Hi all,
I added two pages to Google Doc " Analysis of pageviews trend after correction for overreporting" scroll to bottom https://docs.google.com/a/wikimedia.org/document/d/1kpJrfataS5KAxGXFoygQVhMl...
One updated chart, plus two new charts,and some raw data.
Quick first observations: (may be refined in coming days) The page view reports http://stats.wikimedia.org/EN/TablesPageViewsSitemap.htm are based on webstatscollector data, which are totally based on GET /wiki/ (as was known). They miss out on GET /w/index.php (a considerable amount of page views, of all kind, including search) and GET /w/api.php (not so much). ‘Get other’ is not plotted in the chart. It is partly portal pages, e.g. http://wikipedia./org , partly invalid requests. (I may need to break this up further)
So with above limitations squid based counts 'GET /wiki/' are roughly same order of magnitude as webstatscollector data, but curves do not exactly match.
More importantly data from squid log show no consistent decline over several months. We also see no decline in share of traffic from Google.
All in all it seems we need to look for internal data collecting/reporting bug as explanation. Or review the patch made in December (for filtering bogus traffic) once again, but it was thoroughly and independently vetted by two people already, not a likely cause.
Any further comments tomorrow
Erik
-----Original Message----- From: analytics-bounces@lists.wikimedia.org [mailto:analytics-bounces@lists.wikimedia.org] On Behalf Of Daniel Schwen Sent: Tuesday, January 14, 2014 0:36 To: A mailing list for the Analytics Team at WMF and everybody who has an interest in Wikipedia and analytics. Subject: Re: [Analytics] Page view data with Wikipedia app?
So actually /w/api.php makes about 20% of the requests compared to /wiki/? (a bit confused, sorry)
That does not seem too surprising, as it includes autocomplete as you type requests in the search widget, which must make a large contribution by sheer request number. Daniel
_______________________________________________ Analytics mailing list Analytics@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/analytics