Last modified: 2013-09-17 19:52:07 UTC
A) The title of the pageviews report card [1] says "Pageviews for All Wikimedia Projects", contradicting the legend on the left which says they are for Wikipedias only, e.g. (for Jun 2013): " All Wikipedias (+Mobile) 21.18B English 9.61B Japanese 1.58B Spanish 1.54B German 1.13B Russian 1.39B French 835.86M " B) Also, the "all" numbers contradict the data on stats.wikimedia.org. E.g. [2] gives 20,223M for "Wikipedia total" (mobile+desktop) for June 2013, not 21.18B. (On the other hand, the report card numbers for the language WPs seem to match [3], e.g. 9,611M for enwiki in June 2013.) C) Even if we assume that the "All Wikipedia" legend in the report card was meant to read "All Wikimedia projects", it still contradicts [2], which gives 21,138M instead of 21.18B on the report card. [2] is the source for the "page requests" number routinely included in the monthly WMF report [4]. One possible cause for problem C) might be that the report card includes Wikidata whereas stats.wikimedia.org does not (cf. https://bugzilla.wikimedia.org/show_bug.cgi?id=46380 ), but in that case the discrepancy should really be clarified in the report card description. (just speculating here) -- [1] http://reportcard.wmflabs.org/graphs/pageviews [2] http://stats.wikimedia.org/EN/TablesPageViewsMonthlyAllProjects.htm [3] http://stats.wikimedia.org/EN/TablesPageViewsMonthlyCombined.htm [4] https://blog.wikimedia.org/2013/07/18/wikimedia-foundation-report-june-2013/#Data_and_Trends
A+B) Input for Limn is wikilytics_in_pageviews.csv Section '=== Page view totals per project - non-mobile + mobile ===' lists for June 2013: 21175211793 for 'Total' (total is unspecified, but clearly context implies 'All Wikimedia Projects') This matches header in Limn report [1]: 21.18 Billion So the Limn legend label is in error: it should either * repeat the section header or * present the counts for Wikipedia wikis only: June 2013: 20.2 Billion, in which case a separate 'Total' value could be added. C) There are two discrepancies between [2] and [4] The Report Card total includes two entities, which are not included in the Page View report: * WikiData (for June 2013 that is 26.1 million) * Main Wikipedia Portal www.wikipedia.org (for June 2013 that is 10.4 million) A clarification could be as follows: "Unlike the corresponding report with <a href='http://stats.wikimedia.org/EN/TablesPageViewsMonthlyAllProjects.htm'>page views by project</a> here the following two sites are also counted: <a href='www.wikidata.org'>WikiData</a> and <a href='www.wikipedia.org'>the Wikipedia Main Portal page</a>. Before anyone asks: adding both sites to the other report as separate columns can of course be done. But the underlying code is one of the least maintenance friendly sections in Wikistats. All projects are totally separated except for this one report and instead of major restructuring of reporting code a much faster shortcut was taken) So would that be worth the effort? I doubt it.
Sounds to me like if we changed the title of the graph and the "All Wikipedias (+Mobile)" label in the legend, we could fix the misunderstanding. Would a description be required then? I'm happy to add one, just let me know what you think the best wording is. So we need three things: 1. New Title for the pageviews graph [1] 2. New Label for the "All Wikipedias (+Mobile)" metric 3. (Optional) Description to add under the graph Thanks for spotting this Tilman
Thanks for explaining this! IMHO 1., 2. and 3. (with the wording suggested by Erik) would resolve the problem. Of course it would still be better to have a consistent usage/definition of "all [Wikimedia] projects", but at least the added remark would help to avoid the impression that something is wrong with the numbers. BTW, based on your clarification, I will probably switch the page request numbers in the WMF monthly report ([4]) to the report card version [1], because it is closer to how I understand the scope and reader expectations of the monthly report.
1. I'm leaving the title of the graph as is: "Pageviews for All Wikimedia Projects" 2. I'm changing the label to "All Wikimedia Projects" 3. I've added the explanatory note about wikidata and the portal page I'm deploying now, let me know if you'd like anything else different.
Scheduling and prioritization of this bug is handled on Mingle card https://mingle.corp.wikimedia.org/projects/analytics/cards/1097
The new description in the footer at [1] is great, but the legend in the left sidebar for the total number still reads "All Wikipedias (+Mobile)". Once that is changed to "All Wikimedia Projects", I think we can mark this bug as resolved.
done, thanks for pointing out (I changed the label on the datasource accidentally).
Cool, thanks! Another thing (related and we should have thought of it during the above discussion, so probably not worth opening a new bug): After the change, the terms "English", "Japanese", "Spanish" etc. in the sidebar legend are now ambiguous, and very likely to be misunderstood as meaning "all English-language Wikimedia projects", etc. Can we change them to "English Wikipedia", "Japanese Wikipedia", "Spanish Wikipedia" etc.?
ok, done, but now we should probably let this bug go to sleep :)
Thanks, and good night, Bug 53238!
