Issue metadata
Sign in to add a comment
|
interstitial.malware.decision is polluted with extra SHOW counts |
||||||||||||||||||||||||
Issue descriptionThere are a handful of clients that increment the SHOW bucket on malware interstials without the PROCEED or DONT_PROCEED buckets, and it messes up our CTR metrics. This affected Windows first (all channels) starting 2017-03-08, and then all desktop client on 2017-04-02 https://uma.googleplex.com/timeline_v2?sid=0dd28f5d5fac0b8413f693aeca692998 Conversations about the spike on Windows: uma-users: https://groups.google.com/a/google.com/d/msg/uma-users/En54EU0CGSk/iEaof98tBQAJ ops: https://groups.google.com/a/google.com/d/msg/chrome-safebrowsing-team/M_NfQ-38qpI/v6OkSxQcEAAJ Dremel suggests this is getting worse, now with 187 clients reporting more SHOW-(PROCEED+DONT)>20 in one day. The max discrepancy is 1700, meaning 1700 interstitials somehow without closing the interstitial.
,
Apr 7 2017
AI: Change the crsb dashboard to compute CTR as proceed/(proceed+dont_proceed) instead of proceed/shown
,
Apr 11 2017
CRSB dashboard updated
,
May 8 2017
Looks like a similar thing is now happening on Android for interstitial.phishing.decision: https://uma.googleplex.com/timeline_v2?sid=520f6a725321f92f4a270d65c74e0e1c On Android, "SHOW" was hovering around 52% but recently stepped up to 57%.
,
Jun 12 2017
I'm starting a thread with finch-users to discuss this since it is affecting the PVer4 launch metrics.
,
Jun 26 2017
There's a recent spike in interstitial.malware.decision = {SHOW}
https://uma.googleplex.com/timeline_v2?sid=d3a769023771d0301301bfad91d8eae0
,
Jul 31 2017
FWIW: I am only seeing a mismatch for interstitial.malware.decision, not for phishing See: https://uma.googleplex.com/p/chrome/variations?sid=efb7b6e26720ccdaa23aa34d934cb9e6
,
Nov 10 2017
,
Feb 18 2018
,
Oct 1
|
|||||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||||
Comment 1 by nparker@chromium.org
, Apr 5 2017