The New York Times’ 6-month-old recommendations feature, which uses an algorithm to suggest stories readers might like, is “really exceeding our expectations in terms of usage and clickthroughs,” says Marc Frons, Times chief technology officer for digital operations.
Improvements are in the works for the recommendations engine, which looks at what users have read on NYTimes.com and suggests other stories. (TimesPeople, which uses recommendations from other people, is also getting a big overhaul.)
Why is it popular? Frons says:
“What is new about what we’ve done is how it figures out what to give you next based on what you’ve read. A lot of recommendation engines work kind of like the Amazon recommendation model — you bought a toaster, therefore maybe you’re interested in more toasters. …Our algorithm tries to figure out complementary or even disparate matches that will help expose you to what we think are things you would be interested in, rather than just topics.”
Like the Internet radio service Pandora, NYTimes.com readers in the future will be able to vote stories up or down over time to signal their interests. “We can get a better sense of what worked for you as an individual reader rather than just looking at your clickthrough patterns,” Frons said. However, he cautioned that most readers don’t want to spend extra time casting votes. “Active personalization is great for a small minority of our users,” he said. “The vast majority of people just want a great experience, and they want it done for them.”