Research talk:Daily unique registered editors

Change name[edit]

@DarTar: To fit this metric in the structure of previous metrics, (e.g. newly registered users), I'd like to change the name to unique registered editors (drops "daily" and explicitly specifies "registered") and allow for arbitrary time bounds to be used. Is there a reason that this metric and it's kin are specified as daily? --Halfak (WMF) (talk) 14:47, 8 September 2014 (UTC)[reply]

"Daily" seems quite useful to me. The number of editors in an arbitrary time period has an arbitrarily different meaning... --Nemo 19:12, 8 September 2014 (UTC)[reply]

That's a fair point. Counting "uniques" is much different than counting the number of events. --Halfak (WMF) (talk) 21:02, 8 September 2014 (UTC)[reply]

Singularity time[edit]

Thanks for the graphs, which are yet another representation of the recent editing activity collapse. Especially for it.wiki, however, I'm interested in understanding the timing of the drop with more precision: here it looks to be around 2013 in de/es/fr, but perhaps even in 2012 for Italian? With the smoothing/rounding it gets a bit hard to see. Would be interesting to put every wiki at index 100 in, say, January 2012 and see how they evolve till 2014. --Nemo 19:12, 8 September 2014 (UTC)[reply]

Hey Nemo. You don't want to actually look at the daily trends of this metric since it's super noisy and the lines just overlap. In the plot below, I narrowed the timebounds as you requested and I also normalized each wiki by the average daily unique editors in Jan. 2012.

Relative change in daily active editors since Jan. 2012. The number of daily unique editors is plotted as a factor of the average daily unique editors in Jan. 2012 for English, German, French, Spanish and Italian Wikipedias.

You can see that all but German seem to have the same general shape of decline with a minor recovery on January. I suspect that this is seasonal and Germans are weird. Note that for editors making at least 5 edits, the Italian Wikipedia is declining too fast to experience an actual recovery -- just a slightly slower descent for a bit. --Halfak (WMF) (talk) 21:53, 8 September 2014 (UTC)[reply]

January is always a peak so it makes sense that all other months are below 1, but as you say there is no recovery. What a painful graph! --Nemo 06:58, 9 September 2014 (UTC)[reply]

Effect of Wikidata[edit]

It could be possible that some or all of the changes in 2013 forward are due to Wikidata, since edits concering language links would happen with a much much lower frequency than before. This especially concerns editors who are maybe proudly creating an article in their small homewiki and now going to the large Wikipedias to enter the langlink to their new creation. If we could tag and remove all edits that only touched langlinks, we would get an answer to whether this theory could explain some of the dip. --denny (talk) 19:37, 8 September 2014 (UTC)[reply]

I've not yet looked too much into the "one edit" editors count dynamics but certainly it would be affected by Wikidata at some extent. However it's IMHO impossible for Wikidata to explain the pattern this metric is surfacing (at least for it.wiki), because you see the same across all classes: Research:The sudden decline of Italian Wikipedia#A closer look. Other it.wiki users also think the trend may have started before March 2013, in which case Wikidata would be out of the table. --Nemo 06:58, 9 September 2014 (UTC)[reply]