Page MenuHomePhabricator

Record monotonic time of survey impression
Closed, ResolvedPublic


Currently the QuickSurveyInitiation table records the survey impression. By virtue of the event capsule, we know what time this happened, but only to the second.

With minimal effort, it should be possible to record a precise monotonic time, such as, to get an idea of how long it took for the survey to appear.

In the context of the performance survey, this will allow us to see if there is a correlation between negative responses to "did this page load fast enough?" and how long it took for users to see the survey (either because they had to scroll to it, or because it took a while to load). Which would allow us to prove indirectly whether or not the survey's late appearance influences responses to it.

Event Timeline

Gilles triaged this task as Medium priority.Jun 8 2018, 7:54 PM
Gilles created this task.

Change 439304 had a related patch set uploaded (by Gilles; owner: Gilles):
[mediawiki/extensions/QuickSurveys@master] Record on survey impressions

Change 439304 merged by jenkins-bot:
[mediawiki/extensions/QuickSurveys@master] Record on survey impressions

Krinkle subscribed.

@Gilles Does that patch resolve this task, or is there more?

I also need to write the part that will collect it and send it via EventLogging

Nevermind, that's already the case for this one. Just let me check that the data is correctly recorded.

I don't see the new EventLogging table yet, I presume this hasn't hit the wikis where the survey is running yet.

Indeed, with this week's deployment freeze those wikis are still on wmf.8

Vvjjkkii renamed this task from Record monotonic time of survey impression to cdbaaaaaaa.Jul 1 2018, 1:04 AM
Vvjjkkii removed Gilles as the assignee of this task.
Vvjjkkii raised the priority of this task from Medium to High.
Vvjjkkii updated the task description. (Show Details)
Vvjjkkii removed subscribers: gerritbot, Aklapper.

The data is making it to EventLogging as expected and I'm seeing large values as I thought was possible when the survey is initially rendered below the fold. This will be a great signal to eliminate filter out that case or factor it into the machine learning.