The New York-based startup demolishes conventional wisdom about the best ways to draw in readers--and snaps up $64 million in financing.
Online recommendations systems are usually based on the assumption that if you're interested in one thing, you'll be interested in similar things--a principle called "relevancy." That might work on shopping sites, like Amazon, but it turns out, according to a company called Outbrain, that it doesn't work so well for content sites.
The New York-based company is turning the whole publishing recommendation engine business on its head. Unlike many such engines, Outbrain doesn't serve up links based on how similar the topics of the new stories are to the one the reader is already reading.
It also eschews another standby in the business: serving up links based on the likelihood that a reader will click on them. Instead, the system makes its determinations based on how likely the reader is to stay engaged on the site once they click through.
It's an idea that's gaining traction. Outbrain's recommendation engine is now used on almost 1,000 brand publisher sites in the U.S. and Europe (as well as tens of thousands of smaller sites). And today, the company is announcing that it has closed a $35 million round of funding that brings its total financing to $64 million.
The company's chief insight has been that traditional recommendation strategies don't actually deliver the results publishers are looking for. Publishers don't just want readers to click on links. They want readers to stick around on the site, surfing from one story to another--and, of course, running their eyes across a bevy of ads in the process.
Serving up links based on relevance actually works very poorly, Outbrain founder and CEO Yaron Galai tells Fast Company. Unless the reader is a sports fan who wants to read obsessively about their home team, related links don't actually perform as well as links on seemingly unrelated subjects, but ones that Outbrain's algorithms have discovered somehow appeal to similar readers.
The relevance criterion is "one of the lowest performing in our network," Galai says. "It has hardly any meaningful impact on the content that we're serving."
Similarly, serving up sexy-sounding links also doesn't work very well, Galai says. Over the last few years, headline writers have become devilishly skilled at penning seductive headlines for stories that don't deliver the goods, simply to gather up as many clicks as possible.
But those strategies apparently are beginning to backfire. "When we focused just on clickthrough rate and not on how engaging the experience is after the click, our algorithms got very aggressive at serving links that are sensational," Galai says.
Readers, however, are becoming inured to the kind of link bait that promises much and delivers little. "They're getting good at becoming blind to anything that duped them into having a bad experience," he says. "There's only so many times you can dupe people into clicking on links they didn't actually want to consume."
So instead, Galai says, Outbrain's algorithms focus on measuring how engaged a reader stays on a site after they've clicked through--how many subsequent pages they view--and ranks their catalog of hundreds of thousands of stories accordingly.
Which leads to the second way that Outbrain is turning the content recommendation game on its head: The company's engine doesn't only recommend content on the publisher's site, the way, for example, the New York Times' "Most Popular" widget only lists New York Times stories. It also includes links to external sites.
For example, in the image below, the recommendations on the left ("We recommend") are for content on the host site (in this case, CNN's Health channel). The recommendations on the right ("From around the web") link to external sites.
The host sites don't mind, Galai says, because of the unique business model Outbrain has developed. The hosts get the recommendation engine for free. It's the external sites who fork over the cash to have their content listed (they pay on a per-click basis).
The system works for all parties, Galai says. Host sites--the 1,000 brand publishers along with another hundred thousand or so smaller publishers and blogs--get a recommendation engine that surfaces the content on their site that is most likely to keep readers engaged. (Galai says sites get a 5-10% lift in views.) But they also get a revenue stream when readers click over to other sites.
And the publishers who pay to have their links placed in the engine--which includes several hundred publishers--get their money's worth because Outbrain surfaces stories that have been proven to be correlated with longer stays.
Readers also win because Outbrain's system of measuring how long people actually stick around means that they end up recommending much higher quality content, stories, and sites that readers would genuinely be interested in.
Outbrain's system generates about 200 million monthly clicks, Galai says, which results in a total of 3.5 billion subsequent page views. He won't reveal the company's annual take, saying only that 2011 revenues will be in the "eight figures."
The latest round of funding was led by Index Ventures, and included investments by Outbrain's previous funders, Carmel Ventures and Lightspeed Venture Partners. Index partner Dominique Vidal, a previous CEO of Yahoo Europe, joins the board.
Galai says the company will use the money to expand globally as well as develop their nascent video and mobile recommendation products.
Read also: "Brains And Bots Deep Inside Yahoo's CORE Grab A Billion Clicks"
[Image: Flickr user Davichi]
E.B. Boyd is FastCompany.com's Silicon Valley reporter. Twitter | Google+ | Email