New visitors to www.pandora.com take heart in the site’s bold home page question, “Can you help me discover more music that I’ll like?” Once they take a few minutes to set up their own personal radio stations, the answer is “Definitely”.

The artful science of the Music Genome project has allowed Pandora to effectively analyze songs in a way that can deliver non-stop streams of music similar to a specified artist direct to a user’s desktop. This is just plain fun for the casual listener, but for the music supervisor it can serve as an important tool for discovering new, licensable music that they may not have otherwise heard about.

www.musicsupervisioncentral.com talked with Pandora/Music Genome Project Chief Strategy Officer and Founder Tim Westergren about Pandora and music supervision in this exclusive interview.  

MSC: When did you start Pandora, and why?

TW: I founded it in January of 2000. The inspiration for it was really from my own experiences as a musician. I was in rock bands for 8 or 9 years, then I was a film composer for four years. My passion comes from the world of independent artist and allowing them to connect with a more mainstream audience. That’s where I come from.

MSC: How is Pandora facilitating those connections?

TW: The way the system works is we do a detailed analysis of one song at a time. We have 35 people that work for us, musicians who listen and capture a song’s musical DNA. So the recommendations are done purely on what a song sounds like – the popularity of a tune is not relevant. It’s one of the only places you’ll find music not based on popularity. No one paid to have it put there, or because the aggregate audience likes it, a la the Amazon collaborative filtering model which is essentially a popularity contest. So one thing we’re good at is taking a song you’ve never heard before and putting it on the right playlist. I think Pandora is very unique that way.  

MSC: How does music get into the queue to get listened to in the first place?

TW: There’s a lot of ways that we acquire music. There’s a group of people here whose job is to find it. They scour the charts, magazines, indie mail order places, turning over every stone that is about trying to find music that has done something, somewhere for somebody. We’re not trying to play A&R, but trying to find music that has resonated with an audience somewhere. We have hundreds of thousands of songs. Since we launched this services in November of last year, we’re getting a tremendous amount of recommendations sent to us by people using the service. That’s becoming the thrust for us, with 200,000 songs in the first five months, as well as direct submissions from artists. Artists can see details about how to submit their music to us in the FAQ.

MSC: How long does it take for a song to get listened to once it arrives at Pandora?

TW: It’s probably about 2 or 3 weeks. We do about 8,000 new songs a month, and we get maybe four times that much on a monthly basis from recommendations. We can’t add everything we receive, but we do listen to everything. My attitude about song selection for the Genome project is that musicians have to earn it, and we’re looking for the best stuff we can hear. There’s about 400,000 songs total in the database now.
 
MSC: What is Pandora’s business model, and why are you optimistic for growth in 2006 and beyond?

TW: It will be ad-supported. You see a little bit of that on the site now, like the banner ads, and we’re hoping to get across the finish line that way. Maybe some audio advertising also, but it will be limited, like what you hear on NPR: ‘This hour is sponsored by…’ We’ve already sold some advertising to Honda, for example. That kind of size advertiser is very encouraging – this type of business starts to get big when the advertisers start to pay attention to it. That’s the tipping point.

MSC: What makes this work challenging? What’s been surprising about the process of building out the database?

TW: The challenging part has always been that it’s a very time-consuming method. Tackling something this big with people is kind of crazy. It’s one of the reasons we struggled for so long, because investors were skeptical about that as an approach. That was the big thing, convincing people that this was the right way to do it. Now that we have a critical mass, the skeptics are seeing the value.

Some of the surprising stuff – and something I always hoped would be true – is how strong an appetite there is among people for new music, especially as they get older. I think, there’s a perception that music isn’t as important to people, and I think that is sort of imposed upon people who don’t have a way of getting a satisfactory experience. Your CD collection gets stale, radio doesn’t speak to you anymore, and people get resigned to music playing a smaller part in their lives. With Pandora, you connect people with a tool that they can use to discover new music and they go crazy. People are starved for it. It’s like being reborn, you remember how you used to get so excited, like how finding a new band was like finding a new boyfriend or girlfriend.  

MSC: Do you consider yourselves music supervisors? Why or why not?

TW: I’ve never been asked that question, it’s an interesting one. I’d say ‘no’. I’d say we’re a resource for music supervisors. A music supervisor is someone who’s with a film or commercial, and they’re looking for a particular sound. What Pandora take is the instincts of a supervisor to find songs they know would work, and then find something else similar to that.  

A friend of mine who’s a music supervisor in LA told me the other day that they’re using it as a resource to find songs, and that was news to me. But it’s great to hear, particularly because I know a placement can become a source of income for musicians. If someone can get some money because of this, that’s great. Any artist would give their left leg to get synchronization monies. 

MSC: How do you think Pandora can be a useful tool to music supervisors working in the field for TV, film, video games and more?

TW: The most obvious benefit is that, as a supervisor, a lot of the stuff you’d like to use, you can’t because it’s too expensive. You can use Pandora as a way to easily find cheaper alternatives. Finding music can be an incredibly time-consuming process, but Pandora can make that quicker and easier. We’re just starting with leveraging the Music Genome, and in the future we’re going to allow people to specify in their playlists not just what kind of music they like, but what parts of the music they like – the solo, or the voice, for example. It’s more granular control, and music supervisors will go crazy over that, being able to drill down and get deeper access. I’m excited to see how they react to that.

MSC: What are the criteria that songs are evaluated on?

TW: There’s close to 400, which is an enormous number. It’s all the individual details of each aspect of music. We break music up into different buckets in a process called SHMRFT: Sound, Harmony, Melody, Rhythm, Form and Text. I think that was originally coined by a Berklee professor – each one of those aspects is broken down by its primary colors, so for harmony it’s not just tonality but the whole harmonic characteristics of the tune. Collectively, each detail of harmony becomes one of the attributes for understanding all the aspects of the music. There are over 30 attributes to describe the voice, including range, vibrato, melody, timbre of the voice, how it’s delivered. So whether it’s Tom Waits or Mariah Carey, they can all exist on the same continuum. 

MSC: What are the licensing issues with getting songs on Pandora? Which licenses do you need, and how difficult are they to get?

TW: Fortunately, a few years ago, in 1998, the Digital Millennium Copyright Act was established, and it allows Webcasters to do what we’re doing. We don’t have to have individual agreements with labels. We pay performance monies out to ASCAP, BMI, SESAC, but we can do that with one central agreement. We don’t need 10,000 contracts. We have a fixed royalty, and this was agreed on by the RIAA, as long as we don’t do things like play songs on demand, and there’s a limit on how many songs you can skip forward in an hour, or hear more than four songs by one artist in a three-hour period, things like that. So everything we do is perfectly legal and every artist is compensated, provided they’re a part of BMI, ASCAP or SESAC.

MSC: What genres are you strongest in? Where do you need to build out more?

TW: Certainly mainstream rock, alt rock – everything rock. Where we have to work harder is more niche genres like blues, country, certain genres of jazz. We did an enormous amount of back catalog and keeping up with new releases is not so much of a challenge, but we need to go deep and spread into new areas. And two genres that are totally lacking are classical music and world music, which is a very boorish way of saying anything not English-speaking. We’re about to release a great deal of Latin music into the database.

MSC: What do you think is next for Pandora? What would you like to see it be able to do next?

TW: I’d like to see it be mobile, so you can take it on a phone or a device of a car. I’m very keen to do international music, those are two big priorities for me. I’d also like to start seeing us empower the users more, allowing them to start tagging artists, contribute and aggregate information into the service. I’m excited to see that.

What did you think of this article? Email info@musicsupervisioncentral.com or post comments in the “Discussion” section.