Pandora's founder: We can make the Internet radio business work
Of the companies in this market, Pandora Media has the loudest voice. Pandora allows users to create as many as 100 free radio stations but does not let them pick specific songs. Using a "thumbs up-thumbs down" screening system, users can tell Pandora what they like to hear after inputting the name of an initial artist to narrow in on a genre of music.
With 30 million registered users and revenues tracking towards $40 million for 2009, according to PaidContent, Pandora could be the first truly profitable Internet radio company. That's quite a feat, considering Pandora's multiple near-death experiences. The company struggled to find a business model for years, then labored under what it believes are punitive licensing fees levied by music labels on Internet broadcasters.
Along the way, Pandora founder and Chief Strategy Officer Tim Westergren (pictured) has managed to raise $64 million (as reported by the Wall Street Journal) in venture capital and gain the ear of the U.S. Congress by marshaling a deafening grassroots campaign to convince Washington to lighten up on licensing fees.
It worked and Pandora is still standing as a result. I got a chance to interview Westergren in Pandora's spacious new Oakland offices overlooking San Francisco Bay. (I did almost trip over a drum kit as I wound my way back to the front desk). Here's an edited excerpt of the conversation:
DailyFinance: Is Pandora cannibalizing music sales?
Tim Westergren: The data says it's dramatically additive. By our surveys, 45 percent of Pandora listeners are buying more music than they did before. Only 1 percent of them are buying less. We sell $1 million worth of music per month on iTunes, Amazon and other sites and that number is growing. We are one of the biggest music affiliate referral sites on the Internet.
Can you afford to stay in business? You are paying higher license fees than broadcast or satellite radio for your music catalog.
We now have a survivable license payment rate. We can make the business work and build a good business on it. But it remains a fundamentally unfair system. We pay a far higher per use fee than broadcast or satellite radio does for the same number of plays of any given song. As we compete more and more directly, this licensing structure is getting increasingly anachronistic.
If you are driving in a car, to you its a song on the radio. But we have to pay 60 percent of our revenue in licensing to the music intellectual property holders. With satellite, the license rate is 10 percent and with broadcast the rate is 3 percent. The performer gets paid nothing in the broadcast world. It doesn't make sense. We're hoping we can fix that over time.
I know you have some users paying you directly to remove ads from their streams. How is the advertising business going? I personally never see the ads because I put the window to the back and just listen.
That's actually a typical response. Virtually everyone who uses Pandora says they do not interact very much with it. But in fact, the average user interacts just over seven times per hour. So you probably interact with Pandora more than you think. So we do things to ensure higher performance.
For example, we put in ads when and only when you are interacting with Pandora. Advertisers know that you are more or less engaged at that point. If Pandora is minimized, we're not running ads. We await until you skip a song or thumb a song. Thus far we have been able to maintain premium rates precisely because we can maintain high rates of engagement. Our advertisers say they are happy.
So when do you think a club owner or a band can start buying self-service ads on Pandora.com?
In my best case, some time next year. Two factors are impacting it. First, it's a matter of our priorities. We are in a big general expansion mode and we want to more completely saturate the mobile and automotive markets. We are on all the big handsets already -- iPhone, Blackberry, Palm, Android.
Adjacent to that are connected devices like the XBox, connected DVD players and televisions, things like that. It's a huge family of products. Our goal is to have a Pandora connection in all of those things. Then there's cars, where we are still in very early stages.
Tell me more about your expansion into cars.
We are really trying to expand our presence in the automotive space. Right now people can plug an iPhone into an auxiliary jack and run Pandora over the car stereo. The next step beyond that is to embed Pandora in the dashboard so that all the phone does is bring connectivity to the car and the car runs the app and it has dashboard controls.
The real big benefit of embedded versus on the phone is safety and user experience. You don't want people fiddling with the phone while they are driving.
Before you started this you were a musician, right?
Yes, that's right. I played piano since I was a little kid. I didn't become formally trained until college. My real deep dive into the training was after college. I spent 12 years as a musician and composer. That ranged from playing in cover bands for weddings to Holiday Inn piano bars to touring in rock bands. I was also a film score composer.
I really got into the technical side at Stanford. I didn't major in music, but I spent a lot of time studying recording technology. There was a place called Center for Computer Research in Music and Acoustics at Stanford. This was the hub for programmers and musicians. Outside school, I was a big consumer of all the latest digital music technology.
So you graduated and started playing. Then you launched Pandora.
Yes, I started it with two other guys, a computer scientist and a mathematician. They built the technology and assembled and managed the software engineers.
I'd read that your process for rating and scoring music is actually very subjective because it's basically people doing the ratings.
That's part of our Music Genome Project. And the answer is no, not really. We have a team of music analysts. Their job is to analyze music, slap on a pair of head phones and attribute by attribute assign scores to all the parts of a piece of music. They are highly trained and some have graduate degrees in music and theory. We put them all through a rigorous training process so that if two people rate the same song, they will rate it the same way.
But it's not truly digital then?
No, computers still do a very poor job of recognizing and assigning attributes to music. They can do some things like identify where a downbeat falls and tempo, but in general they struggle to categorize music in the way a human ear can.
How did you end up launching the business?
I initially thought we were building a recommendation technology we would license to other companies. There was a bunch of music sellers like CDNow and CDBaby. And there were these portals that were looking like they would become the music industry of the future. We said "Let's build this recommendation tool and license." We chased that business model for five years. It wasn't until the end of 2004 that we finally changed and built the radio model.
What made you switch?
We knew our business was failing. And we had a huge piece of intellectual property and were casting about for a business. A handful things colluded to make it seem like the right choice. One was broadband penetration. In 2004 it was skyrocketing. This made a huge difference for streaming audio on the Web. Also, we figured that broadcast radio has historically been a much bigger business than music retail so that meant a bigger market opportunity.
Then there was the rise of personalization, which we figured would collide with music. Our product is perfectly suited for this particular task -- putting together a string of songs that somebody likes. All those things put together made us realize this would be an interesintg opportunity. The question was whats the business model behind it? We initially launched a subscription service. People clearly thumbed that one down -- nobody wanted to pay. Quickly within a matter of weeks we had to figure out a free alternative. And so we switch to the current model.
Where do you see radio and the audio world going?
In the last year there has been a big sea change in the infrastructure surrounding radio. That's driven by the ability of cell networks to carry a streaming audio signal. That a phone can now stream Pandora has broken open what until now was a domain occupied only by commercial radio (satellite or broadcast) either in cars or homes. I think this new ubiquitous streaming capability is a hugely disruptive change that will put Internet radio right up alongside broadcast radio.
The ability to personalize Internet radio makes it much more attractive than playlists intended for 100,000 people. We are right at the moment when it's really taking off. Pandora adds 65,000 new registered uses per day. And over half come from a mobile phone. Our mobile listenership has ballooned in short order from nothing to 35 percent of total listening time. Half the people using Pandora on an iPhone are listening to it in their car. It's really amazing how its taken off.
As a former working musician, how do you think Pandora can help working musicians?
For a long time, the last 10 years, people have been talking about what the Web can do for the working musician. There have been myriad sites and services that have popped up designed to deliver help a working artist with little or no label backing and help them make a living. One of the things missing from that is scale and ability to target the information in a way that's useful for the artists. I think Pandora is well on its way to being able to combine those two. We'll be big enough to be able to bring to bear a huge audience in a really targeted way to help the musicians in our catalog.
So what's the next business you are going to start after this?
I'd really rather start a band and just play music.
Alex Salkever is Senior Writer at AOL Daily Finance covering technology and greentech. Follow him on twitter @alexsalkever, read his articles, or email him at firstname.lastname@example.org.