Thanks to its recent forays into curated content, particularly around topics like the Olympics, it’s become obvious that Twitter wants to be more than just a utility for distributing links: it wants to target users based on the “interest graph” that it has constructed around them based on their activity. Prismatic founder Bradford Cross says that’s what he wants to do as well, and he thinks his startup is actually in a better position to do it than Twitter. On Tuesday, the service launched what appears to be a direct attack on the larger network by giving users the ability to follow each other on Prismatic, but Cross says this isn’t an attempt to become a social network — he just needs more data to fine-tune the service’s recommendations, and turn it into the ultimate serendipity engine.
Until now, Prismatic has based its filtering and news recommendations primarily on a user’s activity on whatever social networks they wish to use to connect when they sign up: namely Twitter, Facebook or both. The service mined all of that behavior, including retweets and likes, and fed the data into its recommendation algorithm to produce a stream of content targeted around interests — and users have been able to “follow” topics and sources such as a newspaper or blog. But Cross said something has been missing, and the last piece of the puzzle arrived Tuesday, with the ability to follow specific users. That following data will make it even easier to target content that is relevant, he said:
With most services it’s usually a pain in the ass to do this kind of stuff — there’s either lots of noise, or you have to spend all this time configuring it in order to get it to work right. We’ve got it working pretty well, but the addition of people allows us to inject even more serendipity and a sort of editorial voice into your stream.
Cross said that the addition of people-following will provide even more signals the Prismatic algorithm can use to recommend content, and the goal is to provide as much serendipity as possible — that is, to show users content they may not have explicitly said they are interested in, but that fits their profile or interest graph. “People following gives us another dimension to connect across different interest areas,” he said. “It’s like the internet you’re not seeing on Facebook or other social networks.” Twitter is also clearly focused on the value of its follower graph: that’s what it recently told Tumblr and Instagram they could no longer access.
Prismatic connects to Twitter, but no longer relies on it
Prismatic used to base most of its algorithm on data from Twitter and other social networks, but as the service has grown it has built up its own database around what users choose to share with each other through Prismatic, and that has meant less reliance on Twitter. Cross — a data scientist who used to run a hedge fund based on semantic learning algorithms before he started the company — said that the percentage of users who share content on Prismatic is much higher than it is for Twitter, because “our relevance level is way beyond anything else that’s out there right now.”
Cross said that Prismatic has also been moving away from a reliance on Twitter data over the past six months because it became obvious that the larger network was starting to clamp down on what external services could do with its content, and he realized that posed a potential risk. “It’s just a business risk that you have to look at objectively,” the Prismatic founder said. “We look at it as an awesome partnership, where we get data and we share back to the network, but we aren’t relying on them.” Prismatic is also working with Facebook (s fb) to share more content through that network as well. Added Cross:
“If people want to play social-network war games, that’s fine — we’ll connect with whoever wants to connect with us, and we’ll be committed to all possible protocols you might want to use to distribute your stuff. We’re just going to assume that differentiation wins, and if we can’t build something that’s different enough from what Twitter or Flipboard or whoever is building, then that’s our own fault.”
Everyone is trying to solve the relevance problem, the Prismatic founder said, in part because “whoever gets this right is going to have a huge business, because there are some obvious revenue models” associated with being able to recommend relevant content in something approaching real time — hence the profusion of filters and aggregators such as Zite and News360 and even Flipboard, and Twitter’s desire to do the same. But social networks like Twitter and Facebook, he says, started with social connections and have been trying to add relevance filtering, while Prismatic has done the opposite.
“All we want is to be a matchmaker for people and content — like a giant relevance engine,” he said. “For us, it’s always been about relevance, and the social aspect is just another part of that.” Others may see the addition of following features as Prismatic getting into the ring with App.net and Twitter in some kind of social-network fight to the death, says Cross, but the startup has its sights on a much bigger prize. “We don’t want to just build some kind of copy-cat social network or some new social ghost town,” he said. “We want to completely reinvent the way media works.”