Last modified: 2014-03-11 13:42:08 UTC
Fluctuation in monthly forecast is (too?) high, reported by Goan in 2011 possible cause: number of weekend days in ongoing month Forecasts are now logged, so trends can be analyzed
https://mingle.corp.wikimedia.org/projects/analytics/cards/339
Unrelated to bug 35194. This is about forecasts for current month which occur in page view reports like http://stats.wikimedia.org/EN/TablesPageViewsMonthlyOriginalCombined.htm. There is no logic based on this whatsoever, and the estimate is meant as 'nice to know' only. But of course its value diminishes if estimates are not ballpark stable during the month. BTW the row of forecasts for current month is only presented from the 2nd week in the month forwards. Estimates can be recognized as they are prefixed by plusminus sign.
Statement about 'no logic based on this' needs confirmation. Given that https://bugzilla.wikimedia.org/show_bug.cgi?id=46209 says otherwise. It probably safer to phrase this as 'no logic should be based on this'. Last complete month instead of current month is much better to rank wikis for presentation order.