Revcontent buys content rec engine Rover
By bringing this machine learning-powered detector of user interest in-house, the content rec service hopes to ‘own the interest graph.’
The Silicon Valley-based Rover, formerly known as Flipora, employs machine learning to automatically infer a person’s interests, based on observing over a few days what she reads through its engine and what stories show up on that user’s Facebook and Twitter feeds. Revcontent also analyzes a user’s behavior on the page to determine interest and engagement.
Rover founder and CEO Jonathan Siddharth, who is joining the Revcontent team as senior vice president of technology, told me that Rover content recommendation apps have over 40 million registered users, whose profiles are now brought to Revcontent.
Measured against other content categorization engines, he said that Rover’s was “five times more accurate” in choosing categories, even sub-categories, than competitors’.
He added that the engine’s interpretation of content is so granular it can automatically tell, for instance, “if a story is pro- or anti-Trump.”
This is not the first time Revcontent has used machine learning, but CEO and co-founder John Lemp told me that this is a much more advanced implementation.
This boost is also informing his goals. “Facebook covers a user’s social graph,” he said, “and we want to own the interest graph.”
Here’s a screen showing Revcontent’s content recommendations:
Based in Sarasota, Florida, Revcontent says it delivers about 250 billion recommendations a month for such publishers as Forbes, The Atlantic, Newsweek and The New York Times, reaching 92 percent of all US households.
Its competitors include Taboola, AdBlade and Outbrain, but Lemp said his company is now “more effective at predicting if someone will click” on one of the suggested stories than other recommendation services.
Rover had actually been integrated into Revcontent for almost a year before it was acquired, he said, adding that the result has been a 90 to 100 percent lift in eCPM rates.
Although Revcontent’s and other services tout their ability to recommend content of one kind or another, much of the “content” is actually an ad. Lemp told me that his company’s service is beginning to tilt more in the direction of editorial content.