As pointed out in T186572: uBlock blocks EventLogging, there are adblockers which prevent events from being sent to our server. What proportion of our visitors are Eventlogging-invisible? Can we find this by simply comparing web logs with event logs? Is the proportion similar for anonymous vs. logged-in users?
Description
Description
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Open | BUG REPORT | None | T256463 QuickSurveys should show an error when response is blocked | ||
| Resolved | awight | T256464 What proportion of our visitors have blocked EventLogging? |
Event Timeline
Comment Actions
@nettrom_WMF may remember the answer to this (for logged-in users anyway), because we looked at this not that long ago with regards to GrowthExperiments-Homepage where we have a server-side event that fires alongside several client-side events. I want to say the number is like 20% on desktop and 10% on mobile, but that could be way off.
Comment Actions
This was answered precisely by @AndyRussG in T236834#5696044: When comparing ad-hoc sendBeacon calls to the same event as recorded by EventLogging, there was an 11% drop. Whatever the causes, adblocking is the most likely, and the implication for EventLogging producers is that we should expect to lose the same proportion of data even though JS is working in general and we may have already rendered something we intended to count, for example.
Comment Actions
Thanks for the ping @kostajh! I'm going to throw a bit of nuance on the number that @AndyRussG reports based on the findings for Growth that I reported in T243632#5898720.
Note that my analysis looked at logged-in users because the Homepage isn't available to non-registered users. On a per-user basis, 10–11% of them blocking everything consistently is what we'd expect to see, that's found in both analyses. In the Homepage analysis, we also found an additional 15–20% block to a varying degree, so we have some data for them but not consistently.
If we switch to looking at this on a per-session basis, the proportion increases. I interpret that to mean that more active users are more likely to block. Desktop sessions were more likely blocked compared to mobile sessions, except on Vietnamese Wikipedia where it was opposite.
In summary: how client-side blocking will affect QuickSurvey would depend on the sampling strategy used for it.