@MusikAnimal, a question about the Pageviews, cumulative metric above:

This is a per-page breakout version of the Views to pages improved metric we defined for the the Event Summary reports (T205561, T206692). There, we said this metric would count pageviews “from creation until last data update.” But since these metrics are meant to provide some measure of impact, that really makes no sense for Pages Improved, since all the pageviews to the article prior to the event are irrelevant.

So how can make this metric more relevant? E.g., can we:

• Count from the last edit made during the the event? Or
• Count from the end of the event?

I like the first of those a little better, since the event may be ongoing at the time of the report. But either would be fine. Are they hard to do? Is one harder than the other??

Ready for review: https://github.com/wikimedia/eventmetrics/pull/250

Hmm, I found an inconsistency in the Avg. Daily Pageview numbers in the Pages Improved vs. Event Summary reports. To reproduce:

• Download the Pages Improved csv for the event "At Least One of All Contribution Types" (In the Program joe's testing events).
• Add up the Avg. Daily Pageviews number, only for en.wiki pages. As of 2019-03-28 18:13, that comes to 13,515.
• Compare to the onscreen per-wiki breakdown for the same figure. The total is 12,937.

Is that a bug? I could understand if adding up all the individual page daily averages gives a somewhat different number from whatever method we use to calculate the aggregate daily average. But we're only talking about 12 pages, so a variance of almost 600 pages seems like a lot.

@MaxSem, @dom_walden you did what do you think?

(Note: I checked Bytes and Edits in the same way, and they were both internally consistent.)

@dom_walden, I proofed the Wikitext table against the spec and can confirm that the language and formatting are as specified. That leaves you to concentrate on checking the accuracy.

I also checked to make sure that the number of Pages Improved and the number of Wikidata items reported here are consistent with the numbers reported in the Summary report. (We've decided to include the Wikidata items in this report, since we dropped the originally planned Wikidata report.)

In T210898#5068366, @jmatazzoni wrote:

@MaxSem, @dom_walden you did what do you think?

It seems to consistently report higher averages then those on https://tools.wmflabs.org/pageviews/, on all the wikis, on both the csv and wikitext report. I haven't been able to work out what is going on.

Most of the testing of the accuracy of the figures has been done in the CSV report in T210775.

I took one event and compared the figures, line by line, between CSV and Wikitext. They matched.