At some point in the spring of next year, a San Francisco startup is going to start demoing a product that’s designed to predict the future.
Today, Klout is perhaps best known as a social media monitoring firm. Its “Klout Score,” which measures how important or influential a Twitter account is, has become one of the most popular metrics of its kind, as more and more users look to expand their reach on social media.
But the company is trying to take that concept a little further. Instead of measuring how big a user or topic is on social media right now, Klout is trying to develop algorithms that predict social media trends – which seemingly irrelevant Tweets, videos or web pages are viral hits in the making.
“Predictive social media is really interesting to us,” says Joe Fernandez, chief executive officer of Klout. “One of the things we want to understand is who are the people that are fire starters on a meme or a message – who are the 50 seeds you need to give the message to in order to have it spread?”
There’s an argument to be made that, in the online world, 2012 will be the year of the social soothsayer.
The concept of using massive amounts of user-generated data to predict events and outcomes isn’t new. Google has been consistently successful at predicting flu outbreaks before anyone else, including the U.S. Centres for Disease Control, simply by monitoring its users’ searches for flu-related terms. It turns out that aggregate web data do a very good job of catching major trends before they become obvious.
But until recently, much of that work in predictive analytics was confined to very specific cases. Now, everyone from rock bands to movie studios to apparel makers are looking for ways to predict the success of their various products and campaigns before the fact – and they’re willing to pay millions for the technology.
The problem is, predicting the next viral video or meme is very, very difficult. For one thing, such phenomena rarely start out somewhere obvious, like an influential user’s Twitter feed. Instead, they tend to bubble up on more obscure forums and message boards – a link from an influential web personality may make such an item a big hit, but the trick is in figuring it out before that link goes out.
At Klout, two teams are working on the company’s predictive analytics tool. One team collects as much data as possible from millions of various feeds – Facebook pages, Twitter accounts and blogs, among others. Then a science team staffed by experts in topics such as network theory and natural language processing combs through the data, looking for patterns.
Even though the work is still far from complete, Klout’s research has shown some interesting results. For example, Mr. Fernandez says the data suggest a user doesn’t usually react to a particular message, be it a link or a video or something else, until two or three other users in their social network become aware of it – essentially, until there are people to have a conversation about it with.
Klout is far from the only company doing this sort of work. Dozens of companies specializing in monitoring social media networks have begun setting their sights on predictive analytics. A firm called Meltwater, for example, has begun monitoring social sentiment to gauge how the 2012 U.S. presidential election might play out. At YouTube, researchers have spent countless hours trying to figure out what makes a video go viral. In 2012, such efforts will only intensify, as more people and companies look to the Internet as the primary driver of buzz around almost everything.
“We all have something we’re passionate about,” Mr. Fernandez says. “This is about how to spread those ideas and messages the furthest.”