Readership & Wikipedia

One of the most important slidedecks you will look at this year — at least if you care about the future of Wikipedia and her sister projects — is this one, about declining and flattening Wikipedia readership, from the WMF metrics meeting a couple days ago. I wasn’t able to attend the meeting in person (fortunately, they are recorded), but I’ve been reading up on this issue, and the data presented drives home a point I’ve been thinking for a while: that this trend threatens our projects and our mission, and we need to treat it appropriately.

There’s no summing up our typical reader. It’s not possible to generalize about who they are, except that they all have some sort of internet access, because we’re talking about somewhere around 1/15th of the world’s population (that would be on the order of 400,000,000-500,000,000 readers a month — a bit less than half a billion) that access our projects via non-mobile devices.

But we can count our readers, and it’s the “bit less” that this post, and the data in the slides, is about. You see, for a long time — at least a couple of years — those of us who give talks about Wikipedia could confidently say there were half a billion unique readers a month. We were actually measuring a bit more. Now, there are less (on desktops). Mobile visitors are growing, but they seem to access fewer pages; pageviews are basically flat. This is deeply surprising, for a project whose readership and popularity has grown exponentially for a decade. And it’s counterintuitive, that the world’s biggest source of ready reference information wouldn’t continue to gain readers and pageviews as more and more of the world comes online.

Here’s the summary of what we know to date about readership, from the slidedeck above:

High level summary of Readership
● Mobile is growing, desktop is shrinking
● Globally, pageviews are flat (-0.9% AGR)
● Global North traffic (72% of all traffic) is flat
● Global South traffic is increasing, driven by mobile
● In the US, pageviews are declining (-8.6% AGR)
○ In the US, decline on desktop is not fully offset by mobile web

Globally, pageviews are flat. (Actually, they’re decreasing). For a project whose mission is to reach every single person in their own language, and for a project that depends on readership both to provide a source of new contributors and for financial support, that is a crisis.


Why is it happening? The growth of mobile views is self-evident and makes intuitive sense — mobile is where people new to the internet are mostly coming online, and even for those of us with desktop access, the convenience of mobile continues to grow and dominate our lives. So, as pointed out in the slides above, we must continue to make our mobile site a more delightful and usable (and editable) experience.

But what about the rest of the drop in desktop readership? Some of it — though we don’t know exactly how much — is attributable to new innovations in packaging information online. Things like the Google Knowledge Graph (that displays the answer you were looking for at the top or side of your search results) gives you information that is often based on data in the Wikipedia article on the topic, but without providing clicks to our pages. This service arguably fulfills our mission of sharing information, but it threatens our mission of making it possible for everyone to edit those pages — without that edit tab at the top, or even an obvious link back to Wikipedia, how do we gain new contributors, fix errors, or even fulfill the spirit of free knowledge?

And, I worry about the reasons that we don’t yet know about why readership might be dropping — I worry about the potential that Wikipedia is not as useful as it once was, or that there are other sources of information that people find more useful that aren’t free or don’t have the mission that we do.

I am a librarian. Libraries, when faced with fewer people coming into the library, have a classical set of hypotheses for the reasons why:

  • people have found a source of information that is more convenient than the library (the internet,, etc. etc.)
  • the library is hard to use or inconvenient, so people give up before they get to our resources
  • people do not know about the library and what it offers
  • the library does not contain the information or resources that people are looking for, so they must look elsewhere.

Almost always, when faced with declining traffic, some combination of all of these factors are true. In response, libraries put a great deal of effort into things like outreach and marketing, making their collections and catalogs more user-friendly, and trying to offer services that respond to their community’s needs (whether that’s a bilingual children’s storytime at the public library, or a document delivery service at an academic library). It is rarely easy to make these changes — all libraries are constrained by limited budgets and staffing, there are core services at the heart of our mission that cannot be dropped, many things about the library are inherently not very user friendly (like trying to make a collection of millions of books accessible), and it is often hard to know what to do.

But, the ways that libraries have found to surmount these challenges include listening to and surveying their community — asking, what are the changes big and small that would make the library easier to use and make library users more successful? (Having done a fair bit of this kind of survey work, I’ve found that answers are usually quite practical:  things like more power plugs, longer hours, or a better website login service are often on the top of the list). Libraries know it’s important to communicate to stakeholders (and funders) why their work is important. And libraries have a strong culture of listening to each other to gain best practices. Finally, libraries have learned that it’s important not to be dismissive of the quick wins even as you plan for bigger long-term goals — for instance, I don’t have the budget to remodel my building, but I do have the budget to buy some powerstrips and white boards, which will help the students studying in the library right away.

I argue that we at Wikimedia need to approach our solution to this problem in the same way. We need to listen, and measure, our community — that means better analytics, reader and editor surveys, qualitative as well as quantitative analysis. We need to be sure the problem is clear, so we can define it for ourselves and also communicate it cleanly within our community, so we all have a common understanding of what we need to work on solutions for. (Think about the crispness of some of our other issues, which have also taken a long time to sift through the data to figure out: we need more contributors. There is a gender gap.)

And we need to listen to other projects, online and off. This is a bit harder; we’re unique. But we’re not the only website that tries to present long-form text in a readable way, or present lots of information in a confined space in a mobile device. We’re also not the only website that serves up vast number of pages out of a database. And we are not the only mission-driven community. We do not have to reinvent every wheel.

Finally, let’s plan for quick wins and long-term goals. Quick wins: uploading photos via the mobile app. Long-term goals: easy to use, a joy to contribute, the leading shared source of information for the world. We need both kinds of objectives, and we can’t ignore either.

This is not a cheerful post, but it’s also not lacking in optimism. I believe that we have a bright future ahead of us — that this first decade has only been the beginning of the Wikipedia story, not the end. But we need to pay attention to our place in the world, and make sure that we are serving it as best as we can.

Edited since publication because: I mistyped the # of UVs (500M, not 5M); to incorporate Erik M.’s comments about mobile UVs vs desktop UVs, in the comment section; and to elaborate on what libraries do. 

This entry was posted in "How Wikipedia Works", free culture/free licenses, wik-eh-pedia, wmf board. Bookmark the permalink.

7 Responses to Readership & Wikipedia

  1. Erik Moeller says:

    The comScore data you link to (with the “Now, there are less” comment) does _not_ include mobile data, as is noted below the chart. comScore does not capture mobile viewership globally, only for a few select regions. So the trend we’re seeing in our own data (that mobile growth is offsetting desktop decline globally) is not reflected in comScore’s numbers.

    What comScore’s limited mobile data does suggest is that our global readership (uniques) may actually still be increasing, since unique users on mobile devices tend to view fewer pages per user. So if page views overall are stable and traffic is shifting to mobile, it’s likely that this means that more people are viewing less content on average. We’ll seek to confirm this hypothesis through our own data by developing a WMF “unique clients” measure. This speaks to the importance of creating a truly engaging mobile readership experience.

    If you’re on an Android device, I encourage you to check out the latest nightly builds for some interesting new readership-focused experiments, including significantly improved mobile search & improved article/infobox presentation (still experimental!):

  2. phoebe says:

    Thanks Erik. So at the moment we don’t have a good *complete* UV measure, it sounds like. I heard a mention of user session data in the metrics meeting but didn’t catch the whole comment; do we have user session info?

    I’ll edit the post to try to make it accurate. And yes, yay mobile development! I loved the screenshots of new ideas.

  3. h-stt says:

    I do not agree with your fundamental assumption that WMF’s (primary?) mandate is to raise readership. These nummeric metrics are the problem, not the solution, because they imply that WMF’s audience are the readers, while the true audience should be – no, must be – the authors.

    Only content brings in readers, only readers can be targeted by fundraisng, only funds pay employees. Focus on the authors, and reduce the stuff by half, then WMF would do a better job.

  4. Guy Chapman says:

    OK. to play devil’s advocate. The page view thing may be caused by declining participation (editors typically score larger page view hits than readers), groeth of non-English Wikipedias and so on.

  5. Robert Rohde says:

    @Guy: There are roughly 17 billion page loads per month, but only about 10 million edits per month. Each editor would have to load ~80 pages per edit in order to account for even 5% of the total traffic Wikipedia receives. And that 5% would still amount to more than 10,000 page views per active editor per month. In all likelihood, nearly all of the page views are coming from outside of the editor community.

  6. rupert THURNER says:

    i tried only one sample, “candor”. i had to enter “candor wikidata” and “candor wikipedia” to bring the page on a reasonable position in the search result to get into the wikimedia world. i do not find the quality of the entries appropriate, in none of wikidata, wikipedia, wiktionary.

    then i tried “pferd” (horse in german). on the german google it showed the de:wp article on top. on the english google, it showed it on a lower position.

    i cannot recall that i had to type “keyword wikipedia” to look for a keyword a while ago. nowadays it happens quite often that i do this. are you experiencing the same with the search results?

  7. Oliver Keyes says:

    (COI declaration: I produced the data that…all of this is based on. And am writing up my own writeups as we speak, hence being here!)

    h-stt: one flows from the other. Raising pageview numbers should not be a goal (after all, if readers can get all the information they need with one request, isn’t that a GOOD thing?), assuming that this does not have implications for editing and donating. In other words: if users can get all the information they want in N pageviews, and they have the highest chance of trying to edit or donating in N pageviews, AWESOME. Let’s not worry until we start going below N.

    I disagree that raising the number of readers should not be a goal, for two reasons. The first is: this is why we’re all here. Do you write articles simply to have written articles? Possibly. But there is a platform to write articles on, so that they can be read – so that they can inform people. Making sure that everyone has access to the sum of all human knowledge: that is the movement and the Foundation’s stated purpose. Not making sure that everyone can write that sum of human knowledge, and then who cares if it’s used or not.

    Second: readers turn into editors and donors. If we do not have any readers, we do not have a population of people capable of turning into editors, contributing the work, renewing the projects – and given how low the conversion rate is already we should be paying attention to substantial shifts in demography or numbers around readership. If we do not have any readers, we do not have a population of people who can help fund the projects (although on the other hand, I guess if nobody was using us we’d probably spend less money on servers and Operations engineers) and the platform goes away.

    You recognise this in your post, which confuses me: the amount of actual content on Wikipedia has been growing for years, whatever the population of editors. So if content brings in readers, on its own, why is the reader population not growing at the same pace? Answer: because the availability of the paths people take to get to Wikipedia is a tremendous factor in readership. Something that we wouldn’t know if we were focusing exclusively on editors, and not studying readers.

    Guy: the vast majority of pageviews are not from editors. Ignoring the numerical proportions, geographically mapping editors and readers shows a very different distribution.

    Rupert: this is what I mean when I refer to the availability of paths. Google and other search engines play a big role in driving traffic, and this is something we need to dig into, both through research to let us understand it better, and through reaching out to third-parties to see what information they have.

Comments are closed.