Page MenuHomePhabricator

Wikistats Pageviews differ hugely from pageviews.wmcloud ( 0.1x — 1,000x )
Open, Needs TriagePublic

Description

Exs:
Bethany_Lutheran_Church_in_Oilmont,_Montana.jpg : 5.5k vs 32k
Gtk-dialog-info-14px.png : 10M vs 12M
Email Silk.svg: 370k vs 30k
Logo for e-Science Photo Competition v2 without text.svg : 182M (!!) vs 2.5M
Mona Lisa, by Leonardo da Vinci, from C2RMF retouched.jpg : 1.9M vs 34k
Flag of Austria.svg : 22M vs 30k

There's also a huge jump in Wikistats Pageviews for media (at least for Commons) from ~4k monthly views to ~30k monthly views, which makes no sense statistically

Event Timeline

Restricted Application added subscribers: MusikAnimal, Aklapper. · View Herald Transcript

@Sj Could you provide a link to the corresponding views on Wikistats?

First off, you state "pageviews" differ hugely, but what you're linking on pageviews.wmcloud is Mediaviews, which is not the same metric as pageviews (see Pageviews vs Mediarequests).

I'm not sure how to check mediarequests of individual files on stats.wikimedia.org. I managed to find the "Top files by mediarequests" (there's an apparent bug in "Explore" dropdown; I found the link by checking the "All Wikistats Metrics" page). Looking at that list, I see File:Icon_pdf_file.png which has 302.76 million mediarequests for November 2023, while Mediaviews reports 302.88 million mediarequests. The other titles have similar variations. They are still different, so something is not adding up, but they're at least very similar.

API endpoint used by wikistats: https://wikimedia.org/api/rest_v1/metrics/mediarequests/top/all-referers/all-media-types/2023/11/all-days

API endpoint used by Mediaviews (because it shows daily granularity): https://wikimedia.org/api/rest_v1/metrics/mediarequests/per-file/all-referers/all-agents/%2Fwikipedia%2Fcommons%2F4%2F4d%2FIcon_pdf_file.png/daily/2023110100/2023113000

Mediaviews is supposed to simply sum the mediarequests for each day as reported by the API, so in theory it should be identical to the /top endpoint used by Wikistats. I'll look into this more when I have time, as I'm not sure whether to blame Mediaviews or the backing API.

Aha, thank you ! Yes, that's the difference, I was looking at pageviews, not mediaviews, on Wikistats. Interesting to see the gaps on the two ends of the spectrum between pageviews and mediaviews, and that it's not always in the same direction.

So this isn't a bug (aside from the minor variance, which could be differences in how the "last month" is defined), more a mystery