the nearest integer. The final Site Engagement score for the page before it is closed or another navigation occurs. Score is between 0.0 and 100.0; this metric is rounded down to a multiple of 10 to limit granularity. nyquist@chromium.org qinmin@chromium.org ssid@chromium.org Input and output tensors for executing the ML model identified by optimization targets and version. The input tensors can come from UMA or UKM, with or without aggregation. Check model metadata for more information about each tensor. Since all the ML tensors are floats, they are encoded into int64 and will be decoded later during model training. This metric is recorded after an ML model is executed, and the actual result is collected. UMA or UKM metrics related to the features affected by the ML model, with or without aggregation, after running the features as instructed by the ML model. For example, this can be the 7 day total button clicks after an ML model decides to show it. No encoding is used as this is normally a count or sum of a UMA/UKM metric. The 10th UMA or UKM metrics related to the features affected by the ML model, with or without aggregation, after running the features as instructed by the ML model. No encoding is used. The 2nd UMA or UKM metrics related to the features affected by the ML model, with or without aggregation, after running the features as instructed by the ML model. No encoding is used. The 3rd UMA or UKM metrics related to the features affected by the ML model, with or without aggregation, after running the features as instructed by the ML model. No encoding is used. The 4th UMA or UKM metrics related to the features affected by the ML model, with or without aggregation, after running the features as instructed by the ML model. No encoding is used. The 5th UMA or UKM metrics related to the features affected by the ML model, with or without aggregation, after running the features as instructed by the ML model. No encoding is used. The 6th UMA or UKM metrics related to the features affected by the ML model, with or without aggregation, after running the features as instructed by the ML model. No encoding is used. The 7th UMA or UKM metrics related to the features affected by the ML model, with or without aggregation, after running the features as instructed by the ML model. No encoding is used. The 8th UMA or UKM metrics related to the features affected by the ML model, with or without aggregation, after running the features as instructed by the ML model. No encoding is used. The 9th UMA or UKM metrics related to the features affected by the ML model, with or without aggregation, after running the features as instructed by the ML model. No encoding is used. The input tensor at index 0 represented in IEEE 754 double precision. The input tensor at index 1 represented in IEEE 754 double precision. The input tensor at index 10 represented in IEEE 754 double precision. The input tensor at index 11 represented in IEEE 754 double precision. The input tensor at index 12 represented in IEEE 754 double precision. The input tensor at index 13 represented in IEEE 754 double precision. The input tensor at index 14 represented in IEEE 754 double precision. The input tensor at index 15 represented in IEEE 754 double precision. The input tensor at index 16 represented in IEEE 754 double precision. The input tensor at index 17 represented in IEEE 754 double precision. The input tensor at index 18 represented in IEEE 754 double precision. The input tensor at index 19 represented in IEEE 754 double precision. The input tensor at index 2 represented in IEEE 754 double precision. The input tensor at index 20 represented in IEEE 754 double precision. The input tensor at index 21 represented in IEEE 754 double precision. The input tensor at index 22 represented in IEEE 754 double precision. The input tensor at index 23 represented in IEEE 754 double precision. The input tensor at index 24 represented in IEEE 754 double precision. The input tensor at index 25 represented in IEEE 754 double precision. The input tensor at index 26 represented in IEEE 754 double precision. The input tensor at index 27 represented in IEEE 754 double precision. The input tensor at index 28 represented in IEEE 754 double precision. The input tensor at index 29 represented in IEEE 754 double precision. The input tensor at index 3 represented in IEEE 754 double precision. The input tensor at index 30 represented in IEEE 754 double precision. The input tensor at index 31 represented in IEEE 754 double precision. The input tensor at index 32 represented in IEEE 754 double precision. The input tensor at index 33 represented in IEEE 754 double precision. The input tensor at index 34 represented in IEEE 754 double precision. The input tensor at index 35 represented in IEEE 754 double precision. The input tensor at index 36 represented in IEEE 754 double precision. The input tensor at index 37 represented in IEEE 754 double precision. The input tensor at index 38 represented in IEEE 754 double precision. The input tensor at index 39 represented in IEEE 754 double precision. The input tensor at index 4 represented in IEEE 754 double precision. The input tensor at index 40 represented in IEEE 754 double precision. The input tensor at index 41 represented in IEEE 754 double precision. The input tensor at index 42 represented in IEEE 754 double precision. The input tensor at index 43 represented in IEEE 754 double precision. The input tensor at index 44 represented in IEEE 754 double precision. The input tensor at index 45 represented in IEEE 754 double precision. The input tensor at index 46 represented in IEEE 754 double precision. The input tensor at index 47 represented in IEEE 754 double precision. The input tensor at index 48 represented in IEEE 754 double precision. The input tensor at index 49 represented in IEEE 754 double precision. The input tensor at index 5 represented in IEEE 754 double precision. The input tensor at index 6 represented in IEEE 754 double precision. The input tensor at index 7 represented in IEEE 754 double precision. The input tensor at index 8 represented in IEEE 754 double precision. The input tensor at index 9 represented in IEEE 754 double precision. The version of the ML model used for execution. May not be a counter int. The OptimizationTarget associated with the ML model. This is an enum for classifying different ML models. It represents what an model is trying to optimize. The time from last segment selection to when the UKM is recorded, in seconds. Predicted result from executing the ML model. This is normally a score for determining whether to show a particular feature. The value is a floating number between 0 to 1 and encoded into IEEE 754 double precision here. This is being deprecated. Use PredictionResult1 instead. Will add deprecated flag once all the code are start using PredictionResult1. The 1st output when inferencing the ML model with input data. The value is a floating number between 0 to 1 and encoded into IEEE 754 double precision here. The 10th output when inferencing the ML model with input data. The value is a floating number between 0 to 1 and encoded into IEEE 754 double precision here. The 2nd output when inferencing the ML model with input data. The value is a floating number between 0 to 1 and encoded into IEEE 754 double precision here. The 3rd output when inferencing the ML model with input data. The value is a floating number between 0 to 1 and encoded into IEEE 754 double precision here. The 4th output when inferencing the ML model with input data. The value is a floating number between 0 to 1 and encoded into IEEE 754 double precision here. The 5th output when inferencing the ML model with input data. The value is a floating number between 0 to 1 and encoded into IEEE 754 double precision here. The 6th output when inferencing the ML model with input data. The value is a floating number between 0 to 1 and encoded into IEEE 754 double precision here. The 7th output when inferencing the ML model with input data. The value is a floating number between 0 to 1 and encoded into IEEE 754 double precision here. The 8th output when inferencing the ML model with input data. The value is a floating number between 0 to 1 and encoded into IEEE 754 double precision here. The 9th output when inferencing the ML model with input data. The value is a floating number between 0 to 1 and encoded into IEEE 754 double precision here. The OptimizationTarget of the last selected segment before UKM is recorded. Some features might have several ML models with them. This metric records which model is currently selected for the feature. yyanagisawa@chromium.org chrome-worker@google.com Recorded when a Service Worker fetch handler fails to load a main resource due to a network error (such as "failed to fetch" errors occurring in fetch() inside the fetch handler) or timing out during race-network-and-fetch or normal fetch handler execution. Indicates how the fetch handler was invoked. The specific net::ERR_ code indicating the reason for the failure. (possible values are listed in net_error_list.h). Records the network error that occurred when the Service Worker static routing API's 'race-network-and-fetch' handler deduplicates the network fetch with the fetch() API call to the navigation URL. This metric is only populated when this static routing strategy is active. The specific net::ERR_ code indicating the reason for the failure. (possible values are listed in net_error_list.h). Records the network error of the initial fetch() API call in cases where the deduplication mentioned above does not apply. suzukikeita@chromium.org chrome-worker@google.com Events taken when the main resource is loaded. This only occurs when the static routing API is used. The actual router source type for Static Routing API. This indicates which source type was actually used to get the resource. This will be recorded only when the router has matched the incoming resource. Time spent to conduct cache lookup (right before creating the cache lookup request to right after the browser process receives the cache result). Recorded when the service worker has associated routing info with the `cache` as the rule, and the browser or renderer is processing requests. Duration in units of milliseconds. ServiceWorker running status at the time when the SW is handling the main resource. Recorded regardless of the router existence via Static Routing API. The result of the router evaluation of the Static Routing API. This indicates which source type is specified to get the resource. This will be recorded only when the router has matched the incoming resource. Time spent for the SW static routing API router evaluation to handle a request. Recorded when the service worker has associated routing info, and the browser or renderer is processing requests. Duration in units of micro seconds. The sum of registered router rules with the Static Routing API in main resource. Recorded per navigation when the service worker, which has associated routing info, is activated. This is rounded down to the nearest exponential bucket (with a bucket ratio on 1.15). ServiceWorker running status at the time when router evaluation is conducted. We expect this is equivalent to `InitialWorkerStatus` in most cases. However, those are set in different tasks (ctor and `StartRequest()` in SWMainResourceLoader) and some other task may update the status between them. We keep both metrics just in case. yyanagisawa@chromium.org chrome-worker@google.com Events taken at the end of the loading pages controlled by the service workers. It means that the metrics is recorded only if the loading is successfully completed. True if there are network fallback sub-resources of audios in the page. (boolean value. true: 1 or false: 0) True if there are service workers handling sub-resource loads of audios in the page. (boolean value. true: 1 or false: 0) True if there are network fallback sub-resources of CSSStyleSheet in the page. (boolean value. true: 1 or false: 0) True if there are service workers handling sub-resource loads of CSSStyleSheet in the page. (boolean value. true: 1 or false: 0) True if there are network fallback sub-resources of Dictionary in the page. (boolean value. true: 1 or false: 0) True if there are service workers handling sub-resource loads of Dictionary in the page. (boolean value. true: 1 or false: 0) True if there are network fallback sub-resources of CSSStyleSheet in the page. (boolean value. true: 1 or false: 0) True if there are service workers handling sub-resource loads of fonts in the page. (boolean value. true: 1 or false: 0) True if there are network fallback sub-resources of images in the page. (boolean value. true: 1 or false: 0) True if there are service workers handling sub-resource loads of images in the page. (boolean value. true: 1 or false: 0) True if there are network fallback sub-resources of link prefetches in the page. (boolean value. true: 1 or false: 0) True if there are service workers handling sub-resource loads of link prefetches in the page. (boolean value. true: 1 or false: 0) Enum value to represent how main / sub resource are loaded. True if there are network fallback sub-resources of manifests in the page. (boolean value. true: 1 or false: 0) True if there are service workers handling sub-resource loads of manifests in the page. (boolean value. true: 1 or false: 0) Total number of sub-resources which were matched to `RouterSourceEnum.cache` in ServiceWorker Static Routing API, from navigation start until onload event. This is rounded down to the nearest exponential bucket (with a bucket ratio on 1.15). Total number of sub-resources which were matched to `RouterSourceEnum.fetch-event` in ServiceWorker Static Routing API, from navigation start until onload event. This is rounded down to the nearest exponential bucket (with a bucket ratio on 1.15). Total number of sub-resources which were matched to `RouterSourceEnum.network` in ServiceWorker Static Routing API, from navigation start until onload event. This is rounded down to the nearest exponential bucket (with a bucket ratio on 1.15). Total number of sub-resources which were matched to the `RouterSourceEnum.race-network-and-fetch-handler` in ServiceWorker Static Routing API, from navigation start till onload event. This is rounded down to the nearest exponential bucket (with a bucket ratio on 1.15). True if there are network fallback sub-resources of CSSStyleSheet in the page. (boolean value. true: 1 or false: 0) True if there are service workers handling sub-resource loads of scripts in the page. (boolean value. true: 1 or false: 0) True if there are network fallback sub-resources of speculation rules in the page. (boolean value. true: 1 or false: 0) True if there are service workers handling sub-resource loads of speculation rules in the page. (boolean value. true: 1 or false: 0) An integer ratio (0 - 100) to represent the ratio of network fallbacks from the total subresource load. True if there are network fallback sub-resources of SVG documents in the page. (boolean value. true: 1 or false: 0) True if there are service workers handling sub-resource loads of SVG documents in the page. (boolean value. true: 1 or false: 0) True if there are network fallback sub-resources of text tracks in the page. (boolean value. true: 1 or false: 0) True if there are service workers handling sub-resource loads of text tracks in the page. (boolean value. true: 1 or false: 0) Total time spent to conduct cache lookup. Recorded when the service worker has associated routing info with the `cache` as the rule, and the browser or renderer is processing requests. The field is the sum of all cache lookup time of subresources. i.e. if there are X resources whose router source is `cache` for example, this value will be the sum of X cache lookup durations regardless of cache hit or miss. Duration in units of milli seconds. Total time spent for the router evaluation to handle a request. Recorded when the service worker has associated routing info, and the browser or renderer is processing requests. Duration in units of micro seconds. The number of network fallbacks during sub resource load. This is rounded down to the nearest exponential bucket (with a bucket ratio on 1.15). The number of all sub resource load of the page controlled by the service worker. This is rounded down to the nearest exponential bucket (with a bucket ratio on 1.15). True if there are network fallback sub-resources of videos in the page. (boolean value. true: 1 or false: 0) True if there are service workers handling sub-resource loads of videos in the page. (boolean value. true: 1 or false: 0) True if there are network fallback sub-resources of XSLStyleSheet in the page. (boolean value. true: 1 or false: 0) True if there are service workers handling sub-resource loads of XSLStyleSheet in the page. (boolean value. true: 1 or false: 0) jeffreycohen@chromium.org chrome-creation@google.com Event representing the user action of selecting a piece of text on a page and selecting the option to create a link which would contain the necessary text fragment information to link back to this specific part of the page. Record the type of error which prevented a successful link generation. Only recorded if the link creation failed. Whether the generation of a link with text fragments was successful or not. jeffreycohen@chromium.org chrome-creation@google.com Event representing the user action of selecting a piece of text on a page and selecting the option to create a link which would contain the necessary text fragment information to link back to this specific part of the page. Record the type of error which prevented offering link to text to user. Only recorded if the link was not offered to the user. Whether the link with text fragments was successful offered to the user or not. jeffreycohen@chromium.org chrome-creation@google.com Event representing the browser action of highlighting (incl. scroll-to) text fragments that were part of the URL. This happens only once during page load. Record the type of source for the current navigation, which has text fragments. At the moment, this maps to identifying whether the source is a known search engine, Send Tab To Self, or unknown. Whether the highlighting of a set of text fragments was a complete success or not. yaoxia@chromium.org cammie@chromium.org Metrics for Shared Storage worklet. Each document can create zero or more worklets, and the recorded URL represents the URL of the top-level page that creates the worklet. (Note: The Storage.SharedStorage.Worklet.NumPerPage histogram tracks the number of worklets per page.) Recorded when a worklet is destroyed. See https://wicg.github.io/shared-storage/#worklet. The total duration (ms) that a Shared Storage worklet is actively processing operations (i.e, from the worklet's creation until either its last operation completes or it is destroyed). The number is exponentially bucketed. knollr@chromium.org Collected when a user finishes a Click to Call journey. Specifies the entry point of the user journey (i.e. left click, right click, text selection). Boolean value indicating whether any (>0) apps are available that the call could be sent to. Boolean value indicating whether any (>0) devices are available that the call could be sent to. These belong to the same account and are running Android. Whether the user selected a device, an app or nothing (i.e. dismissal) in the user interface. For some entry points this might always be the same value (e.g. right clicks always select devices). yuezhanggg@chromium.org chrome-shopping-eng@google.com Recorded when we identified a set of URLs that users might be doing comparison in using the client-side heuristics, and after checking if these URLs are comparable based on a server call. This event is recorded one per URL for all URLs in the set. Whether the URL in this client-identified cluster is also comparable based on the response from querying the server. An ID corresponding to the event that will enable joining all active tabs for this event together. meiliang@chromium.org ayman@chromium.org chrome-shopping-eng@google.com Recorded when a user navigates to a product details page with an available discount that is eligible for auto popup. This gives us the true auto popup population. Whether the eligible popup bubble was suppressed. This is controlled by a feature flag. Logging the feature flag value to create two groups for data analysis. tommasin@chromium.org privacy-on-the-web@google.com Recorded when the merchant trust bubble is opened. A boolean signaling that the bubble is opened. Only records true values. zhiyuancai@chromium.org ayman@chromium.org Recorded when user finishes a navigation while the merchant trust data is available. A boolean signaling that data is available. Only records true values. zhiyuancai@chromium.org ayman@chromium.org Recorded when a merchant trust message is clicked by the user. A boolean signaling that message is clicked. Only records true values. zhiyuancai@chromium.org ayman@chromium.org Recorded when a merchant trust message is displayed to the user. A boolean signaling that message is displayed. Only records true values. zhiyuancai@chromium.org ayman@chromium.org Recorded when a store info row in trusted surface is clicked by the user. A boolean signaling that store info row is clicked. Only records true values. zhiyuancai@chromium.org ayman@chromium.org Recorded when a store info row in trusted surface is displayed to the user. A boolean signaling that store info row is displayed. Only records true values. tommasin@chromium.org privacy-on-the-web@google.com Recorded when the merchant trust side panel is opened. A boolean signaling that the side panel is opened. Only records true values. yuezhanggg@chromium.org mdjones@chromium.org Records if and how a page has been determined to be a product details page. It is recorded per eligible navigation (e.g. limited to main frame navigation, http/https check, etc), and is recorded after the page has finished loading. The state of if and how the page has been determined to be a product details page. yuezhanggg@chromium.org mdjones@chromium.org Recorded when an user takes action on shopping features in Chrome in the current page. Whether the user has clicked in the discount bubble to copy the discount code. Whether the user has clicked and expanded the discount bubble in omnibox. Whether the user has clicked the page action entry point to request to show the price insights. Whether the user has tracked the product on the current page using any of the on-page entry points. yuezhanggg@chromium.org mdjones@chromium.org Recorded when an user navigates to a shopping-related page and reports shopping-related information about this page. Whether a discount from Chrome is available for this page. Whether price insights is available for this page. Whether the page is price trackable. Whether the page is classified as shopping content. Which, if any, contextual shopping feature is promoted in expanded state or in-product highlight. dibyapal@chromium.org pwa-team@google.com Records various metrics about whether a particular site has been installed as a PWA or not. Measured if the site matches the scope of any installed PWAs present in the profile. Measured if the app is partially installed via sync. This means that the user has installed the app on another device, and it exists in their sync profile, but it is not 'fully' installed on this device. This is expected to be a strong signal that the user will want to install the app fully on this device, as they already have it installed elsewhere. dibyapal@chromium.org pwa-team@google.com Records various signals regarding the fields present inside a manifest. If the DisplayMode is undefined, HasStartUrl is kEmpty, and all the other fields are false, then that denotes a site that has no manifest. For all other use-cases, the existence of a manifest is confirmed. Records the display mode populated inside the manifest as per the manifest specifications. If it does not exist, this is measured as Undefined. Records whether the manifest has the background_color field populated. Records whether the manifest has icons of purpose ANY. Records whether the manifest has maskable icons. Records whether the manifest has the name field populated. Records whether the manifest has valid screenshots specified. Records whether the manifest has a valid URL in the start_url field. Records whether the manifest has the theme_color field populated. dibyapal@chromium.org pwa-team@google.com Records various signals about a site that boost its chances of being installed as a web app, like valid fetch handlers (boosting offline capabilities), service worker data usage and whether the site has favicons. Measures the service worker cache storage size as determined by the Quota system for a site in bytes. Records whether the site has non-default favicons or not. Non-default favicons are those which are either explicitly set in the client side, or are icons obtained from the manifest. The default favicon set by chrome, that follows document_url/favicon.ico is not counted as part of this metric. Records if the service worker exists and has a valid fetch handler. Measured the size of the service worker script in bytes. If the service worker does not exist, then a script size of 0 is measured. pkotwicz@chromium.org chrome-counter-abuse-alerts@google.com Recorded when a toplevel page navigates. This records which user-page-familiarity-heuristics would trigger for a given site, if any. Whether any familiarity heuristics would trigger on the site. Whether the site is on the safe browsing high confidence allowlist. Site engagement score in the range [0, 100], rounded down to a multiple of 10 to limit granularity. The most stringent navigation-history related familiarity heuristic that would trigger on the site. If no navigation-history heuristic matches then the value NO_HEURISTIC_MATCH is recorded. thanhdng@chromium.org jiameng@chromium.org Records user activities that are related to charging such as battery percentage, time, device type, etc. Percentage of the battery. The percentage of the battery at the beginning of the last charge. The percentage of the battery that the last charge reached. Event's day of the month (e.g. 1-31) in local time zone. Event's day of the week in local time zone. Sunday is 0. Mode of the device (e.g. closed lid, tablet). Type of the device (e.g. laptop, tablet). Duratioin of the last time the device was charged in minutes. Duration of audio playing in the last 30 minutes (bucketized and in minutes). Duration of video playing in the last 30 minutes (bucketized and in minutes). A unique number that represent order of an event. Whether there's any shutdown/suspend actions between the last charge and current event. Whether the device is being charged or not. Event's month (e.g. Jan-Dec) in local time zone. January is 1. Number of key events in past 30 minutes. Number of mouse events in past 30 minutes. Number of stylus events in past 30 minutes. Number of touch events in past 30 minutes. Reason for the logging of an event. Brightness of the screen in percent. Time of the event in minutes since midnight in the local time zone. Time since the last time user unplugged the charger in minutes. Timezone difference from the last charge. It is equal to current_timezone - timezone_from_last_charge. The valid range of time zone value will be -12 (UTC -12:00) to +14 (UTC +14:00). Charge voltage in mV. goto@chromium.org reillyg@chromium.org ayui@chromium.org yigu@chromium.org Records performance metrics for SMSReceiver API. Flag indicating whether the report comes from a cross-origin frame or not. Records the result of a call to the SmsReceiver API. Records the status of parsing an incoming SMS when using the WebOTP API. It records one sample per incoming SMS. Records the duration from when the API is called to when an SMS has been successfully received. Records the duration from when the API is called to when the user presses "Allow" to pass the incoming SMS to the site and proceed with the SMS verification flow. Records the duration from when the API is called to when the user presses "Deny" to terminate the SMS verification flow. mmocny@chromium.org speed-metrics-dev@chromium.org Records soft navigation metrics including LCP, CLS, INP. Soft navigation is DOM changes that are perceived to be navigation, but there is no page load. Soft navigation metrics like LCP are equivalent to their counterpart of the PageLoad event. One-based offset in the list of all user interactions of the interaction which was the INP (https://web.dev/inp). The time (in milliseconds from navigation start) of the interaction which was the INP (https://web.dev/inp). The number of distinct user interactions from the time when the soft navigation happens to the time when the next soft navigation happens or the page load end. See definition of user interaction: https://web.dev/better-responsiveness-metric/#group-events-into-interactions An approximation of a high percentile of user interaction latency from the time when the soft navigation happens to the time when the next soft navigation happens or the page load end, in ms. For this metric, we measure the latency of user interactions by Responsiveness.UserInteraction.MaxEventDuration. Measures the maximum cumulative layout shift (bit.ly/cls-explainer) that has occurred from the time when the soft navigation happens to the time when the next soft navigation happens or the page load ends in a session window. The gap between two consecutive shifts in a window is not bigger than 1000ms and the maximum window size is 5000ms. This metric's integral value is 100x the fractional cumulative layout shift score described in the explainer. Deprecated as of 2026/02/17. Records the navigation id of a soft navigation. It is used to differentiate soft navigations. Type of the soft navigation. Corresponds to the first URL change. Indicates whether the soft navigation occurred within a PageLoad event (value 1), a PrerenderPageLoad event (value 2), or a HistoryNavigation event (value 3). See page_load_metrics::PageLoadType. Measures the FirstContentfulPaint in milliseconds from the start time of the soft navigation to the presentation time. Measures the LargestContentfulPaint in milliseconds from the start time of the soft navigation to the presentation time. Measures the ratio of image file size (in bits) to displayed pixels for the LCP candidate. Bucketed into roughly exponential intervals to account for the wide range of encountered values (less than 1:100000 to greater than 10000:1). This metric is recorded whenever the LCP candidate is an image, and omitted when it is text. Measures the time in milliseconds from soft navigation's start time to the time when the image LCP element is requested. This metric is not recorded when the LCP element is text or video. Measures the time in milliseconds from soft navigation's start time to the time when the image LCP element finishes loading. For animated images, this is when the image resource finishes loading completely, not the first frame time. See web.dev/optimize-lcp for more details. This metric is not recorded when the LCP element is text or video. Measures the time in milliseconds from soft navigation's start time to the time when the image LCP element starts loading. See web.dev/optimize-lcp for more details. This metric is not recorded when the LCP element is text or video. Resource loading priority used for the LCP candidate. This metric is recorded whenever the LCP candidate is an image, and omitted when it is text or video. Indicates that type of the LargestContentfulPaint candidate. This value is a packed set of flags, defined in third_party/blink/public/common/performance/largest_contentful_paint_type.h Records the start time of a soft navigation. The start time is a time delta in milliseconds relative to the last hard navigation time. It is used to analyze the frequency of soft navigations. agale@chromium.org top-chrome-desktop-ui@google.com Recorded when a new split view is created. An ID corresponding to the event that will enable joining both tabs in the split together. The orientation of the split which can be horizontal or vertical. agale@chromium.org top-chrome-desktop-ui@google.com Recorded when a split view is updated which could mean a navigation happened within a split view or another tab is swapped into a split. An ID corresponding to the event that will enable joining both tabs in the split together. The orientation of the split which can be horizontal or vertical. vkrot@google.com iwa-team@google.com Recorded when chrome starts. Sub apps count per parent app, more then 0. vkrot@google.com iwa-team@google.com Recorded when subApp install either succeeds or fails per subApps.add call. Result enum of install operation of a single sub app. vkrot@google.com iwa-team@google.com Records the result of web app manifest updating a security sensitive field (name and icons) for a sub app. Result enum of apply pending update operation of a single sub app. vkrot@google.com iwa-team@google.com Records the result of web app manifest silently updating during page navigations for a sub app. Result enum of manifest silent update operation of a single sub app. csharrison@chromium.org Recorded for page loads where subresource filtering policy was evaluated. Enum that records the policy decision to activate subresource filtering for a page load. 0 = Unknown, 1 = Activated, 2 = Disabled, 3 = Deprecated value, 4 = URL whitelisted, 5 = Activation conditions not met. 'Activated' indicates that subresource filtering was activated. All other reasons indicate that subresource filtering was not activated. Records '1' if subresource filtering was activated in dry run mode. In dry run mode, subresource filtering policy is evaluated against resources loaded on the page, but subresource filtering is not actually applied. An enum denoting where in the redirect chain there was a URL that matched one of the subresource_filter activation lists, and wasn't marked as warning. See subresource_filter::RedirectPosition for values. nidhijaju@chromium.org blink-network-stack@google.com Event recorded when a response is received that is content-encoded with zstd and is a subresource load. Set to 1 if the subresource served zstd-compressed content. rajendrant@chromium.org mcrouse@chromium.org Records coverage information for src videos. Recorded once for every src video that is fetched from network. The full content length of the video retrieved from the first range request for the resource. This full content length represents the total bytes when the entire video is played. The bytes are rounded down to the exponential bucket with factor of 1.3. Enumerates the different results possible for subresource redirection, such as redirectable or different reasons of ineligibility. skrakowi@chromium.org chrome-counter-abuse-alerts@google.com Recorded when the user takes an action on a suspicious notification whose contents were replaced by a warning. The type of action that the user has taken after receiving a notification content detection warning. The 'suspicious' score given to notification contents that came from the URL. Ranges from 0 to 100. sayman@chromium.org User took one or more screenshots of a page. A boolean signaling that the event has occurred; only records true values. madhavpruthi@google.com ckitagawa@chromium.org clank-tab-dev@chromium.org A tab's state has changed. Only recorded for loaded tabs with WebContents. Whether the tab is currently pinned. This is stored only for tabs which are pinned, i.e. this metric will always be true. chrisha@chromium.org lpy@chromium.org Collects the duration in MS from when the tab is backgrounded to when a JavaScript alert is fired. Only recorded when the tab is in the background. Indicates whether the alert is fired from main frame. Duration in MS from when the tab is backgrounded to when alert is fired. chrisha@chromium.org lpy@chromium.org Collects the duration in MS from when the tab is backgrounded to when audio starts to play. Only recorded when the tab is in the background. Indicates whether the audio stream belongs to the main frame. Duration in MS from when the tab is backgrounded to when audio starts to play. chrisha@chromium.org lpy@chromium.org Collects the duration in MS from when the tab is backgrounded to when the favicon is updated. Only recorded when the tab is in the background. Duration in MS from when the tab is backgrounded to when the favicon is updated. chrisha@chromium.org lpy@chromium.org Indicates whether the notification creation is from the main frame. Duration in MS from when the tab is backgrounded to when a non-persistent notification is created. chrisha@chromium.org lpy@chromium.org Collects the duration in MS from when the tab is backgrounded to when the title is updated. Only recorded when the tab is in the background. Duration in MS from when the tab is backgrounded to when the title is updated. chrisha@chromium.org Collects the duration in MS from when the tab is backgrounded to when it is brought to foreground or closed. Boolean value indicating whether or not the tab is discarded since it was backgrounded. Boolean value indicating whether or not it's triggered because the tab is brought to foreground again. An int64 random number generated when TabManager.TabMetric is logged. The same number is also logged here, so that feature label can be paired. Duration in MS from when the tab is backgrounded to when it is brought to foreground or closed. zhenw@chromium.org Collects data when a tab is switched to from another tab and then finishes loading in the foreground during session restore The event is only recorded when a tab is switched to from another tab within the same tabstrip. As a result, the initial foreground tab is not included in this event since it was not switched to from another tab. The event is not recorded when the session overlaps with background tab opening. The ID of this event's sequence in current session restore session. The ID of this session restore session. Number of tabs that are restored. Number of all tabs of the system, which includes all browser windows. Tab switch load time in MS. It is defined as the time between when the user switches to a background tab, and the time when that tab finishes loading in the foreground. If the user switches away before the tab finishes loading, a metric will not be recorded unless the user switches back, in which case the tab switch load time is measured from that point in time. zhenw@chromium.org Collects the expected queueing duration of tasks and metadata for the foreground tab during session restore. The event is not recorded when the session overlaps with background tab opening. The event is not recorded when there is only one restored tab, in which case, there is no background restored tab and we are not interested at that. In this way, we cap the volume while keeping all interesting data. Duration in MS for expected task queueing duration. The metric reflects the responsiveness of a tab. A lower value means the tab will respond to inputs faster. This metric is equal to RendererScheduler.ExpectedTaskQueueingDuration. It is emitted once for all tasks in each 1000-ms window. The ID of this event's sequence in current session restore session. The ID of this session restore session. Number of tabs that are restored. Number of all tabs of the system, which includes all browser windows. chrisha@chromium.org Collected when a tab is closed, at most once per source. The time when the tab was closed, expressed as the amount of time in MS that has elapsed since the main frame navigation providing the content of the tab. michaelpg@chromium.org Collects information about a browser or app window to associate with the TabManager.TabMetrics entries for tabs in that window. Whether the window is the active (frontmost) window. Enumeration of the window show state, such as fullscreen or minimized. Values are enumerated in metrics::WindowMetricsEvent::ShowState. Number of tabs in the tab strip. Rounded down to the nearest exponential bucket (with a bucket spacing factor of 1.5). Will be 1 for windows with only one top-level WebContents, such as app windows. Enumeration for the type of the window. Values are enumerated in metrics::WindowMetricsEvent::Type. Session ID of the browser or app window this entry represents, unique for the browsing session. anthonyvd@chromium.org chrome-catan@google.com Recorded when a tab changes state to/from "active", "background", and "closed". The state the tab is in after the transition. The count of tabs with the same main frame URL as this one in the current browser session. Used to identify instances of duplicate tabs. Bucketed in 20 buckets (0 to 20) + an overflow bucket. The count of times this tab was revisited since it was created. Bucketed in 20 buckets (0 to 20) + an overflow bucket. The state the tab was in before the transition. Whether the tab is in the same tab group as the current active tab. Whether the tab is in the same browser window as the current active tab. The time spent by the tab in `PreviousState`, before transitioning to `NewState`, in seconds. The total time the tab has spent as the active tab since it was created, in seconds. olivierrobin@chromium.org rohitrao@chromium.org Recorded when a Sad Tab is displayed or interacted with. This allows correlating specific URLs and page contexts with renderer crashes that lead to Sad Tabs and user feedback. The type of the SadTabEvent that occurred (e.g. Displayed, Button clicked, Learn more clicked). True if the Sad Tab is in feedback mode (typically due to a repeated failure). False if it is in reload mode. sauski@google.com kmg@google.com Recorded each time User Bypass is activated to bypass 3PC blocking and re-enable 3PC for the page. Only applicable if the user is blocking 3PC by default. Whether the page attempted to call the FedCM API on the current page load, prior to the user re-enabling 3PC. The number of times the user refreshed the current page (without navigating elsewhere) before re-enabling 3PC. An integer clamped to the range 0 to 10. Whether the user has previously re-enabled 3PC on this page via the in-context controls. The site engagement level of the page when the user re-enabled 3PC. Whether the page had requested access to storage via the Storage Access API on the current page load, prior to the user re-enabling 3PC The state of site data access (e.g. blocked, allowed, etc) for the the current page load prior to the user re-enabling 3PC. nyquist@chromium.org ssid@chromium.org User pressed the 'Share' in the top toolbar, which shares the current URL. A boolean signaling that the event has occurred; only records true values. kazinova@google.com friedrichh@chromium.org kolos@chromium.org Recorded when a user takes an action on the Touch To Fill sheet. The action a user took when interacting with the Touch To Fill sheet. kolos@chromium.org friedrichh@chromium.org Recorded when the Touch To Fill sheet is shown to the user. A given form's state (e.g. there are empty fields in the form, or there are only two fields and they both are filled by Chrome) that indicates whether the form is ready for submission. Recorded when a Touch To Fill sheet is shown for that form. kolos@chromium.org friedrichh@chromium.org Recorded when a successful login is detected after filling by the Touch To Fill. The time in milliseconds between filling by the Touch To Fill and a successful submission (To ensure filling and submitting correspond to each other, the usernames should coincide and the timestamps should be within one minute. A successful login after a failed submission or manual typing is not reported). Used to compare the times when automated submission is enabled and disabled. Thus, we measure the time saved for a user. Recorded when the password manager detects that the login was successful. hamelphi@chromium.org Metrics related to a Translate event. These metrics are described in TranslateEventProto. curranmax@chromium.org megjablon@chromium.org chrome-language@google.com Summary of the user's interaction with Translate from the beginning of the page load to when the event is recorded. This event is recorded when a page load is completed and every time that Chrome is backgrounded during the course of the page load. This means we can record multiple events for a single page load, but it guarantees that we will still record data in instances where the user backgrounds then kills Chrome. The source language at the time the event is recorded. In most cases this will match the initial source language, but the user can change the source language manually if they want. Note that the semantic_type attribute is included in order to remain consistent with the previous Translate UKM proto. At the time this event is recorded, the state of Translate. This includes whether the page is translated or not and if no UI is shown, just the omnibox icon is shown, or the full translate UI is shown. The target language at the time the event is recorded. In most cases this will match the initial target language, but the user can change the target language manually if they want. Note that the semantic_type attribute is included in order to remain consistent with the previous Translate UKM proto. The first error to occur within Translate for this page load. The user's first interaction with the Translate UI. The HTML language content attribute specified by the document. Note that the semantic_type attribute is included in order to remain consistent with the previous Translate UKM proto. The HTML language attribute specified by the document. Note that the semantic_type attribute is included in order to remain consistent with the previous Translate UKM proto. The initial source language that Translate determines for the page. Note that the semantic_type attribute is included in order to remain consistent with the previous Translate UKM proto. Whether or not the initial source language is in the user's never translate language list. By default the UI language, languages the user has translated to, and languages added to the content languages are on the never tranlsate list. However, users can add or remove languages manually as well. At the beginning of the page load, the state of Translate. This includes whether the page is translated or not and if no UI is shown, just the omnibox icon is shown, or the full translate UI is shown. The initial target language that Translate thinks it should translate to. Note that the semantic_type attribute is included in order to remain consistent with the previous Translate UKM proto. Across all translations during this page load, the maximum amount of time (in milliseconds) it took to translate the page. The language that was estimated by the language model based on the page's content. Note that the semantic_type attribute is included in order to remain consistent with the previous Translate UKM proto. The score, or confidence, of the language detection model's prediction about the language of the page content. The score is between 0 and 100, higher means more confident. The number of times a translation was reverted. The number of times that the target language was changed by the user over the course of this page load. The number of errors to occur within Translate for this page load. The number of times the page was translated. The number of times that the user interacts with the Translate UI. Decision of the Ranker whether to show the UI or not. Version of Ranker used for this page load. In case multiple events are logged for one page load, we track the sequence in the order each event is logged. The amount of time (in seconds) this page was in the foreground and not translated. The amount of time (in seconds) this page was in the foreground and translated. The highest priority trigger that determined the initial state of Translate for this page load. Whether the page content used to detect the page language was empty or not. peconn@chromium.org Collected when a page is opened in a Trusted Web Activity. A boolean signaling that the event has occurred (typically only records true values). ishitatsuyuki@google.com uekawa@chromium.org cros-virt-devices-guests@google.com Page load metrics for the first page load in a Trusted Web Activity (TWA). The startup is considered Cold if the browser process was not running and had to be started, and Warm if the process was already running in the background. Timings are relative to the Activity initialization timestamp passed from android-browser-helper. The version code of android-browser-helper used by the TWA wrapper associated with the origin. The timestamp of the first navigation commit, relative to the Activity initialization timestamp passed from android-browser-helper. Split by startups where the browser process was not in the background. The timestamp of the first navigation commit, relative to the Activity initialization timestamp passed from android-browser-helper. Split by startups where the browser process was already running in the background. The timestamp of the first contentful paint, relative to the Activity initialization timestamp passed from android-browser-helper. Split by startups where the browser process was not in the background. The timestamp of the first contentful paint, relative to the Activity initialization timestamp passed from android-browser-helper. Split by startups where the browser process was already running in the background. The timestamp of the largest contentful paint, relative to the Activity initialization timestamp passed from android-browser-helper. Split by startups where the browser process was not in the background. The timestamp of the largest contentful paint, relative to the Activity initialization timestamp passed from android-browser-helper. Split by startups where the browser process was already running in the background. Timestamp of performance.mark('mark_fully_loaded'), relative to the Activity initialization timestamp passed from android-browser-helper. Recorded only if the web page supports the respective performance marker. Split by startups where the browser process was not in the background. Timestamp of performance.mark('mark_fully_loaded'), relative to the Activity initialization timestamp passed from android-browser-helper. Recorded only if the web page supports the respective performance marker. Split by startups where the browser process was already running in the background. Timestamp of performance.mark('mark_fully_visible'), relative to the Activity initialization timestamp passed from android-browser-helper. Recorded only if the web page supports the respective performance marker. Split by startups where the browser process was not in the background. Timestamp of performance.mark('mark_fully_visible'), relative to the Activity initialization timestamp passed from android-browser-helper. Recorded only if the web page supports the respective performance marker. Split by startups where the browser process was already running in the background. Timestamp of performance.mark('mark_interactive'), relative to the Activity initialization timestamp passed from android-browser-helper. Recorded only if the web page supports the respective performance marker. Split by startups where the browser process was not in the background. Timestamp of performance.mark('mark_interactive'), relative to the Activity initialization timestamp passed from android-browser-helper. Recorded only if the web page supports the respective performance marker. Split by startups where the browser process was already running in the background. shaseley@chromium.org speed-metrics-dev@chromium.org Metrics associated with a unloading in preparation for a navigation. Measures the time in milliseconds of running the beforeunload handlers for a page. If the beforeunloads are dispatched from the browser, this reports the longest running one since these handlers will be run in parallel when the page has frames in different processes. If the navigation is renderer initiated, it will be the duration of running beforeunload handlers in the navigating frame and its local descendants in the renderer plus any latency caused by dispatching beforeunloads to remote frames from the browser. Measures the time in milliseconds from when the browser sends the request to run beforeunload to the renderer to when the renderer begins dispatching the handlers. For browser-initiated navigations, we report the queueing duration of the beforeunloads that took the longest since they are dispatched in parallel to remote and local frames. For renderer- initiated navigations this metric will only be filled in if we dispatched beforeunloads from the browser to remote frames which we had to wait on. Measures the time in milliseconds from when the browser sends the request to dispatch before unload handlers (for browser-intitated navigations) or from when the renderer started running before unload handlers (for renderer-initiated navigations) to when the commit navigation was sent to the renderer process. For cross-process navigations this represents the total amount of time the navigation is blocked since it was initiated. Measures the time in milliseconds from when the browser sends the request to dispatch before unload handlers (for browser-intitated navigations) or from when the renderer started running before unload handlers (for renderer-initiated navigations) to when the navigation was committed. For same-process navigations this represents the total amount of time the navigation is blocked since it was initiated. Measures the time in milliseconds to run the unload handlers in the old document where the navigation was committed. Only records for same-process navigations since there are no unload handlers to be run in the new process during a cross-process navigation. Measures the time in milliseconds from when the browser sends the commit navigation request to the renderer to when the renderer dispatches unload events. Only records for same-process navigations since unload handlers are run in a different process for cross-process navigations. jiameng@chromium.org Collects user activity events after staying idle for a short period of time. Percentage of battery, bucketed to 5 percents, i.e. [0, 5%) is mapped to 0, [5%, 10%) is mapped to 5 etc. An enum representing whether the device is managed, defined in |ash::power::ml::UserActivityEvent::Features::ManagementType|. An enum representing the mode of the device, defined in |ash::power::ml::UserActivityEvent::Features::DeviceMode| An enum representing the type of the device, defined in |ash::power::ml::UserActivityEvent::Features::DeviceType| Time taken in seconds from when we start logging the features to the moment when the event type is known and we finish logging the complete event. An enum representing the reason of the event, defined in |ash::power::ml::UserActivityEvent::Event::Reason|. An enum representing the type of the event, defined in |ash::power::ml::UserActivityEvent::Event::Type|. The number of key events reported as user activity in the last hour. This is bucketized so that it is recorded precisely when it is in [0, 99], rounded to nearest 100 when it is in [100, 999] and rounded to nearest 1000 when it is in [1000, 9999]. It is capped at 10000. An enum representing the last activity day of the week, defined in |ash::power::ml::UserActivityEvent::Features::DayOfWeek|. Last activity time as hours since midnight in the local time zone. Last user activity time as hours since midnight in the local time zone. A boolean value whether model prediction is applied in the screen-dim decision. False means powerd dims the screen regardless of the model prediction. A threhold in the range of [0, 100]. Higher threshold makes screen-dim less likely. Probability that user will remain inactive as predicted by the model. It is in the range of [0, 100]. If score is less than the decision threshold, dim will be deferred. Whether the model recommends the screen dim should be deferred (NO_DIM) or allowed (DIM), or encountered an error (MODEL_ERROR). The number of mouse events reported as user activity in the last hour. This is bucketized so that it is recorded precisely when it is in [0, 99], rounded to nearest 100 when it is in [100, 999] and rounded to nearest 1000 when it is in [1000, 9999]. It is capped at 10000. Boolean value to represent whether the device is currently on battery power. Number of times user reactivated, not including the current event. Number of times user has remained idle, not including the current event. Duration of activity in seconds up to last activity. Recent video playing time in seconds. Delay after which the screen will be dimmed in seconds. Whether screen was dimmed at the start of the event. Whether screen dim occurred due to inactivity timeout. Delay after which the screen will be turned off in seconds, as measured from screen is dimmed. If dim is disabled, it is the delay from the start of inactivity. Whether screen was locked at the start of the event. Whether screen lock occurred. Whether screen was off at the start of the event. Whether screen off occurred due to inactivity timeout. The ID of this event in the current session. Time from last key event in seconds. Time from last mouse event in seconds. Time from last touch event in seconds. Time from the end of the last video playing in seconds. The number of touch events reported as user activity in the last hour. This is bucketized so that it is recorded precisely when it is in [0, 99], rounded to nearest 100 when it is in [100, 999] and rounded to nearest 1000 when it is in [1000, 9999]. It is capped at 10000. jiameng@chromium.org Records an activity that happened while a page was open. The ID of the activity that happened while a page was open. Enumeration for the MIME type of the page. Values are enumerated in metrics::TabMetricsEvent::ContentType. Boolean value indicating whether the page has any user-input text. Boolean value representing whether this tab is selected in its containing browser. Boolean value representing whether the containing browser is in focus. Boolean value representing whether the containing browser is visible. Boolean value representing whether the containing browser is the topmost one on the screen. Site engagement score in the range [0, 100], rounded down to a multiple of 10 to limit granularity. djw@chromium.org Metrics related to a page visit as a user perceives it, regardless of how the page was visited. More metrics may be associated with a particular page visit, these can be found in specific UKM events with the same UKM source ID. Currently the methods of visiting a page that will populate this event are: 'normal' page load (more metrics in the PageLoad event), Back-Forward Cache Restore (more metrics in the HistoryNavigation event), and AMP documents (more metrics in the AmpPageLoad event) Determines if this page visit event should be counted towards calculations involving Core Web Vitals. If true, this page visit event should not be counted. If false or NULL, the page visit event should be considered. Currently this is determined by checking the mime-type of the page visit event. Determines the final status for this page visit. That is, records whether the page was never foregrounded, was foregrounded but did not reach FCP, or did reach FCP. The metric is recorded at the end of each page visit. App backgrounding on Android to is considered to be the end of page visit as well, as the app could be evicted by Android anytime after that happens. True if this page visit was initiated by user input. For renderer navigations, only true if the code can trace back to the original input event - there are cases where this known not to be possible. etiennep@chromium.org Records information on memory and timing of a full GC cycle, reported at the end of the cycle. Main thread only duration of the whole garbage collection of the unified heap. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Main thread duration of the atomic pause of the garbage collection of the unified heap. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Main thread duration of the atomic pause of the garbage collection of the managed C++ heap. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Main thread only duration of a whole garbage collection of the managed C++ heap. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Time between the start of the full garbage collection and the end of previous full cycle. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Overall duration of a whole garbage collection of the unified heap. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Overall duration of a whole garbage collection of the managed C++ heap. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Amount of global consumed memory right after the previous full garbage collection cycle, used as baseline for limit computation. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Amount of global consumed memory right before the current full garbage collection cycle. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Delta between the current and baseline amount of global consumed memory for the current full garbage collection cycle. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Delta between the limit and baseline amount of global consumed memory for the current full garbage collection cycle. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Allocation limit on golbal consumed memory used before the current full garbage collection cycle. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Heap Growing Mode at the start of the current cycle, that affects heap limits computation. Indicates whether the cycle started while input handling state was true. Indicates whether the cylce started while loading state was true. Amount of allocated memory freed by the garbage collection of the managed C++ heap. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Amount of allocated memory freed by the garbage collection of the managed C++ heap. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Amount of used memory after the garbage collection of the unified heap. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Amount of used memory after the garbage collection of the managed C++ heap. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Amount of used memory before the full garbage collection of the unified heap. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Amount of used memory before the full garbage collection of the managed C++ heap. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Amount of used memory freed by the garbage collection of the unified heap. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Amount of used memory freed by the garbage collection of the managed C++ heap. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Amount of old generation consumed memory right after the previous full garbage collection cycle, used as baseline for limit computation. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Amount of old generation consumed memory right before the current full garbage collection cycle. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Delta between the current and baseline amount of old generation consumed memory for the current full garbage collection cycle. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Delta between the limit and baseline amount of old generation consumed memory for the current full garbage collection cycle. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). Allocation limit on old generation consumed memory used before the current full garbage collection cycle. This is rounded down to the nearest exponential bucket (with a bucket ratio of 1.3). The priority the current full garbage collection cycle was running at. Reason the current full garbage collection cycle was started. Reason incremental marking was started before the current full garbage collection cycle if any. Indictaes whether ReduceMemoryFootprint flag was set. ecmziegler@chromium.org Records information on baseline compilation of WebAssembly modules and its resource consumption. The compilation was running asynchronously. The compilation was re-using already cached native module code instead of JIT compiling it. The size of the native module code in bytes after baseline compilation finished. The compilation was deserializing the native module code from the disk cache instead of JIT compiling it. The lazy compilation flag was enabled and therefore not all functions might have been compiled. The number of functions that Liftoff could not compile and that therefore needed to be compiled by TurboFan. The wire bytes were streamed while compiling. The module was successfully compiled. The total time the compilation took in microseconds. ecmziegler@chromium.org Records information on WebAssembly modules included in websites. Number of declared functions in this module. The size of the wire bytes of the module. The wire bytes were streamed while decoding. The module was successfully decoded. The total time the decoding took in microseconds. ecmziegler@chromium.org Records information on instantiation of WebAssembly modules and its resource consumption. The number of functions imported by the module. The module was successfully instantiated. The total time the instantiation took in microseconds. ecmziegler@chromium.org Records information on top-tier compilation of WebAssembly modules and its resource consumption. The size of the native module code in bytes after top-tier compilation finished. The lazy compilation flag was enabled and therefore not all functions might have been compiled. The total time the compilation took in microseconds. marja@chromium.org v8-loading@google.com Data about which JavaScript functions are compiled. The collected data is obfuscated to not reveal script URLs. The data is split into separate int64s. sanjeetnd@google.com charleszhao@google.com Records videoconferencing usage on web apps. Only records for apps which have captured camera, microphone, or screen at least once while open. The total time duration (bucketed) the VC app spent capturing camera. Duration in units of ms. App captured camera. App captured microphone. App captured screen. The total time duration (bucketed) the VC app spent capturing microphone. Duration in units of ms. The total time duration (bucketed) the VC app spent capturing the screen. Duration in units of ms. Duration (bucketed) for which the VC app was open. Duration in units of ms. shend@chromium.org Collects the used input types with the Virtual Keyboard. Type of a input field which the user is trying to use the Virtual Keyboard on. hanfeng@google.com Records boarding pass barcode detection results. Whether a boarding pass barocode is detected. yfriedman@chromium.org hartmanng@chromium.org pkotwicz@chromium.org WebAPK usage metrics recorded on Install. Records the display mode of the added WebAPK. Distribution source of the WebAPK (browser, device policy, etc). Recorded when this WebAPK is installed. From where the user initiates the install: app menu, install API, etc. yfriedman@chromium.org hartmanng@chromium.org pkotwicz@chromium.org WebAPK usage metrics recorded when the WebAPK is backgrounded. WebAPK app version that the user has installed. Distribution source of the WebAPK (browser, device policy, etc). How long the user spent inside the WebAPK, in milliseconds. This will use exponential bucketing (ie, count in 10s of milliseconds at first, then 100s, then seconds, 10s of seconds, minutes, 10 minutes, hours, 10 hours, days, weeks, years.) yfriedman@chromium.org hartmanng@chromium.org pkotwicz@chromium.org WebAPK usage metrics recorded on Uninstall. WebAPK app version that the user has installed. Distribution source of the WebAPK (browser, device policy, etc). How long between Install and Uninstall of a particular WebAPK, in seconds. This will use exponential bucketing (ie, count in seconds at first, then 10s of seconds, minutes, 10 minutes, hours, 10 hours, days, weeks, years.) Records how many times the WebAPK was launched between install and uninstall (in coarse buckets). Recorded when this WebAPK is uninstalled. yfriedman@chromium.org hartmanng@chromium.org pkotwicz@chromium.org WebAPK usage metrics recorded during a user's interaction with the site. WebAPK app version that the user has installed. Distribution source of the WebAPK (browser, device policy, etc). Recorded every time this WebAPK is launched. Launch source of the WebAPK. eirage@chromium.org> hartmanng@chromium.org mwi-team-core@google.com A2HS usage metrics recorded when users started the add to home screen flow on Android. Records the display mode of the added Webapp. Record whether `start_url` for the Shortcuts is the same as the current page url. Record an StatusCode for why this site is not WebAPK compatible. dmurph@chromium.org pwa-team@google.com Features of a web app recorded at most once per day if a user interacted with the web app. Recorded for installed and/or promotable (installable) web apps unless noted otherwise on the metric. Keyed by start_url of the app. Note: Changed in Feb 2023 to only record when App sync is enabled. Records the total duration (seconds, bucketed) that the app was open in the background in the last calendar day. Summed during the day and emitted if the date changes. Recorded only for installed web apps that are currently open in a web app window, and all preinstalled web apps. In m128 a bug is fixed where on resuming from sleep background time did not resume counting. Whether the user has enabled link capturing to open in-scope links in the app window. Recorded only for installed web apps that are currently open in a web app window, and all preinstalled web apps. Effective display mode of the web app (may be set by a manifest and manifest may be overridden by the user). Records the total duration (seconds, bucketed) that the app was open in the foreground in the last calendar day. Summed during the day and emitted if the date changes. Recorded only for installed web apps that are currently open in a web app window, and all preinstalled web apps. As of m128, this time is cut short if Chrome determines that it is not active anymore (see the DesktopSessionDurationTracker). This is to keep the cumulative sum of this value comparable with Session.TotalDuration. Whether the web app is installed. Install source of the web app (omnibox, menu item, policy, etc). Records the total number of times the app was focused in the foreground in the last calendar day. Recorded only for installed web apps that are currently open in a web app window, and all preinstalled web apps. Whether the web app is currently considered 'promotable' (passes Chrome checks to be considered an installable web app). Records that the web app was used. Always true (or the event would not have been emitted for this WebApp). dmurph@chromium.org pwa-team@google.com Recorded when a site attempts to install another site that isn't the current document. There are two possible entry points to the install - The Web Install API (navigator.install) and the install element. For any given install, either the first *or* second metric pairs will be recorded. Records the result of one activation of the install element for the URL of the target site that we attempted to install. As of M153, keyed on the origin of the caller-supplied manifest URL (earlier milestones used the install URL). Only recorded when the manifest URL is HTTPS. Records the result of one activation of the install element for the URL of the requesting page that the element was activated from. Records the result of the Web Install API for the URL of the target site that we attempted to install as a web app via the API. As of M153, keyed on the origin of the caller-supplied manifest URL (earlier milestones used the install URL). Only recorded when the manifest URL is HTTPS. Records the result of the Web Install API for the URL of the requesting page that invoked the API. kenrb@chromium.org chrome-webauthn@google.com Information about a WebAuthn Get Assertion API call with conditional mediation enabled. The amount of time (ms) that has elapsed between a page's DOMContentLoaded event and the API call being handled. This is only recorded for the first conditional request on the page. kenrb@chromium.org chrome-webauthn@google.com Information about the completion of a request to register a new Web Authentication credential. This was added for desktop platforms in M128 and for Android in M131. The outcome of the request, indicating whether it was successful, and the cause of failure if not successful. The mode of WebAuthn register request that has been completed. kenrb@chromium.org chrome-webauthn@google.com Information about the completion of a request for a signed assertion by a Web Authentication credential. This was added for desktop platforms in M128 and for Android in M131. The mode of WebAuthn assertion request that has been completed. The outcome of the request, indicating whether it was successful, and the cause of failure if not successful. yigu@chromium.org goto@chromium.org web-identity-eng@google.com Records in the browser process when a document is destroyed. Records whether a document collected phone number, and/or used WebOTP, and/or used OneTimeCode (OTC) during its lifecycle. solenberg@chromium.org webrtc-dev@chromium.org Collects usage patterns that may indicate attempts to gather users' IP addresses. A heuristic in the webrtc library tries to detect usages that may indicate attempts to gather users' IP addresses rather than setting up communication via WebRTC, and reports possible attempts. The usage pattern detected (an int encoding bit values). The bits are defined in third_party/webrtc/pc/peerconnection.h, in the "PeerConnection::UsageEvent" enum. hbos@chromium.org guidou@chromium.org Logged when an application uses complex SDP in a WebRTC PeerConnection or experiences an error attempting to parse SDP. Such cases will result in errors if the application assumes the wrong default SDP format. This is an important concern during the transition of the default SDP format from Plan B to Unified Plan. A value that represents how complex/unparsable SDP is used. The categories are defined in blink::ComplexSdpCategory. evliu@chromium.org chrome-media-ux@google.com Recorded when a page navigation invokes either SpeechRecognition or SpeechSynthesis. Used to measure website adoption distribution and speech recognition backend selection. The explicit speech recognition quality level requested by the site (Command=0, Dictation=1, Conversation=2). The speech recognition backend selected/used for the session (Cloud=0, SODA=1, SmallExpertModel=2, GeminiNano=3). Logged as 1 if SpeechRecognition was initiated during the navigation. Logged as 1 if SpeechSynthesis speak() was invoked during the navigation. arichiv@google.com katabolism-finch@google.com This metric is recorded when WindowProxy usage occurs (e.g., a frame accesses its opener, parent(s), child(ren), or a window it opened). Records the method being used on the window proxy. Boolean value that indicates if the frame accessing the window proxy and the frame referenced by the window proxy are in the same page. Records the context of the frame tree accessing the window proxy. Records the context of the page accessing the window proxy. Records the user activation state of the frame accessing the window proxy. Records the context of the frame tree referenced by the window proxy. Records the context of the page referenced by the window proxy. Records the user activation state of the frame referenced by the window proxy. Records the similarities and differences between the storage key of the frame accessing the window proxy and the frame referenced by the window proxy. billorr@chromium.org xr-dev@chromium.org Records information about WebXR methods that were used in a single navigation. The event will be recorded as the WebXR functions are called - since the event is singular, the data that will eventualy be sent will contain a union of metrics that were set in an event. Boolean value that indicates that the WebXR input API was used. Boolean value that indicates that the API to enumerate devices was called. This API is only exposed through WebVR. Boolean value that indicates that poses were requested. Boolean value that indicates that presentation was requested. Boolean value that indicates that the WebXR Device API was used. Boolean value that indicates that a device was returned by the API to enumerate devices. This API is only exposed through WebVR. Boolean value that indicates that a device was returned by the API to enumerate devices, and that the device supports presentation. alcooper@chromium.org xr-dev@chromium.org When session ends, records data for a WebXR / WebVR session. The approximate amount of time (in seconds) that elapsed from session creation to session end. Times are reported accurately when low, for example, under a minute, and get rounded to minutes and then hours as they get larger. This is done by SessionTracker::GetRoundedDurationInSeconds. For inline sessions, this might get pretty large. State of the 'bounded-floor' feature request. State of the 'local' feature request. State of the 'local-floor' feature request. State of the 'unbounded' feature request. State of the 'viewer' feature request. Whether the 'bounded-floor' feature was used during this session, regardless of whether it was used successfully. Whether the 'local' feature was used during this session, regardless of whether it was used successfully. Whether the 'local-floor' feature was used during this session, regardless of whether it was used successfully. Whether the 'unbounded' feature was used during this session, regardless of whether it was used successfully. Whether the 'viewer' feature was used during this session, regardless of whether it was used successfully. An enum value specifying WebXR session's mode. See `device::SessionMode` for enum type declaration. 0: Unknown, catch all. 1: Inline session. 2: Immersive VR session. 3: Immersive AR session. xr-dev@chromium.org cassew@google.com Recorded when promise returned by `requestSession` gets resolved / rejected. State of the 'bounded-floor' feature request. State of the 'local' feature request. State of the 'local-floor' feature request. State of the 'unbounded' feature request. State of the 'viewer' feature request. An enum value that indicates that the `requestSession` API was called to request session with this mode. See `device::SessionMode` for enum type declaration. 0: Unknown, catch all. 1: Inline session was requested. 2: Immersive VR session was requested. 3: Immersive AR session was requested. An enum value that describes the result of the call to `requestSession` method. See `` for enum type declaration. 0: Success. 1: Other failure.