Viral math

Curiosity-gap headlines are a way of hacking real-world friendships for profit—one Facebook won't long abide

This chart, from Newswhip via Derek Thompson, has been doing the rounds, and causing a bit of debate:

The question: What on earth is Upworthy doing so right? How is it that Upworthy’s articles shared a good order of magnitude more often than anybody else’s?

Part of the answer is that Upworthy simply doesn’t publish that many articles overall — a couple of hundred a month, each one carefully and laboriously optimized, through extensive A/B testing, to be as socially infectious as possible. But that doesn’t fully explain how Upworthy’s articles can be so much more viral. For that, Upworthy needs the help — either on purpose or inadvertent — of Facebook.

Facebook is the monster in the publishing room: a traffic firehose which can be turned on or off at Mark Zuckerberg’s whim. Right now, it’s turned on, and while a lot of sites are feeling the love none is doing so more than Upworthy. (Except, maybe, ViralNova.) So, how does Facebook give Upworthy such a big boost?

Let’s start with the basic mathematics of virality. Start with an article, any article; let’s stipulate that it gets 1,000 pageviews, naturally, just by dint of being published on a certain website. Now, let’s say that 1% of that article’s readers decide to share it with their friends, and that each reader has 100 friends. That means 10 people sharing, and 1,000 new people seeing the link. How many of those people will click the link? Let’s say it’s 10%. Which means that the article gets a boost of 100 new pageviews. Those extra pageviews cause their own viral loop, which generates an extra 10 pageviews, and that’s where the cycle pretty much peters out. Thanks to sharing, the article has been viewed 110 times, over and above the original 1,000 pageviews.

This requires a formula. Call the basic strength of the website PP, for publisher power: that’s the number of pageviews you can expect to get when you publish an article on your website. You then multiply that by S, or shareability: the likelihood that a reader will share your article on Facebook. That in turn gets multiplied by F, or the number of friends per reader, and then by C, which is the clickbaitiness of the headline.

The key number here is S·F·C, or shareability times friends times clickbaitiness. In our model, that’s 0.01 * 100 * 0.1 = 10%. If you increase any of those numbers — if you make people more likely to share your article, or more likely to click on the headline — then you’re going to increase the virality of the piece. For instance, if you double the proportion of people sharing the article and also double the probability that someone is going to click on the headline after seeing it, then S·F·C becomes 0.02 * 100 * 0.2 = 40%. If you start with 1,000 pageviews, then you’ll get another 400 viral views which in turn will generate another 160, and so on: your viral boost goes up from 110 views to 660 views.

You can see that a relatively small tweak to the variables in the S·F·C formula can make a very big difference to your total pageviews. Pretty soon you can double your initial pageviews, or treble them — and, then, when S·F·C exceeds 1, you achieve escape velocity: your article just keeps getting shared more and more and more. Getting S·F·C > 1, then, is the goal of all would-be viral content, and it’s by no means impossible: if 5% of an article’s readers share it, and those readers have 200 friends each, and 25% of people who see the headline click on it — well in that case, S·F·C is a whopping 2.5, or 250%.

At those levels, it almost doesn’t matter what PP is — how many pageviews you seed your article with before it goes viral. PP still matters, however — which is why so many viral sites have pop-up boxes which try to harvest your email address. It turns out that emailing lots of people with links to new content is a great way to start the ball rolling.

But there’s a fly in the ointment, here — something which makes achieving escape velocity much more difficult. Let’s call it FBT, for Facebook Throttle. If you share an article on Facebook, and you have 100 friends on Facebook, that does not mean that your 100 friends are all going to see that article in their newsfeed. Far from it. After you click “share”, Facebook then decides whether the article you just shared is going to appear in your friends’ feeds or not. (This is a very big difference between Facebook and Twitter, which shows you everything your friends are sharing.)

As a result, the important formula isn’t S·F·C; rather it’s S·F·FBT·C, where FBT is the probability that the article you’re sharing is going to actually appear in your friends’ feeds. In recent months, Facebook has been taking its foot off the throttle quite dramatically — but no one knows how long that’s going to last.

Which brings me to Upworthy. We know that Upworthy spends a lot of time optimizing for maximum S and maximum C. It more or less invented the “curiosity gap” headilne, for instance, which turns out to be a great way to boost C. In other words, Upworthy is maximizing the variables under its own control.

What’s less well understood is that there seems to be a direct correlation between C and FBT. While Facebook controls its own throttle, it does so in response to user behavior: it wants to show its users more of what they want to see, and less of what they don’t want to see. And it’s easy to tell what Facebook’s users want to see: just look at what they’re clicking on. As a result, there’s a direct feedback loop between C and FBT: the higher your clickbaitiness (C), the less that Facebook will throttle you, and the more likely that your articles will be seen by your readers’ friends.

To put it another way: at the moment, Facebook assumes that people click on exactly the material that they want to click on, and that if it serves up a lot of clickbaity curiosity-gap headlines, then it’s giving its users what they want. Whereas in reality, those headlines are annoying. Curiosity-gap headlines are a bit like German sentences: you don’t know what they mean until you get to the end, which means that the only way to find out what your friend is saying is to click on the headline and serve up another pageview to Upworthy. (Or ViralNova, or Distractify, or whomever.) It’s basically a way of hacking real-world friendships for profit, and there’s no way Facebook is going to allow it to continue indefinitely.

All of which is to say that the massive advantage which Upworthy has, as seen in the chart at the top of this post, is certain to go away. It’s a temporary phenomenon, a function of the fact that Upworthy is better than anybody else at turbocharging virality by using artificially-optimized curiosity-gap headlines as a way of sending a (false) message to Facebook that those headlines are the stories its users really want to read. Upworthy’s formula will work until it doesn’t. Which is why I think that Dennis Mortenson is going to win his bet against James Gross.

Has America ever needed a media watchdog more than now? Help us by joining CJR today.

Felix Salmon is an Audit contributor. He's also the finance blogger for Reuters; this post can also be found at Tags: , , , ,