Interview with Tim Westergren, CEO of Pandora

Pandora, is an online radio service that allows someone to quickly and easily create a radio station that plays music they like. It’s powered by the Music Genome Project, which is something we’ve been working on for almost seven years now. The Music Genome Project is an enormous collection of songs that have been analyzed, one song at a time using 400 unique musical attributes, so it’s like musical DNA.

The way Pandora works is you type in a song or an artist that you like, and the Music Genome Project looks at that artist’s or song’s musical DNA and, in a sense, it goes looking for other songs in our collection, which is well over half a million songs now, at musical neighbors and starts creating play lists based on musical similarity.

There are close to 400 attributes, and these are essentially are all the little details of every aspect of a song that collectively gives that song its sound. It’s everything from the details of melody, harmony, rhythm and instrumentation to the compositional form, vocal performance or vocal harmony. Each one of those aspects of a song is broken down into a form of musical primary colors.

As an example, we describe every voice along a little over 30 unique attributes, and those describe, for example, the use of vibrato or falsetto, the range, the timbre and the delivery. Plus, any voice, whether it’s a gravelly Tom Waits sound or Ella Fitzgerald, can be essentially described along those same attributes, along different levels of each one.

It was funded by the venture investors, and continues to be, mostly in Silicon Valley. We started it seven years ago and we trade in different kinds of businesses. We tried to build a business on different ways of using the Music Genome. We did turn a profit, but it was only about two years ago that we decided to build the radio station that became Pandora. So, we were one of those struggling start-ups for many, many years.

They are all done using trained musicians; we don’t capture a single attribute using machines. Every song gets listened to by one of our music analysts, and these are people that typically have at least a four year degree in music and are all active, well trained musicians. They spend as much as 30 to 40 minutes doing a single song. They listen to it many times and they manually capture all those details.

They’re paid. The typical musician works somewhere between 20 and 25 hours a week, which is about how long you can really sit there on a daily basis and listen to songs, one after the other. They get paid an hourly wage and they also get health insurance, so it’s great gig for a performing musician.

Yes, it’s a crazy idea! We don’t plan to keep up is the short answer to your question. We do somewhere between 13, 000 and 15, 000 songs a month. We have a little over 40 musicians working for us. I would say that in North America alone that probably accounts for a third of the music that is produced in any given month. So, we’re already behind and falling further and further behind with each month.

Yes. We’ve been at this for seven years and we have built up an enormous catalogue. One of our first tasks, of course, when we did this was to go back and pick up everything from, basically, the beginnings of recorded music in the US–the 20s and 30s–any record that had reached any kind of public visibility from the very beginning, and did all that stuff. You’d be surprised, the volume of music, even if you take everything that’s ever been on Billboard since the 50s by itself is not too much because popular music only has room for a fairly narrow sliver of the overall pie. We’ve been filling out our catalogue and going back decades for years now in every genre.

Yeah, I’m a musician myself, and the reason I started this company is because I spent ten years playing in independent rock bands and trying to build a career. So I got very interested in trying to figure out how to solve the problem that faces every musician, every artist, really, but in this case, every musician. They’re trying to rise…they’re like a needle in a haystack. The web offers the perfect opportunity to help them find their audience, and vice versa.

We do pay a licensing fee for every song that we stream. We actually pay two different fees. One is a publishing fee, which is to ASCAP, BMI and CSAC, which are the main royalty collecting agencies for artists. We also pay a mechanical fee, which is a specialty category for webcasters like us, which also goes to artists by a different organization called Sound Exchange. So we pay for every incremental song or album that we stream. Our opinion is, we fully intend to make this a profitable, successful business, is to subsidize it through advertising. The visual advertisement you see on Pandora is all paid advertising. Our objective is essentially to sell enough of that advertising to support the experience.

It’s all headed in that direction. The service is only ten months old and we’ve been hiring, furiously hiring an advertising sales team, which is just starting to get up to speed. Their mission is to get advertising revenue caught up with our costs.

We looked all over the place to find music so they’ll type in a song style they like and if that’s not available then try to find out the music that would work. Folks who book concerts are using it to find opening acts, they’ll type in a headliner and see what comes next, and look for an independent band. So within the music industry, it’s got lots of applications that once we make that data available, it will be a really valuable resource for artists that are trying to see, for example, results by seeing the response to their music. It’s also I think good for educational applications, there are quite a few teachers who are using Pandora now in their classrooms, and they are using it for all sorts of different things, but certainly to teach music theory, and I imagine over the years that application will only grow.

In some ways I’ve always been an entrepreneur, as I was a self-employed musician. We did actually do some preparation. I was a film converter for a while too. I had no idea, what I was starting. I just had the idea for it, and was watching everything that was going on online with the surge in online music sites, and what seemed to be this really rapid transition toward online music. I was living in San Francisco in 1999, which was the heyday of the dot-com era, and I shared the idea with a friend, who had already started and sold a company successfully. He just said “What the hell? That’s really interesting, let’s try it.” Before I knew it, two months later, we had a business plan. We were working out of a one-bedroom apartment, looking for seed capital. It wasn’t really planned out.

I think one of them is as an entrepreneur you need to think about the sustainability of your life. You need to create an environment where you can remain focused on your business and enjoy what you do, but not be in a position where you are worried every day about your survival. There’s an enormous amount of stress on you, just the general things of every day life. Because the amount of work that goes into a start-up is basically endless, everything else in your life needs to be reasonably on track. It will really absorb all of your energy, everything that you have to give.

Other than that, it’s finding something that you really, really enjoy doing so that the work itself doesn’t feel like work. You have to work 100-hour weeks, and you have to do it month after month after month. You need to be in a good sort of place in terms of enjoying your work and being secure to be able to do that.

I think also you need to find a good partner. I think that it’s incredibly hard to do this with people, and I can’t imagine having done this without partners. I actually founded a company with two people, one of whom I knew very well and the next was a friend of his. So, it was two guys I had a lot of respect for, who were very smart and were very good at different things than me. It was really an important support structure to have.

Well, one of them was technical, so he was our founding chief technology officer. I had no technical training, so he managed the whole software development part of what we were doing. The second person, who I had known from college, brought a lot of experience in just starting companies. He headed us up and was able to short cut all the typical learning that you have when you get going: how many people, building a strategy, getting office space, health insurance, incorporating all the details of setting up a company and thinking about organizational structure and all those things. He had all those skills already, so we were spared a lot of hard lessons.

I’ve never really quite thought about these questions. I’ve had a very particular kind of experience because we founded this company in March of 2000, literally three weeks before the collapse of the whole dot com bubble. In my particular experience with Pandora – we actually weren’t called Pandora when we started – but this company has been providing a long, extended physical track.

I think the reason I started it is I am personally very compelled to do things that I really want to do. I am restless doing jobs where I’m not in control of my own destiny where I can do what I want. I’m not a person who is used to routine or working for somebody. I like the excitement and improvisation that goes along with entrepreneurship. It’s fabulously rewarding when and if you can make it happen.

I feel that all the things we went through were necessary parts of what we became as a company. Sure, I wish I could have waited four or five years to start the company, and not started it right when everything collapsed. At the same time, I think the reason our company is in such a good position now is because we did survive and built this underlying asset through the Music Genome Project over the years. We wound up with a very sensible and valuable piece of intellectual property. I think that we went into all of the complications of pre-financing and refinancing and all those rounds that I would have liked to have known more about that before going in, but at the same time, I think if I really knew a lot about that, I wouldn’t have started it.

I think there is sort of a healthy naivet? that I think an entrepreneur needs when they get going. You really do not want to know the odds, because you would never start. I cannot say that I would do a ton of things differently, to be honest with you.

I believe it was a handful of things. Most importantly were the individual contributions of the employees. A lot of people in the company, maybe 35 or 40 people, at various times worked with no salary. Some people, myself included, worked for almost two years without getting paid. That was an extraordinary fact, not for me as a founder, but for the employees it is extremely unusual.

I think that the reason they did that was because they believed that in all this madness there was a good idea that was eventually going to have it’s day. They were able to inspire people. It is a combination of convincing and inspiring people to believe that eventually this leadership team was going to get them there. If they just grabbed an oar and leaned into it, that this is a team they can really count on. That belief was tested to the absolute limit.

I think we were really able to inspire them and lead this team. I think that we never had enough money to get drunk on money, so we stayed small. We got small when we had to. We were able to sign just enough little licensing deals to give us hope. We had enough good deals along the way to keep us going.

We are really ambitious with our plans. I have to be nervous. I actually think that we can radically change the whole business. What I hope among other things is that we will create a musician type class.

That would essentially mean getting exposure to this huge seed of independent musicians who are typically completely ignored, who don’t have any kind of outlet. In actually helping them to find an audience, in this case not a mass audience, but a large enough audience to help start their life.

I think that will mean not only transforming the whole vocation of being a musician, but on the listener side, we can bring music back into classes where people get disconnected from it. As we get older, we loose interest in music, or loose their connection to it. Their CD collection goes stale. It is a terrible, terrible loss for people. I think we can change that on a massive, massive scale.

Sure. We are actually looking at a partnership with EarthLink and Music.com and TicketMaster and all the music guys. We are definitely open to finding good, mutually beneficial partnerships.

Yeah, we have been talking about that for a long time, and a lot of listeners have expressed an interesting in contributing to that or if that is something that we do. There were certainly on top of mind for us. One of the challenges in this particular pursuit is that the stuff that we are measuring, the musical attributes, have an inherent subjectivity to them. It required a lot of varied, concerted efforts to create a consistent way to, for lack of a better word, scoring these attributes.

Our experience with having people analyze music that are not trained, the data that we get from that is normally really not possible to parse. It would be too difficult to get consent. I think that the contribution that users can make is in terms of tagging is the collection of extra meta filters on top of everything.

We do collect all of the thumb feedback from all of our listeners. We have close to 300 million thumbs up or thumbs down. That is kind of the aggregate effect of the audience. If a song has a particular relevance on a given station, the audience is going to essentially correct or enhance [the playing station?]. So we are taking kind of collaborative feedback of our listeners to help things out, but on a daily collection side, it is a little more tricky.