Forum:Wiki Stats 2020

Well met, stats fans! While it's been years since I've done this and, as we'll see, it would be better to wait another several months for a complete picture, I've realised that Quantcast has started removing its earliest data, providing us with only a year's worth. So, it's better to get it done sooner before more is lost, while I've got a chance and can't procrastinate further.

The data is provided by various sources on the FANDOM site, and by the Quantcast company. FANDOM uses Quantcast for website traffic data and analysis, and some of us admins have been allowed to look at it too. There is a shocking amount of data on demographics and browsing and shopping activity. I don't know what I can and can't share here, or what's exact data or AI estimation. Anyway, I'll focus on the FRW traffic where possible.

I'm compiling and graphing the data in Excel, which I'm not great at. Maybe a more thorough statistical analysis could be done, but I'm 19 years after failing my stats course and 9 since access to Mathematica, and this is time-consuming enough as it is. But I know, statistics is how we sort truth from spin.

Anyway, let's see how the wiki has been doing...

Rankings
In short, pretty well. On 19 July 2020, on FANDOM's Wiki Activity Monitor, the Forgotten Realms Wiki has a WAM Score of 98.12/100 and was ranked 58 out of the 5000 top wikis on FANDOM, and 28th out of the 2031 top wikis on the Gaming vertical. That's mostly video games (right now, we sit between the Assassin's Creed and Xenoblade wikis) and can be odd company (the advertising we get tends to be computer games and hardware). This of course reflects the current popularity and profile of D&D. But how do we compare to other D&D wikis?

Googling 'forgotten realms wiki' gives the FRW as first hit, above Wikipedia, and no real competition as wiki for the Forgotten Realms. Googling 'dungeons dragons wiki' gives a few: Wikipedia, us, Dungeons and Dragons Wiki, and a few on other sites that seem to be largely homebrew and a 1d4chan. I'd say the FRW is best, but I am biased. I don't have stats for the others, so I'll look at the (English-language) competition on FANDOM in the Gaming vertical: Dungeons and Dragons Wiki is 1488 and scores 27.03, while D&D Lore Wiki is 941 and scores 51.31 and Dungeons & Dungeons 4e Wiki is 1280 and scores 36.20, and they also get a fraction of our traffic on Quantcast. So, I feel confident saying the FRW is indeed the top-ranked and most-popular D&D-related wiki around. PS: According to Quantcast, the chief site also visited by our readers is D&D Wiki, which covers core, crunch, homebrew, and other settings, so there's not so much overlap.

I've been recording the WAM data randomly, but it fluctuates daily and weekly averages aren't provided, making the graphs are ugly and confusing, so I'll keep them to myself. I probably need to record every Tuesday or something. But we hit our current peak rank of 15 on 3 March 2020, with a maximum WAM Score of 98.72, rising to a new maximum of 98.84 the next day. That was achieved in the wake of the big Baldur's Gate 3 gameplay reveal. Otherwise, our position within WAM has remained at a consistent high all year.

The WAM Score is "a combination of traffic, engagement and growth", but exactly how it's calculated and what it really represents is a secret known only to FANDOM. Nevertheless, traffic is what we get naturally and growth is what we do anyway, so the only remaining factor to enhance is engagement. This is likely to be heavily based on discussion, that is, use of User Talk, Talk, and Forum pages, as well as Blogs, Article Comments and Message Walls. Even wikis for less well-known (at least AFAIK) franchises that are smaller and get less traffic can get crazy high WAM scores and rankings probably because they have high levels of communications. For example, the Adopt Me Roblox wiki (the kids play Roblox and I don't even know what it is) only has 705 pages and a quarter of our unique visitors per month, yet currently has a WAM score of 99.65 and is ranked 6 overall. But its Recent Changes is nothing but chat, trolling, vandals, blocks, and shitposts, so this form of "engagement" is not really desirable. Other high-ranked wikis seem similarly placed with high levels of discussion, good and bad. Whatever its form, FANDOM seems to be chasing interaction.

To boost WAM higher, we probably need more activity on User Talk, Talk, and Forum pages (not just in our Slack group). We could also think about using Blogs more, using Polls, and activating Chat, Article Comments, and Message Walls. Yet this is a reverse missile: it means more trolling, vandals, and need for moderation. Of course, we don't need a higher WAM ranking, WAM ranks are no measure of quality, it nets us nothing and being 50, 15, or 5 is arbitrary when we're already the top D&D wiki. However, as we've seen on Twitter and other forums, fans don't think to or don't seem to know how to talk to us and will go elsewhere and ask other people, so having more avenues of discussion could attract more editors.

PS: Moviesign added "Whatever value the WAM score has for comparing wikis has been diminished since the introduction of Discussions. The WAM metrics do not reflect any additional traffic on Discussions. Once all wikis migrate to the UCP, I suspect we will get new metrics that are more accurate and reliable." So there. In any case, once the migration is complete, the communications options available, and the ones we have a choice in, will change, so it does seem better to wait and see.

So, what is a more accurate picture of our ranking? From Quantcast, I can see a list of all the covered FANDOM wikis, ranked by their monthly "uniques". Uniques are the "number of distinct devices that visit this property". If a person visits the site on their computer 10 times, that's one unique. If person visits on their phone and computer 5 times each, that's two uniques. Obviously, that's not quite the same as people and there's a lot of overlap from people using both phone and computer, but it's the next best thing to counting individual people reading the wiki. The true number of people is somewhere less than this. (Quantcast used to do an estimate of individual people, but seems to have abandoned it.)

I've recorded these lists (discarding FANDOM and Gamepedia totals) at some random times, and compiled bar charts of the top 40 wikis by their monthly uniques. By the way, the first is during the final season of Game of Thrones, hence its huge column being an outlier that overshadows the rest.

In the first two, the FRW sits at 39th out of 40 (with 737,500 and 937,800 monthly uniques, respectively, and currently it is 34th, with 1,590,000 uniques in the last month. (By comparison, WAM gives rankings of 57th, 49th, and 58th for the same periods, no doubt diminished by their engagement metric.) Spelling it out, 1.6 million different computers, phones, and tablets accessed the FRW in the last month. We've moved up several places in that ranking just since October, while all wikis have experienced a great upswell in visits, which I'll explore further later.

Popular Pages
Next, I'll look at Special:Insights/popularpages. The values are updated weekly, and starting in May 2016 I recorded them for the Main Page and a few selected top pages each week (though I missed a few weeks) and added some more as they became interesting. The dates are my Mondays and cover the preceding week, right before they update. I divided by 7 to produce a daily average. I can't explain all the dips, and suspect some data-points are erroneous, but there are some interesting peaks and troughs.

First, the obvious spikes. Simply, Stranger Things Season 2 came out on 27 October 2017 and Season 3 on 4 July 2019. With its D&D references, people searched for Demogorgon, mind flayer, and Vale of Shadows, and found the FRW, prompting huge increases in page views in the week after. This also happened for Season 1 on 15 July 2016, but I didn't know about the series and wasn't recording the relevant pages then. I know Vale of Shadows was the most popular page at the time, so I thought it was in error. So was the series – Plane of Shadow would've been more accurate. Anyway, I imagine we educated them on the D&D lore behind the series. Some of these visitors seem to have checked out the rest of the site, judging by the small bumps to the Main Page.

The rest of the time, Demogorgon is about as popular as Drizzt. Curiously, not even Drizzt novel releases cause a blip in his page views. Maybe Drizzt fans don't need the refreshers, or already know our Drizzt Do'Urden page isn't nearly complete.

There are more mind flayer spikes, corresponding to Baldur's Gate 3 news: the first trailer on 6 June 2019 produced a bigger bump to mind flayer than even Stranger Things, and the gameplay reveal on 28 February 2020 produced another. I started recording Baldur's Gate, but much too late and it's still obscured by Drizzt and Demogorgon.

You'll notice some fluctuations over time and a general rise. Partly, I assume this is due to the increased popularity of D&D driving visitors and our own higher profile and quality retaining more readers. Some notable dips in the middle of Decembers are of course Christmas and more generally the holiday periods. I suspect there are longer-period seasonal variations (e.g., US summer holidays), but they seem obscured by more local activity and overall growths, which I'll look at in more detail later.

There's also a clear divergence in the popularity of the topics I picked initially. I presume interest in general D&D and FR topics has grown, while interest in Drizzt novels has not. Of course, WotC has focused recent adventures on Waterdeep and drow, forcing more interest in some topics than others. Streaming game groups like Critical Role also drive interest in various monsters and races on occasion (morkoth is currently top viewed). Otherwise, the top viewed pages in Popular Pages consistently show the main areas of interest are playable races and deities (suiting players making characters) and common monsters (suiting DMs making campaigns). We've got a lot of good monster coverage, but I believe common player races need development, and almost all our gods need a good deal of work.

Traffic
The Wiki Analytics Admin Dashboard (admin-only, sorry) tells us we currently get around 330,000 total views per day, 2,300,000 over the last week, and 9,725,600 over the last month. They wobble a lot but generally peak on weekends. They only go back a week and a month, and I'm afraid I haven't been recording these. I remember when these were around 100,000 and this is sitting at a maximum.

In any case, Quantcast records comparable data over longer periods, so I'll focus on its data. This where things get interesting, and kind of disturbing.

First, the daily total page views, from 20 June 2019 to 18 July 2020:

There's quite a lot of spikiness here. Once again, the peaks are on the weekends, centered on Sundays. I had thought this was due to weekend game days and campaign prep times, and I was right, more or less. This pattern is repeated for FANDOM as a whole and for Gamepedia and specific gaming wikis, and not for TV or movie franchises, unless the TV series comes out weekly. So people are more likely to play games, whether video games or tabletop roleplaying games, on their weekends, not surprisingly.

I'll note that the blue line above is US data and the grey line below is the rest of the world. (It turns out I could've merged these into Global data, but it's too late now and the distinction is more instructive.) These are flipped for FANDOM as a whole and in non-US franchises like Doctor Who. This is probably because D&D is still much more well known and popular in the USA than in other countries. That is, the US readership for the wiki is greater than for the rest of the world combined. We see significant dips at the 4 July (cancelling the expected Stranger Things S3 peak) and Thanksgiving that don't appear outside the US, and at Christmas, which does. There's also the BG3 peak again, obscured but more obvious in the non-US line.

Here, I'll take a moment to look more closely at the international comparison. Here are the top six nations for uniques visiting the wiki, as estimated by Quantcast from demographics. Obviously, the top four English-speaking nations, and Brazil and Germany where D&D is also big. The USA has more than half the wiki readers, and thus D&D players and fans, in the world, and twice as many as the next five nations combined. Within the US, the top states are estimated to be California, Texas, Washington, Illinois, and New York.
 * United States: 55.14%, 426,000 uniques
 * Canada: 8.77%, 67,700 uniques
 * United Kingdom: 8.28%, 63,900 uniques
 * Australia: 4.86%, 37,500 uniques
 * Brazil: 3.35%, 25,900 uniques
 * Germany: 2.24%, 17,300 uniques

Next, to smooth out the data, the weekly total page views, from 20 June 2019 to 18 July 2020:

Now, we see some interesting behaviour. The lines are roughly constantly if slightly downward-sloping through the second half of 2019, which is more noticeable in the smoother monthly data. I wish I'd done this before Quantcast dumped everything back to October 2018 so I could see if this is seasonal but, IIRC, it was showing a steady rise the whole time. Then they reach a minimum in the Christmas period, as expected, and pick up for the start of 2020. And they go up and up, then down, then up, and plateau, then down, then up, and up... What happened? The pandemic did, of course.

But Quantcast has stopped cooperating and loading data, so I'll have to leave it here and come back and dig down into the data in more detail next time. — BadCatMan (talk) 12:36, July 20, 2020 (UTC)

Back again, and I'm going to show the daily and weekly uniques (that is, individual devices, a proxy for people), from 21 June 2019 to 19 July 2020 (one day shifted from the above):

This shows all the same behaviors as for page views, of course, but without quite the same downward trend in the second half of 2019, so hardly any readers are leaving, they're just reading less pages. Then we got a lot of new readers in 2020. Let's look at them.

Pandemic
It's difficult to say when the rise began, as it meets the post-Christmas recovery rather smoothly. I could put it in mid/late-January, once the uniques have returned to their prior level and the news of COVID-19 became prominent and alarming. After all, a new height was reached in the week ending 24 January, and it hasn't dropped below since. But, most likely, this is just a return to regular levels. When I look at similar data for FANDOM as a whole, Gamepedia, and other wikis, they don't show this upswing.

But, in all of them, there is a significant rise with a distinct turning point in the middle of March. The WHO declared COVID-19 officially a pandemic on 11 March and many countries issued lock-down/stay-at-home orders over the second half of March, providing a nice clear trigger for a change in behavior over the following weeks, namely, people spending more time at home online and reading wikis. In fact, looking at the overall FANDOM data, I'd place it at 15 March, some time before most stay-at-home orders came into effect, showing people listened to the news and the WHO and stayed home before governments started making it official.

And, when Quantcast recovers, I'll try to look at that in more detail. — BadCatMan (talk) 06:22, July 21, 2020 (UTC)

Back. It seems I get throttled for using the Quantcast site too much. By the way, to clarify, I'm not saying the FRW in particular is more popular because of the pandemic, but rather people are spending more time online in general, and I'm using the FRW as example to illustrate that.

Those unique devices can be broken down into desktop computers and mobile phones/tablets. So, here is the cross-platform site traffic in terms of weekly uniques, again from 21 June 2019 to 17 July 2020. Mobile and desktop computer usage diverge significantly, in both the US and outside, with desktop increasing markedly while mobile stays level or declines. This is supported by FANDOM's Wiki Analytics: desktop is now 59% compared to mobile and tablet at 41%, whereas they were formerly about even. So, the obvious conclusion, people in lock-down or isolation, or just staying home for safety or because everything's shut, are of course more likely to use their computer than their phone to browse the web.

Now, similarly, here is the cross-platform site traffic in terms of weekly page views, from 21 June 2019 to 17 July 2020. We can see people at home are reading a lot more pages, and the rest of the world even reaches parity with the US.

I want to investigate this engagement further: is it just an overall rise in internet usage, or are people spending longer on the site and reading more pages? Are the new readers passing by or are they staying? Unfortunately, Quantcast's engagement data (calculated from cookies and statistics) only covers the last 30 days, so I can't track this over time and the course of the pandemic. So, I've just got this snapshot. IIRC, the passers-by percentage used to be higher, so the percentage of regulars is probably up, which I'd expect.

Finally, I'd like to look at the fluctuations in uniques and page views since the pandemic. Is this following number of infections or based on lock-down conditions? The USA is the best example, as it has the highest numbers to give the clearest data, and is also the most seriously affected, giving the starkest contrasts. Because individual states have been affected differently, at different times, and responded differently, it would be more instructive to look at a single early-affected state like New York, but unfortunately only national data is provided. The other countries are too low to show anything clearly. So, I'll look at data for the US as a whole.

So, since the pandemic was declared and lock-downs came into effect, more and more people stayed home and stayed online. Then the rates slowed and either plateaued or peaked at the start of May. I imagine a maximum saturation was reached – anyone who's going to look at wikis for things they're interested in is already doing so. After that, things fluctuate and it's difficult to say what the ups and downs are linked to. I tried plotting against new coronavirus cases (as seen and obtainable at Our World in Data, but it didn't show anything like a correlation. Our data is still too sensitive to things like BG3 news and the Diversity and Dungeons & Dragons debates on 17/18 June to reveal any clear connections. The overall FANDOM data is smoother, but also doesn't show anything as obvious as the mid-March increase. But it's clear it has declined since mid-May, which could be due to lifted lock-downs and restrictions in some parts, the Black Lives Matter protests, or just people getting sick of the internet again. The current level may just be the "new normal" of internet usage. Unfortunately, the situation in the US is too patchwork to be reflected in the national data and I've reached my limits for data and analysis.

Strangely, the last two weeks have seen another sharp uptick in the FRW traffic, which isn't in the overall FANDOM data. What that is, I don't know. Am I missing some news?

In this analysis, I had hoped to see some interesting correlations between the pandemic, stay-at-home conditions, internet usage, and our own wiki traffic, and to a degree, we did. From mid-March, there have been significant increases in internet traffic, at the FRW and across FANDOM. Specifically, there was increased usage of desktop compared to mobile. Both are most likely due to people staying home and staying online, not surprisingly. This is to be expected and has already been reported, such as in this New York Times article. Yet visits to a real-world-oriented wiki, namely Wikipedia, have risen and then declined to pre-pandemic levels, while gaming sites like Twitch and FANDOM increased and stayed high.

Addendum
I thought I was done, but it turns out I could still download the entire dataset from Quantcast. So of course I went back, downloaded the US and Global data, and subtracted one from the other to get the non-US data. It goes back to 30 October 2018, but I've remove the first seven days from the 7-day average data and the first thirty days from the 30-day average data so they weren't distorted by the initial lack of data, but I suspect the early data is still erroneous. By the way, these are rolling averages, so there's a new average each day for the previous seven or thirty days, not a data point for each week or month. I couldn't work out how to get Excel to give me that. It seems smoother and more illustrative than a fixed average would be.

In the data, Quantcast provides the Uniques (devices), Views (of individual pages), and also Visits. This is the number of distinct sessions of browsing, e.g., if they visit once in the morning and once in the evening, that's two visits. It's not an option on the website. So because more data is better, I've graphed the weekly and monthly uniques, visits, and views going back to November 2018. Bear in mind, these graphs don't exactly align as I've removed 21 more days from the monthly graph.

Once again, you can spot some familiar dips and peaks, like Christmases, Stranger Things, and BG3 news, and lately, the pandemic. But smoothing it out further into the monthly data reveals the overall and seasonal trends I was looking for. Traffic dips in the holiday season, with a minimum at Christmas time, then picks up sharply by the end of January, then plateaus through February and March, before declining to another minimum at May/June, and again picking up sharply in time for July, plateauing around August and September, and descending toward December, perhaps with a bump in November. It's much more pronounced in the US data, of course, but might be glimpsed outside. The fluctuation has an alternating period of five/seven months and has minima at the US winter/summer holidays, or rather, the breaks between school and college semesters (I assume, I'm not in the US), as the demographics estimate the majority of visitors are teenagers and young adults at high school or college.

This seasonal trend is overlaid on an overall rise in traffic, producing a step-wise growth. While this only goes back to late 2018, it continues the growth we've seen in Popular Pages and earlier data sets, though it seems more emphasized now. While this could just be due to the increased popularity and profile of D&D since 5 edition in recent years, as I've said before, the fans wouldn't come here if we didn't have something worthwhile. And I believe the data shows this: increasing numbers of people come to the wiki, they visit more often, and they read more pages while they're here.

Especially during the pandemic. The pandemic hump obscures the current cycle since March and enhances the growth rate significantly. The recent uptick isn't anything unusual or pandemic-driven, just the July rise out of the June dip, even if that dip is still higher than what we'd have normally.

Finally, what I really want to observe is that people are not just stuck at home and bored and surfing the web. Instead, people in lock-down seek escapism, in isolation they want exploration, and in a dangerous world they'd rather dive into another. It's an obvious conclusion, but not one I've yet seen made. Ultimately, I feel rather satisfied that D&D and the Forgotten Realms Wiki are providing another avenue of escape for people around the world.

As I said in past Clacks, stay in and explore.