Pandora Founder Westergren Speaks at SCU

By Eric Goldman

I *love* Internet radio. I typically listen to it all day long while I’m pounding out emails, articles and blog posts (like this one). I’ve tried a number of services, including Accuradio, and However, now I listen exclusively to Pandora, which in my opinion is the best of the bunch. Pandora really does get better based on the feedback I give it, which gives me enough incentive to customize it to my tastes. (I won’t embarrass myself by exposing my idiosyncratic music preferences here, other than to say that I don’t think I’ve seeded Pandora with any band that started after 1990. Hey, I’m a product of my time.)

In any case, when the High Tech Law Institute was contacted by the Santa Clara University’s Center for Innovation and Entrepreneurship to co-sponsor a talk by Tim Westergren, one of Pandora’s founders, we jumped on the opportunity. Tim gave an outstanding talk last night to an enthusiastic audience of at least 200 people.

Pandora was founded in 1999 and raised $1.5M in early 2000. That money ran out in 2001 when no more VC money was available. For the next 2 1/2 years, the company went on 300+ VC pitches with no luck, but was able to run on fumes in large part by mostly deferring employee salaries (a total of $1.4M deferred). This is an amazing feat for a number of reasons, including the fact that most employees need to put food on their table, plus I don’t think the labor laws really tolerate employee salary deferrals. In any case, in 2004, the company raised new money and revamped the business model. In October 2005, the company launched a consumer-oriented Internet radio, and the rocket ship took off.

This roller-coaster ride was partially tied to the company’s search for a viable business model. Pandora first thought it would be an online music retailer (typical thinking circa 1999). Then, it focused on B2B licensing of its services to other online retailers (typical thinking circa 2001). By 2004, there had been enough external changes to contemplate a consumer radio service–broadband had become more widespread, and the webcasting license fees were set at a more palatable level. In their October 2005 relaunch, they first started with a consumer subscription model but quickly realized the futility of getting people to pay for Internet radio. Now, they are effectively ad-supported, and Tim thinks that they can reach profitability at the end of next year assuming the webcasting royalty rate will be reset. (After the tribunal issued the new webcasting rate, the company held a board meeting to decide if they should just fold up their tents and give the remaining money back to investors–instead, they pushed for a listener grass-roots campaign, which was wildly successful).

Pandora’s main competitive differentiator is its “Music Genome Project.” 50 trained musicians with at least a college degree in music (called “music analysts”) listen to songs all day long and rate each song on 400 different musical attributes. See the 2005 WSJ article discussing them. By profiling songs this way, the system can predict that a person who likes an artist’s song might like other songs with similar musical attributes. From listening to Pandora for many, many hours, IMO the system isn’t perfect, but it does a pretty good job, and it has definitely hooked me on music that I wouldn’t have listened to otherwise.

However, the human capital required to build this database is significant and expensive. Sure, the supply of people with a music degree willing to be paid to listen to music all day long should be favorable, but still, a low salary multiplied by 50 people is still a big number. As Tim acknowledged, this is the opposite of scalability–a song can take up to 1/2 hour to catalog–which reduces Pandora’s ability to comprehensively catalog the “long tail.” On the other hand, they already have a database of 500,000 profiled songs, and they are adding 15,000 songs a month. Plus, having found a way to survive while building such a large database, the database is now a big barrier to entry, because any competitor seeking to take a similar approach (patents permitting) would have to incur serious upfront costs to replicate a competitive system.

Pandora does a little search engine marketing, but principally they rely on viral marketing–build a better product and let the customers evangelize it to their friends. They claim 8.5M registered listeners, growing at the rate of 500k new registrants per month. Listeners interact with Pandora on average 7-8 times per hour to give feedback or configure things (that sounds about right from my personal experience), which is remarkable as a stickiness measure. The result of this interaction is a database of 1B thumbs up/down opinions, a gargantuan database of user preferences. Tim said that Pandora uses this database for some collaborative filtering, but it sounds like this database is could be a globally important resource if made available more widely. (I’m sure the privacy folks are running through all the ways this data could be misused.)

A couple of other interesting factoids:

* at peak hours, Pandora’s traffic represents 1.5% of global Internet data

* they block International users because of the lack of Internet webcasting statutory licenses parallel to the rights enacted in the DMCA

* 94% of their database of 500,000 cataloged songs are played every day–powerful evidence of the long tail effect that consumers will enjoy otherwise obscure content if the transaction costs are low enough

* Tim said Pandora’s biggest competitor is ClearChannel. Clearly, they want to own the entire radio industry, not just the Internet radio market.

Check out Pandora and see what you think. If you want a jump start and you promise not to laugh at my tastes, I can email you my heavily customized stations.