How they know what you like before you do

The other night, a few friends sat in Tracey Kennedy's Rock Island, Ill., living room listening to music. A song by a band no one but Ms. Kennedy knew started to play, and everyone wanted to know who it was.

Kennedy revealed that it was Silversun Pickups, an under-the-radar Los Angeles band she'd found using an Internet music service called Pandora.com. For her, the website's personalized music recommendations have sparked new listening habits. "It's like I've come back to life," says Kennedy, a 30-something computer programmer. "I'm getting all these vitamins I need." Since she started listening to Pandora at work in late October, Kennedy has bought about 35 new albums.

That's music to the ears of those who make recommendation technology. By 2010, one-quarter of online music sales will be driven by such "taste-sharing applications," predicts a study released in December by the Berkman Center for Internet and Society at Harvard Law School and research firm Gartner.

Over the past decade, e-commerce has taken a cue from the notion that friends give the best recommendations. Personalized suggestions have become more commonplace as various forms of media converge, industry professionals say, and this could both change the entertainment industry and give consumers more power.

What started with Amazon.com's "collaborative filtering" approach, which made product suggestions to consumers based on what they bought, has become a more precise science.

Kurt Beyer, president of Riptopia, a digital media processing company, divides recommendation technology into two general schools: theoretical and empirical. The theoretical approach bases its recommendations on qualities inherent in a product. The empirical approach is similar to what Amazon.com does, gathering large amounts of data about the buyers of a product to make recommendations based on demographics and interests.

Recommendation technology is "exploding," claims Daren Gill, vice president of ChoiceStream, a Cambridge, Mass., company that powers recommendations for AOL, Yahoo Movies, and eMusic, among others.

ChoiceStream makes recommendations based on about 25 attributes, such as "macho," "romantic," "mainstream," and "obscure." Eight editors monitor the technology to make sure that when new music or movies arrive, the automated system places them in the appropriate category. Then algorithms create recommendations for users based on their previous choices.

MusicStrands, a free online music service based in Corvallis, Ore., launched last year and is working to make "music discovery" a social activity. Last week, the company rolled out a new version that lets users see what their friends are listening to in real time.

"They don't want to sit down and listen to what other people are programming for them," says Gabriel Aldamiz-echevarria, MusicStrands vice president, in a telephone interview.

With a library of more than 5 million songs, MusicStrands provides instant recommendations based on what someone is listening to at that moment. Listeners can build and share playlists and "tag" music with terms such as "contemplative" or "driving."

Page: 1 | 2