Jump to content

Wikipedia:Wikipedia Signpost/2015-08-26/In focus

Add links
From Wikipedia, the free encyclopedia
The Signpost
← Back to Contents
View Latest Issue


In focus

An increase in active Wikipedia editors

Contribute  —  
Share this
By WereSpielChequers

According to one possibly over-simplistic measure, the core Wikimedia community, and in particular the core community on the English Wikipedia, has recently stopped declining and might even have started to grow again.

Editors saving more than 100 edits per month on the English Wikipedia
Month 2014 2015 change
January 3,232 3,312 80
February 2,957 3,051 94
March 3,131 3,309 178
April 2,979 3,156 177
May 3,051 3,223 172
June 2,981 3,245 264
July 3,024 3,399 375


Editors saving more than 100 edits per month on all Wikipedia languages
Month 2014 2015 change
January 10,331 10,625 294
February 9,508 9,779 287
March 9,936 10,446 510
April 9,533 9,986 453
May 9,689 10,075 386
June 9,276 9,891 615
July 9,420 10,280 860

For some years now the English Wikipedia and the Wikimedia movement generally have been losing active editors faster than they have been recruiting them.

But one interesting indicator has now started to climb and indicates that the core community may actually be growing again. Though a range of other indicators from the appointment of new admins on the English wikipedia, the number of new accounts created, and the number of editors doing more than five edits per month are still flat or in decline.

The number of editors saving more than 100 edits each month is a long-standing metric published about Wikipedia and other WMF projects. For seven consecutive months, from January to July 2015, that indicator has been positive on both the English Wikipedia community and the whole Wikimedia project—though the situation is more complex on some other sites, such as the German Wikipedia.

We know there are seasonal events that affect the community, and months themselves vary in length, so February 2015 was shorter than January or March; but more editors were contributing more than 100 edits that month than in February 2014; similarly, in January 2015 there were more active editors than in January 2014, a trend that has now run for seven months. Last month, 12,349 editors made more than 100 edits across all projects, 10,280 editors across all versions of Wikipedia, and 3,399 editors on the English Wikipedia, as opposed to 11,257, 9,420, and 3,024 editors respectively in July 2014.

The matter has been discussed on the research mailing list, Wiki-research-l, during the past two weeks.

As with any data over time, there is always the risk that this could just be anomalous, but Wikimedia Foundation data analyst Erik Zachte has now said of the phenomenon: "The growth seems real to me." Zachte has also pointed to the late 2014 speed-up of editing on the Wikimedia sites as a potential contributor to the increase. Implementing HHVM speeded up the saving of edits, which should logically have more impact on wiki gnomes doing lots of small edits than on editors who make just a few saves per hour.

Another theory suggested on the research list and elsewhere has attributed the increase to the improvements to Visual Editor, though with barely ten percent of the most active editors on English Wikipedia using it, it is unlikely to be a major or sole reason for the apparent increase.

The different leadership style of new Foundation executive director Lila Tretikov may be bearing fruit, in terms of better relations between the Foundation and the most active editors.

There is also some concern that Editors saving over 100 edits per month is a simplistic metric; for example, it will include users of highly automated tools such as AutoWikiBrowser, STiki, or Huggle who may achieve that edit count in less than an hour per month, but omits an editor who spends an evening every week writing or rewriting one or two articles, but who might only save an edit every half an hour in that evening.

Should the trend continue, and assuming that someone doesn't find a software bug that has caused the anomaly, future lines of analysis could include examining how much of the increase is due to fewer editors leaving, more inactive editors returning, more new editors joining, and a greater number of casual editors increasing their editing frequency to more than 100 edits per month.

August figures are expected in about a month. It will be very interesting to see whether the trend continues.

+ Add a comment

Discuss this story

These comments are automatically transcluded from this article's talk page. To follow comments, add the page to your watchlist. If your comment has not appeared here, you can try purging the cache.

The Meme of decline

  • There are many ways to criticize this metric, and clearly the 100+ edits a month statistic is an imperfect way to measure the count of "really useful editors", paraphrasing Thomas the Tank Engine. But it is a useful metric, and I see no evidence that the ranks of disruptive or obsessive editors racking up 100 edits or more a month is increasing at the expense of genuine content creators. This data calls the "Wikipedia editor count is plunging!!" meme into question. Cullen328 Let's discuss it 06:24, 28 August 2015 (UTC)[reply]
    • It hasn't just been a meme. The raw figures had been falling for some time, and while part of the decline was due to the edit filters, and part to increasing speed of vandalism reversion, a lot of people believe that most or all of the decline was "real". Yes there will be a time in the distant future when edit count on individual language projects will be as useful an indicator as tonnage of horse shit removed now is of London traffic. If and when editing of infoboxes moves to Wikidata then inevitably the edit count on individual language wikipedias will dip, just as the change to a hub and spoke system for intrawikis lost us lots of perfectly good bot edits. If we ever reduced the default from four warnings to three before we blocked vandalism only accounts, then we would not only lose more good edits than bad (one rollback and one userpage warning for each lost vandalism) but our cherished and longstanding metric of editors doing 5 or more edits in the month would take a knock. By the mid twentieth century the AIs could have taken over this site and be generating each language version of an article from a common core except only where the local language community had opted for local human control of an article of special cultural significance to their language. I'd like to see us measure something more nuanced such as unique hours in which a volunteer saves an edit. But as long as we treat it as an indicator, and remember that what it indicate does change over time, then I think these statistics are worth collecting and monitoring. ϢereSpielChequers 10:10, 28 August 2015 (UTC)[reply]

Edits per article, Wikipedians per article

A nice piece: the downward trend had been showing signs of reversing for a while.

One thing to bear in mind though is that the number of edits per article has clearly declined. In March 2007 there were 4.8 million edits spread across 1.6 million articles (about 3 edits per article). In March 2015, there were 3.1 million edits spread across 4.8 million articles (about 0.6 edits per article).

So articles as a whole are stabilising. In some cases, this may be because they have matured and become really good (I believe featured articles for example generally see fewer edits) or "good enough", in other cases it may just be a question of fatigue on the part of editors who used to fight over content – then you get articles that look like abandoned battlefields (I can think of a few).

In addition, new articles added see fewer edits than they used to in the past – possibly because they are in niche topics, with shorter content and fewer people interested in them.

What has also sharply declined, of course, is the number of highly active Wikipedians per article. This has potential implications for quality control. For example, hundreds of thousands of articles are on no active editor's watchlist. Andreas JN466 14:32, 28 August 2015 (UTC)[reply]

Hi Andreas, I'd agree there has been a fall in edits per article, though I'd point out that part of that will be the things that contributed to the overall fall in the number of edits. Edit filters, faster vandalism reversion, the revised intrawiki system etc. To some extent improved edit filters and pending changes can also compensate for the decline in watchlisting, my own watchlisting picks up rather less vandalism than it once did. I suspect we are also filling in the gaps with lots of uncontentious articles that rarely change, it would be interesting to do stats on the proportion of contentious articles at any time - my suspicion is that the number of disputed articles grows far more slowly than the number of articles. But I do agree that there is a problem, and my preference is that at some point we introduce flagged revisions as it works on DE wiki. That way we can be sure that every edit by an IP or newbie is looked at by at least one regular. ϢereSpielChequers 16:48, 28 August 2015 (UTC)[reply]
I interpret this low-edits-per-article-overall aspect of the statistics very pessimistically. And indeed, I interpret the recent unmistakeably-upward trend in active editors extremely pessimistically.
the pessimist's guide to uptrending very-active-editors , coupled with flatlined-edits-per-article
First things first: when you see that the average article is getting less than one edit per month, to me that doesn't indicate that wikipedia finally has all the articles we need and all those existing articles are complete barring addition of rare ongoing news-events, as clicking Special:Random twenty times in a row will clearly demonstrate is utterly false. To me, that stat indicates that we have hundreds of thousands of articles edited solely by one or two humans. There was recently a sock-investigation, where several hundred sock-usernames were shaking down good-faith-but-COI-encumbered wikipedians for cash off-wiki, in exchange for "rescuing" specific COI-encumbered-on-wiki-articles, usually by storing the advertorial-puffery at some little-used existing-redirect-location in mainspace, and also usually by unattributed-copyvio. The result of this scheme: "new" article is created (usually a spelling-variation or existing long-stable redirect), edited less than ten times to insert puffery by an SPA, and then the SPA evaporates, and the COI-encumbered editor makes one or two changes per year. Articles like that are, by definition, low-edits-per-article, which get approximately one edit every few months.
Now, clearly throwaway-SPA-usernames, and one-article-focus COI editors, obviously aren't increasing the active-100+edits-per-month category. But they *are* very much decreasing the average-edits-per-article statistic. Which leads us to, second things second: why does that seemingly-unrelated issue, low-edit-count SPA and COI editors creating and WP:OWNing low-traffic articles, make me distrust the recent uptrend in very active editors? For one simple reason: anecdotal evidence.  ;-)     Okay, okay, not just that.
But there is anecdotal evidence, to my eyes at least, that disclosed-paid-COI-editing has been exponentially increasing the past three years, and that undisclosed-paid-editing has been *rapidly* increasing the past few years, often by staffers, but sometimes (and increasingly often) by quasi-professionals that specialize in paid-wikipedia-edits (overt or covert). If this anecdotal evidence of highly-active paid-editor-population, is taken at face value, it is still just anecdotal evidence. But logically, if you combine the statistical trend, that we have a lot of unwatched articles, and a metric truckload of articles edited by just one of two people ever, PLUS we also have anecdotal evidence of highly active highly organized paid editing increasing, the conclusion one might draw, varies depending on whether one is an optimist or a pessimist about humanity.
One possible conclusion: suddenly, wikipedia is once again a warm and fuzzy place where lovers of knowledge can convene in beautiful harmony. I'm not seeing beautiful harmony at AfD. I'm not seeing it at AfC. I'm never really expecting to see it at the dramah-boards, either. But worst of all, I don't see it on article-talkpages any more: they are either silent, because nobody is there, or they are silent, because they are the DMZ of a contentious article with wiki-gangs waging wars of attrition, or they are noisy with a wiki-gang battle gone hot temporarily ("to the wiki-mattresses" said Don Wiki-Corleone). All too often, the talkpage is silent, because an article filled with puffery was created in one fell swoop, and then left in mainspace unmolested for years and years, or at least, months and months.
Anyways, I would be delighted if the uptrend was a legit indicator that, despite the harsh wiki-culture, gradually the general public is learning to love wikipedia for itself, and there is a growing awareness that to improve wikipedia, average citizen-volunteers will need to roll up their sleeves, learn the bazillion wiki-policies by heart, and cite every sentence fragment they include to seven different mainstream wiki-reliable publications lest their contribution be insta-deleted. I fear it is vastly more likely that, rather than a growing movement to *save* wikipedia from degradation, which is causing an unexpected influx of citizen-volunteers, there is instead a steadily-growing hardening of public opinion that wikipedia is a horrible nasty place to contribute as a beginner, because everything is deleted, and everyone speaks in crazy wiki-jargon, that invariably translates into You Cannot Do That Here So Obey Or Be Blocked.
Thus, a more pessimistic conclusion: what we instead are witnessing, is the crass commercially-driven influx of a mob of paid editors, who are here to make an off-wiki buck here on-wiki, and are thus willing to suffer the revert-first-ask-questions-never wikiculture, and the byzantine bureaucracy, because they are being paid in cold hard coinage for the frustrating and difficult work they do: getting around the wiki-policies and finding the quiet corners of the 'pedia to add puffery unmolested by the wiki-cops.
Now, for reasons methinks are obvious, I would very much like to be proven wrong, about the pessimistic conclusion being the more plausible one. But, since I see zero mentions of "paid*" nor "disclos*" nor "sock*" in the statistics being offered here, nobody seems to be *trying* to disprove me, and I'm not sure whether that is because my fears are implausible and my anecdotal evidence flawed, or because other people are more optimistic about humanity than myself, and have not considered the alternative explanation for the uptrend. Questions: what percentage of the uptrend is paid editing by disclosed paid editors? What is the estimated ratio of disclosed-paid-editor-count, to undisclosed-paid-editor-count? Have those figures changed since 2013, when the current uptrend kinda-sorta-began? Those are my pessimistic questions. I realize that the wiki-stats available won't have my answers, since to get exacting answers would require analysis of every single edit by unbiased neutral humans. (Catch-22! those are the people we want *making* new edits ... not spending time analyzing *past* edits.)
But I do think that some of the figures we have could be analyzed, to detect agenda-and-promotional-and-battleground-type editing, in a specific fashion. For instance, of the July-uptick, how much or how little of that uptick is related to the 381-strong sockfarm just revealed? We know the sock-list now, so this should be a straightforward SQL query, for somebody with backend-privs. As another idea, more likely to give us rough answers of the kind we're looking for: what if we picked a random sampling of 1000 edits from the June'15 dataset, and analyzed those alone (e.g. by publishing the list of the 1000 randomly-selected edits here on-wiki then crowd-source analyzing them all in divide-and-conquer-fashion). We could categorize them into sourced-improvement, quasi-sourced-quasi-improvement, neutral-grammar-or-similar-tweak, unsourced-addition-factual-looking, unsourced-addition-promotional-looking, quasi-sourced-addition-agenda-pushing, blatantly-vandalism, and maybe a few others?
Then, do the same for 1000 random edits from June'14, from June'13, from June'12, and so on back to 2005 or something, to give us a nice long ten-year-trend. I'd rather see those type of statistics, grouped by type of edit, than the generic anonymized statistics grouped by edit-count-itis-standing. That would give us some insight into whether the uptrend is an increase in good-faith contributors... or instead, to quote a memorable phrase used above out-of-context, a metric related to the tonnage of horseshit that needs to be removed. Gracias, 75.108.94.227 (talk) 16:17, 1 September 2015 (UTC)[reply]
Some of these things are difficult to test, and just because no one is volunteering to test them doesn't mean they agree or disagree with them. I have tested the direct effect of Orangemoody on the >100 a month metric. I've checked a dozen of those accounts at random and only one was even close to 100 edits in any one month, and while many now have more deleted edits than live, the pattern includes gnomish edits as well as article creation, so many of them will have live if reverted edits. So I doubt there will be much direct effect on the >100 edit metric, as unless my sample is ridiculously skewed the great majority of those editors won't have reached the 100 edit total in any one month. Indirectly of course there should be some effect, deleting a bunch of articles created over the last few months will take out all the edits to those articles, and there could be categorisers or other gnomes who drop below 100 live edits in one or more recent months as a result of this. Whether that exceeds normal attrition will be hard to spot, but sometimes AFD deletes contentious articles with a lot of contributors and if that happens it could eclipse the effect of these deletions. I would not be at all surprised if the July uptick disappears when we get the August figures, many articles created in July won't have been deleted until August, and these will be skewed towards one that were categorised, wikified and deletion tagged in July. So when August comes out if July 2015 has dropped below July 2013 but above July 2014 I won't be surprised but I'd suspect it is more likely normal deletions than Orangemoody - not least because any effect of Orangemoody will be spread across the several months that those edits were made. As for the rest of your comments, I don't know how to estimate how much undisclosed paid editing is going on, and I doubt anyone else does either. My suspicion is that spam generally increases in line with perceived value to marketers of an entry in Wikipedia, and that probably lags one or two years behind our total readership, but how do you measure covert activity? ϢereSpielChequers 13:05, 3 September 2015 (UTC)[reply]
Where I hope Orangemoody will have an effect (a negative one in terms of number of edits and number of articles, but a positive effect on the much less easily measurable factor of quality) is inducing us to remove the similarly promotional articles inserted by other promotional editors, and especially the many thousands added in the past when our standards were lower. The only real way to prevent such abuse is by more careful scrutiny, and I hope the effect of this concentrated effort to remove some of this material will be to encourage people to work much harder at detecting and removing others. DGG ( talk ) 23:25, 3 September 2015 (UTC)[reply]

Something that brought me back

I had really slowed down my content editing up until about 1 month ago, largely because I am working for the Foundation on WP:TWL, but also burnout. I know the meta:100wikidays challenge has helped me feel more tied to coming back to contributing (and contributing content). I have noticed an increase in the number of similar collegial activities in the last couple years: whereas when I really burnt out and backed off in 2012/13, those were nearly as accessable. I wonder if we should be encouraging more editing contests, to keep people engaged (there is even a tool kit for running such events being worked on, based on community learnings). On English we don't have as many of these kinds of spaces as we could (geographically focused language communities tend to do it better), Sadads (talk) 17:03, 28 August 2015 (UTC)[reply]

Wikiprojects are ghost towns

I quit editing a while back, maybe 5 years. Trying to make Abe Lincoln an FA broke my will to edit. I've been checking my watchlist daily for the a last few weeks now, though. I see some of the same people around still active, like Black Kite and BOZ. I miss the editors from the old inclusionist/deletionist wars we used to have who aren't around anymore, like Ned Scott and Matthew. Win or lose, I don't think WP has ever had anything as addictive as that particular fight. Now it's over.

Anyways, I still have a bunch of wikiproject pages on my watchlist, but they barely ever come up anymore. WProject comics used to have maybe 10 comments a day. Now it's a few a week. Same with the rest. Things are way less active than they used to be. But, I'm back, doing a little bit here and there, though. Maybe we hit rock bottom and will start moving up. I think wikinews never came back, though, as an example, so who knows. - Peregrine Fisher (talk) 03:07, 3 September 2015 (UTC)[reply]

Welcome back! I find the active pages, editors and even wikiProjects change over time, so looking at a bunch of pages that were active several years ago will likely include more of the ones that have become less active than the newly active ones. My experience of WikiProjects is that many have watchers who return if there is a query and respond if it interests them. So while I agree that overall we probably are less active than at peak, it is easy to get a skewed impression and difficult to really measure how the community has changed. ϢereSpielChequers 12:33, 3 September 2015 (UTC)[reply]

The trend continues for August

August stats are in and the trend continues. I'm starting to monitor this at User:WereSpielChequers/100+ editors ϢereSpielChequers 17:36, 8 September 2015 (UTC)[reply]

There is an interesting discussion at JimboTalk about picking ten random articles, and then comparing their article-quality today, versus their article-quality one year ago in 2014, versus their article-quality 5 years ago in 2010, versus their article-quality ten years ago in 2015. Although it is somewhat biased by the eye-of-the-beholder phenomenon, it does seem to be a useful metric.
    Is there a way to run a similar test on the usernames with 100+ edits per month? Which is to say, randomly pick ten of those usernames, and then examine their contrib-quality for September 2015, compared to September 2014, September 2010, and September 2005? That might give us a rough idea of what sort of editors, doing what sort of work, are causing the uptrend. 75.108.94.227 (talk) 05:47, 11 September 2015 (UTC)[reply]
I don't know how to get that data myself, but I fear it would have the same drawback as ten random articles in that it wouldn't be statistically valid; Especially if you were looking for sub trends such as FA contributions. Ten editors might give you a rough idea between gnomes and content writers, but in the unlikely event that a single FA writer happened to be among the ten that would skew things hugely. It would also be incredibly skewed if you limited it to ten editors who have been active all the way through as opposed to ten random active editors for each year. Yes there are editors who have been consistently active since 2005 - I've heard that until this year the 2006 cohort was the most productive in every year since 2006. But those editors have themselves changed over time, some may have gone from vandalfighting adolescents to article writing postgrads. Others may have written the articles they were interested in and since hung around maybe doing an hour or two on AWB or Huggle once a fortnight. I've only been here since 2007 so I wouldn't make your sample, but I guess I'm in the group who found some niches, in my case categorisation and typofixing, and much of my editing by volume is still in those niches. But my earliest edits definitely won't have included citations, and while I add little content nowadays, when I do I cite it. So if you were looking at it from that angle I would suggest a much bigger sample size, and decide if you were looking at how longterm editors change over time, or at how the currently very active community changes from year to year. Personally I'm taking a very different approach, looking for anomalies and patterns, hypothesizing various things that could have changed the stats over time and then trying to quantify that effect. For example there was a suggestion that the Orangemoody related deletions would have a measurable effect on this by deleting hundreds of articles created in the first half of this year. User:WereSpielChequers/100+ editors is my evidence that the effect of Orangemoody on the >100 edits stats was very probably trivial, or just conceivably, big but almost exactly balanced out by an otherwise unobserved and unexpected move in the opposite direction. ϢereSpielChequers 11:07, 11 September 2015 (UTC)[reply]

Wishful thinking

We have years of decline, then one indicator goes contrary. If it is reflective of anything it is the bad actors beginning their steady march. As another editor mentions above WikiProjects have mostly failed. Barnstars and WikiLove are nice and all but they have nothing to do with the integrity of the project. - Shiftchange (talk) 07:21, 5 October 2017 (UTC)[reply]

It isn't just one indicator now, though we held back the publication of the article for months while people double checked the stats on this one. At the other end of the scale look at Wikipedia:Time Between Edits. For the last year we have been pretty consistent at about 60 days per ten million edits, in 2014 we had three consecutive ten millions that took over 70 days. As for Wikilove, thanks are very nice, but they aren't logged as edits and I'm sure that sometimes they replace them. ϢereSpielChequers 23:35, 24 October 2017 (UTC)[reply]
Keep up with The Signpost on Twitter, Facebook or Mastodon.
Retrieved from "https://en.wikipedia.org/w/index.php?title=Wikipedia:Wikipedia_Signpost/2015-08-26/In_focus&oldid=1193875151"