So-called “dark social” traffic turns out to be mostly Facebook

Despite the fact that our ability to track and measure almost every aspect of user traffic has never been stronger, there has been a strange gap in our knowledge for some time — a phenomenon that Alexis Madrigal, then with The Atlantic, referred to a couple of years ago as “dark social.” In a nutshell, it was traffic — in some cases a substantial amount — that couldn’t be identified. According to an update from Madrigal and web-measurement firm Chartbeat, most of that traffic turns out to be coming from Facebook’s mobile apps.

In the original piece that Madrigal wrote for The Atlantic, he noted that according to analytical tools like Chartbeat, more than half of the website’s social traffic was coming from somewhere other than the usual suspects — namely, the top social platforms such as Facebook and Twitter — but where exactly it was coming from couldn’t be conclusively determined.

Facebook is more dominant

Many analytics programs identified this traffic as coming “direct” or from typed-in links, but Madrigal was convinced that much of it was coming via chat or instant messaging apps. When Chartbeat looked at its aggregated data across other sites, it found that in some cases “dark social” was accounting for more than 65 percent of a website’s traffic. Madrigal argued that this meant Facebook wasn’t quite as prominent as many media companies had thought.


As it turns out, however, Facebook is in fact even more dominant as a traffic driver than publishers originally thought — because much of the “dark social” or unattributed traffic to websites looks to be coming from Facebook’s mobile apps, including its flagship app, but also other apps such as Paper. For a variety of reasons, that traffic isn’t being identified properly in web analytics terms. As Madrigal put it in a post for Fusion:

[blockquote person=”” attribution=””]”The takeaway is this: if you’re a media company, you are almost certainly underestimating your Facebook traffic. The only question is how much Facebook traffic you’re not counting. The bad news is that, if you didn’t know before, it should be even more clear now: Facebook owns web media distribution.”[/blockquote]

[pullquote person=”Alexis Madrigal” attribution=”Alexis Madrigal” id=”898220″] “It should be even more clear now: Facebook owns web media distribution”[/pullquote]

As Madrigal explains it, when people read Facebook through a browser and click a link in their newsfeed, the browser sends data to the site whose link was clicked on — known as a “referrer” — which identifies where the user came from and when, among other things. But when someone is going through their newsfeed on a mobile device and clicks a link, they just go straight to the site, without any referrer being sent. So without a little extra effort and some detective work on the part of the site, it’s difficult to tell whether that’s Facebook or some other form of “dark social.”

Dark social = Facebook mobile

In a blog post at the Chartbeat site about how it has changed its measurement tools in order to account for this phenomenon, one of the service’s engineers describes how mobile is the source of the problem, with more than 50 percent of the traffic coming to some of its clients’ websites lacking a referrer of any kind. It’s not just Facebook, either — mobile apps from Google and others also fail to pass along that kind of information (according to Parsely, another analytics service, Facebook is aware of the problem and is making some fixes).

Although there is no standard referrer data sent by the app, what Chartbeat and a number of media companies noticed when they looked more closely was that these apps sent “user agent” strings unique to the app, which made it easier to track down what was coming from where. BuzzFeed data expert and former VP of growth Dao Nguyen — who is now also the site’s publisher — said the company has been tracking UA strings for some time, and so has The Guardian, which has built its own Chartbeat-style tool called Ophan.

[tweet 540591736570281984 hide_thread=’true’]

As Madrigal notes in his post, when Chartbeat started pulling out the user-agent strings and accounting for those in its analytics — which it started to do this week — the amount of “dark social” traffic to The Atlantic fell by more than 40 percent, and the amount of traffic from Facebook rose by almost the same amount. In other words, the site’s dependence on Facebook is actually substantially higher than it already assumed it was.

That kind of dependence on a single site raises all kinds of issues, as we’ve described before, especially given how little publishers know about the Facebook ranking algorithm works and how changes will affect their content. So while it may be nice to know what some of that “dark social” traffic really is, the answer probably isn’t what most websites were hoping for.