Inside Yahoo’s Content Algorithm

Yahoo is betting on its Content Optimization and Relevance Engine algorithm (C.O.R.E.), which powers its Today homepage module, to reignite positive buzz around the company’s services. Yahoo’s C.O.R.E module, via the Today homepage, publishes 45,000 unique versions of the Today module every five minutes. According to Yahoo, the company is pulling 1 billion clicks per month on the module.

“With all of the content and media available to you on the Web today, personalization makes a lot of sense,” said Raghu Ramakrishnan, chief scientist for search and cloud platforms at Yahoo. “It connects people to what they care about most. All publishers should be looking at ways to filter, aggregate and tailor the content they show their readers to some extent.”

C.O.R.E., said Ramakrishnan, was born of a collaboration of Yahoo Labs and their engineering team. The blend of humans working behind algorithm-driven machines allowed Yahoo to create content that drove significant CTR increases, Ramakrishnan believes.

Higher engagement rates are Yahoo’s goal according to Ramakrishnan, and “that’s exactly what advertisers want too.” “Higher engagement leads to more time spent, repeat visits and clicks, and additional ad inventory,” said Ramakrishnan. “It offers advertisers quality places where they can advertise alongside interesting content that users like. And in some cases, clients can also benefit from these insights, crafting unique, customized programs that are likely to appeal to their target audiences,” says Ramakrishnan.

The CORE algorithm determines which stories will attract which specific audiences, according to Yahoo. Yahoo creates a profile for each user based on information that they’ve volunteered if they’ve registered with Yahoo. Based on this information, C.O.R.E. weights a selection from the 50-100 packages of content that include headlines, links and text  that Yahoo’s editors possess, according to its potential appeal to an audience.

“For instance, if you’re trying to reach young men, you might be surprised to learn how interested they are in dating stories that are ostensibly aimed at women,” said Liz Lufkin, Yahoo’s vp of front page programming. “Older men are more interested in female celebrities than you might think. And women actually do care about sports – just certain kinds of stories.”

Yahoo’s  C.O.R.E. blends a mixture of human an algorithmic intelligence to create a human-monitored content strategy that allows editors to override algorithmic insights when necessary.”We’re using the insights gained from C.O.R.E. dashboards to help further refine our programming on both the homepage and Yahoo!’s media properties,” said Lufkin. ” Essentially, our sites and content pages for individual stories are becoming launching pads for all kinds of related content for people to explore, which is a great driver of further engagement across all of Yahoo!’s network.”

According to Lufkin, Yahoo hopes to leverage the personalization instincts of their 680 million users to inform the shape of the company’s content strategy.

“On the science side, there remains a tremendous amount of exciting work to do,” said Lufkin. “We’re working on adding other data into the algorithms, like social signals, and are looking at ways to create interest graphs for our users so we can help them discover relevant content related to the topics they’re most interested in, like golf or Madonna and build those into their experiences across Yahoo. We’re exploring ways to optimize for more than just click throughs, with the goal of maximizing metrics such as revenue, user engagement and long-term loyalty. We’re also extending the range of settings in which C.O.R.E. can be applied, and combining it with more traditional search ranking and ad targeting techniques.”

That reach into targeting using the C.O.R.E. algorithm means that Yahoo will be expanding its data-driven strategy into all segments of Yahoo content, including Yahoo’s news section.