By any chance, did the steep decline in Commons' new active editors happen to begin on approximately 14 July 2016? And is it now getting steeper as of 05 August 2016?
Topic on Talk:Wikimedia Audiences
Commons new active editors
@Whatamidoing (WMF), thank you! That's a very interesting hypothesis, and there's nothing an analyst loves more than that :) As a side note, do ping me in the future when you ask data questions—I wouldn't have seen this if someone else hadn't pointed me to it.
However, it doesn't really look like cross-wiki uploads are responsible. If you look at the data below, you see these big spikes (bolded) that last for a month or two and then disappear. They usually happen in September and May-June, so I'm pretty sure that Wiki Loves Earth and Wiki Loves Monuments are responsible.
In addition, consider that cross-wiki uploads are likely to be made by users who are already making edits on another project. To get the global number of active editors, I aggregate contributions across all projects so as long as you make 5 edits on some combination of wikis, you'll be counted. So, while I think the block on cross-wiki uploads is bad for new users and likely to harm us long-term, I don't think it's causing these big spikes.
|month||new actives on Commons|
The cross-wiki upload tool is primarily used by new editors, and the abuse filter is focused on editors whose accounts are less than six months old and who have made fewer than 50 edits at Commons. Looking at CentralAuth for the last handful of users caught in the abuse filter, 60% of them (in my sample) currently have ≤4 edits (ever) across all projects. If they stop today, then those 60% won't be "new active editors". If their uploads had worked, then some fraction of them probably would be.
@Whatamidoing (WMF) yes, I definitely agree that the cross-wiki upload block will reduce the number of new active editors at the margins just by preventing those specific edits to Commons, not to mention that another limitation on the capability of new editors won't help our problem with retaining promising new users. But it isn't the major factor in the big month-to-month drop we just saw.