Algorhythm and Blues

Peter Niessen isn't complaining. The 36-year-old director of strategic planning at American Express concedes he has a pretty nice life: a juicy job, a passport with stamps from Cambodia to Ecuador, and a slick Manhattan apartment with a roof deck.

Like many busy professionals, though, Niessen has paid for these blessings in a currency you can't stockpile in a 401(k): time. And as a former rock bassist, one of the most disheartening casualties of his maxed-out life is his ability to discover new music. Carless in New York, he hasn't listened to radio in five years. The lazy afternoons at Tower Records are long gone. "I used to depend on friends," he says, "but I can't look to them as much anymore. It's not that I value music less. I just don't have time to find it."

That dirge of the wage slave is one Tim Westergren has heard before. "People don't lose their love of music," says the founder of Pandora Media, a small Oakland, California, company. "They lose their ability to connect with it."

Westergren, now 39, built Pandora, an online music-matching service, to help the Peter Niessens of the world, the people whose CD-swapping (and beer pong) years are behind them. For $36 a year, after a free 10-hour preview, the company pledges to help them discover sounds they might otherwise miss--and to do it so precisely that its recommendations are better than an actual human being's.

Most matching services depend on other customers' buying patterns. Amazon's or iTunes', to name two, simply note that people who bought CDs by Death Cab for Cutie also picked up the latest New Pornographers' release. But Pandora's Web-based service (there's no download) is driven by a homegrown software package, the Music Genome Project, which maps the DNA of individual songs based on as many as 400 different elements, or "genes." Drop the name of a song or band into the search box and Pandora instantly creates a "radio station" dedicated to that artist, using shared attributes such as acoustic harmony or electronic instrumentation to create a string of songs you might like (see diagram on previous page).

When Niessen set up a station based on the Flaming Lips, for example, Pandora came back with Wilco and Mercury Rev--the usual suspects, given the Lips' sound. But the site's genius is its ability to go deeper as well, dredging up obscurities like the Great Depression and the Lightning Seeds.

Inevitably, Pandora produces some duds. "In between Wilco and the Flaming Lips, Lisa Presley popped up," says Niessen. "I was horrified." But that's actually part of the plan: Users can reject a selection at any time--and the genome rejiggers its algorithm and promises, red-faced, "never to play a song like that again." If you like a song enough to buy it, there's the inevitable link to iTunes and Amazon.

Pandora's a bespoke Internet radio station that learns from its own mistakes.

The result is a bespoke Internet radio station that learns from its own mistakes, one infinitely more diverse than the homogenized playlists promulgated by Clear Channel--and devoid of noxious ads or DJ chatter. Despite what its name might suggest, Pandora is nothing less than a mechanism for filtering and shaping the chaos of an exploding supply of digital music.

"The power of the Internet is that everybody can receive his own unique streaming content," says Joe Kennedy, the company's CEO. "Pandora has all the advantages of traditional radio in terms of speed and ease of use, plus surprise and serendipity."

The Creation Myth

"Serendipity" wasn't a word that got thrown around much in Pandora's early days. In fact, for a long time the company seemed to have been born under a dark star.

Like many ideas at the height of dotcom mania, Pandora was hatched when a flight of fancy met a couple of beers. In the fall of 1999, Westergren, a strapping Stanford poli-sci grad who'd spent 10 years as a keyboardist and songwriter in a rock band, was making a living scoring films--which basically meant pairing music with a director's personal tastes. Then, shortly before moving to Los Angeles to pursue his film career in earnest, he read a newspaper story about musician Aimee Mann; she had legions of loyal fans but not enough to satisfy her label, Geffen Records, which had shelved her.

"I thought, if I could do a kind of Myers-Briggs [personality map] for music and tell people what songs they'd probably like based on musical similarities, the Internet could solve the problem of access," Westergren recalls. "It would be like what eBay did for pink flamingos."

The year being 1999, it didn't take much more than a night in an L.A. bar and a couple of cocktail napkins before Westergren and Jon Kraft, a Stanford classmate who had already launched and sold one business, nailed down a $1.5 million round of angel funding--including a sizable chunk from Guy Kawasaki's Garage.com.