[Chartbeat] took visitors who showed up without referrer data and split them into two categories. The first was people who were going to a homepage (theatlantic.com) or a subject landing page (theatlantic.com/politics). The second were people going to any other page, that is to say, all of our articles. These people, they figured, were following some sort of link because no one actually types "http://www.theatlantic.com/technology/archive/2012/10/atlast-the-gargantuan-telescope-designed-to-find-life-on-other-planets/263409/." They started counting these people as what they call direct social.

The second I saw this measure, my heart actually leapt (yes, I am that much of a data nerd). This was it! They'd found a way to quantify dark social, even if they'd given it a lamer name!

On the first day I saw it, this is how big of an impact dark social was having on The Atlantic.

We've been thinking hard about The Dark Web -- the Web that's not visible to crawlers -- and Dark Social is a variation on that theme since it talks about socially shared links that are not publicly shared links.

Thanks for sharing Alexis's article, Rohit.

Will use Dark Social in PandaWhale's traffic analysis going forward.

I like this idea, but they really need to separate the traffic from their own email lists which has to be a very significant number for them.

Big data correlation! Someone would get rich off their profile id, email lists, and web logs.

Does "Dark Social" really exist?

Alexis writes: According to new data on many media sites, 69% of social referrals came from dark social. 20% came from Facebook.

So far on PandaWhale, we're seeing Dark Social as a much smaller percentage, usually between 10% and 20% for most articles.