Page MenuHomePhabricator

Make event statistics specific to a wiki
Closed, ResolvedPublic5 Estimated Story Points

Description

When an event takes place over multiple wikis, we want statistics broken down by each wiki, along with totals: https://meta.wikimedia.org/wiki/Community_Tech/Grant_metrics_tool#/media/File:New_Grant_metrics_3_w_breakout.png

Currently we are only calculating statistics across all wikis that the event belongs to.

Event Timeline

@Niharika With T182885 I've realized the EventStat model is set up only to contain statistics for the event as a whole, and not individual wikis. This means we cannot break down the stats by wiki as we do in the mock :(

Not easy work... but I'm assuming we definitely want this feature, in which case I'll get started on it now. Shall I proceed?

MusikAnimal moved this task from Ready to In Development on the Community-Tech-Sprint board.
MusikAnimal set the point value for this task to 5.
MusikAnimal moved this task from Backlog to In progress on the Grant-Metrics board.

We discussed this on IRC a bit.
@MusikAnimal One thing I realized just now is that the way retention has been defined is across all wikis. So we can't have different retention metrics across wikis. We should compute retention as we are doing currently.

Similarly "new editors" cannot change because a user can register on any wiki.

Pages created and pages improved are the only ones that will be affected by this.

MusikAnimal changed the point value for this task from 5 to 8.Feb 5 2018, 10:01 PM

Hmm okay. I guess we need a new model entirely, EventWikiStat, in addition to EventStat. This has turned into an 8 pointer!

I was thinking this might be easier to do now that we don't have to calculate separate stats for everything, just for pages created and edited. 🤔

MusikAnimal changed the point value for this task from 8 to 5.Feb 6 2018, 3:08 AM

I think this does it: https://github.com/wikimedia/grantmetrics/commit/4e80db4b03ed69baaabdb1c2272d4798e1d8c61d Also changing back to 5 points, because that was definitely more accurate.