Creative Generalist Q&A: Tim Westergren
Tim Westergren is an award-winning composer, an accomplished musician and a record producer with 10 years of experience in the music industry. He has recorded with independent labels, managed artists, scored feature films, produced albums, and performed extensively. His main instrument is the piano, but over the years he has played the bassoon, drums and clarinet and his musical background spans such genres as rock, blues, jazz and classical music. Tim is obsessed with helping talented emerging artists connect with the music fans most likely to appreciate their music. To that end, in 2000 he founded the Music Genome Project and, more recently, its popular online offspring Pandora. Tim serves as Pandora’s Chief Strategy Officer.
So what inspired you to start the Music Genome Project?
Well, I’m a musician. I’ve been a musician since I was a little kid. When I graduated from college I spent about 10 years playing in rock bands and so I kind of had the firsthand experience of being an independent musician and trying to find an audience and sort of build a career – and how hard that is. And so I really got interested in trying to solve that problem. Music has always been a kind of feast or famine business and it’s essentially a reflection of the economics of the music business. It’s not happening because there aren’t a lot more talented musicians or because there aren’t people who would buy and go see a wider range of artists, it’s just that it’s so hard to connect audiences with lesser-known musicians. And really, the primary inspiration for the Music Genome Project was to solve that problem. And I kind of came to it through this method of doing it through my own experiences recommending music to other people. Not only informally to family and friends but also I was a film composer for a while and I had to figure out what a film director wanted when they gave me a dozen songs they liked. And so I sort of had this idea of a genome in my head informally for many years and it just started to become clear at the end of ’99 with the internet boom – and of course online music had so much energy at that time – and I just had the idea to sort of codify it somehow with this online system.
Has anyone from the genetics, biological genome project taken offense to you commandeering their terminology for music?
I did give a lecture at the NIH a little while ago which was kind of fun but no I haven’t had much interaction with the biologists except for that.
Can you describe the inner workings of the Music Genome Project? How is it set up and what is its main goal?
Sure. The genome itself is a sort of a very large musical taxonomy. It’s a collection of about 400 musical attributes that collectively essentially describe a song, that altogether comprise the basic elements of a song. So it’s all the details of melody and harmony, rhythm and instrumentation, and form and vocal performance – a song broken down into all the most basic of parts. It’s a bit like musical primary colors in a way. And we built that over about a year and then we have a team of musicians that sit down and listen to songs one at a time and analyze each song along these attributes. So they literally score each attribute one at a time based on their musical breakdown of the song. So it’s a bit like musical DNA. And then once you have this kind of fingerprint for a song, this musical fingerprint which is all done by human analysis, you then can calculate how close two songs are together by comparing the scores on these genes on which they’re all described. So we have a mathematical algorithm that takes those numbers and turns it into a proximity measure and that’s what we use to generate playlists.
And the main goal is really, back to my own experiences, this was a way for you to type in a song or an artist that you know and like and be recommended music that you’d never heard of before. So, it would be a way of giving a chance, an opportunity to musicians who don’t already have an audience to be introduced to folks who like music like theirs. And because it’s based on musical attributes it doesn’t need anything more than that. It doesn’t need a purchasing history. The artist doesn’t need to be popular. It just has to make sense musically in the context of that better-known artist. And that’s how I think the Music Genome can really be an effective tool for surfacing a broad catalogue of music.
Pandora’s not an overnight success. You’ve been at it a while (since January 2000). What have been some of the biggest lessons to come from growing a company of this kind?
Gosh! All the hours?! Well, so we founded the company a couple months before the dot-com bust happened. We raised our first round of investing literally a couple weeks before the bottom fell out of the dot-com boom. You know along with that went everything else. The music business entered a very chaotic few years and so we wound up coming of age in a terribly difficult time to finance and build a company. I think timing is something that is really hard to control of course. Very hard to predict. It’s like buying a house. You never really know if you’re buying at the top, the middle, or the bottom of the market. And so it’s hard in hindsight to say, “oh you know you should have waited”. But I think probably what I learned most of all in the course of those years is really how to manage a company, how to recruit people, how to hire people, how to oversee a team of individuals and keep them motivated and productive and inspire them through adversity. And be really, really disciplined about how you spend your time and how you define your business and really being very careful to be efficient and to be really honest with yourself. To be very rigorous in all the decisions that you make with a small company because when you’re small it doesn’t take very much to trip you up and make you fall flat on your face. I learned so much about putting together a company and keeping it together and then sort of adapting to change. There’s so much change.
I was interested to learn of the town halls you hold with Pandora fans at the various places you travel. I would imagine you also deal a fair bit with record labels and music industry people. From that unique and peculiar in-between vantage point, how would you describe the state of music today?
Well, it’s kind of schizophrenic. I think that, as an industry, it’s struggling. The traditional music business is under a lot of pressure from lots of different sides and it’s entrenching and it’s consolidating. It’s definitely shrinking just like traditional industry. I think the business of music though has really never been healthier, ironically. And I think we’re really in the early phase of a complete transformation of the business. I don’t think we’re going to recognize it five years from now. You know, I think if you’re prepared to adapt and be creative and sort of reinvent yourself, there’s an enormous amount of opportunity and I think the next version of the music business is going to be a lot more inclusive. I think that all the things that digital has wrought – everything from making recording cheaper and distribution cheaper to offering brand new ways of promoting your music without having to spend enormous amounts of money – are all going to lead to greater participation from more musicians and more listeners. So, I think it’s a great time to be in the music business provided you’re ready to embrace change.
Do you find that when you talk to fans and musicians that they’re optimistic and then when you speak with industry people that they’re very pessimistic?
I would say that it breaks down a little bit along those lines. I’d say that a lot of listeners are confused right now. You know, the digital music thing is a very confusing space. I think what Apple did was make it very simple. That’s really the genius of that whole iPod/iTunes system. It solved the problem. It solved digital music for a lot of people. But you have countless devices, different codecs, bitrates, all sorts of digital rights management standards, and I think for most consumers it’s a very confusing space. So, that’s certainly what I hear at music get-togethers.
Pandora basically plays the role of matchmaker, introducing listeners to new music. What are the most important factors to doing this linking and connecting role effectively?
The basic premise of the Genome was to try to capture every possible detail about a piece of music that could be playing a role in why somebody likes it. So, let’s say you and I like the same song. We could like it for very different reasons. You might like it because you like the guitar parts or you like the lyrics or it’s got some really great musicianship. I may like it because I really like the sound of the voice or the folk influence or the sound of the acoustic guitar. Different things. I think the key to really nailing a recommendation system is to have a deep enough understanding of music that you can that you can distinguish that. And then you can understand over time what somebody likes about music, but in the meantime that you can offer them with very little effort a listening experience that really works for them. So, you can’t offer a system that requires a lot of training or feedback or a lot of time on the part of the listener. To get it good for them, you have to start off pretty darn good and then get better. And that’s really how we conceptualize the whole Pandora experience.
What are the benefits of a manual listening and cataloguing process versus an automated one?
I don’t think there’s really a religion around this. I do think different things work for different people, so I’m not going to say that this method is better than another method. Because there are people who do prefer a radio station that was created by collaborative feedback, where they type in a song they like and a station is created by what everybody else has typed in. For some people, that’s the answer. I think that the only way to answer the question of is something better or not is to ask what is the reaction of listeners. The lat year we have a fairly unambiguous thumbs-up from millions of people now. It’s growing faster than it ever has before and it’s growing because people are telling each other about it, which I think is the highest form of endorsement. I think the reason people do like this a lot is because the Genome is really good at recommending music you’re going to like – with very little input. One song or one artist. I think that’s a real magic of the Genome.
Do your analysts specialize in a particular type of music or are they encouraged to listen widely?
We hire analysts who have as broad a musical background as possible. They’re all trained musicians so they typically all have at least a four-year degree in music. But they do have to take multiple exams to analyze each successive genre of music. So when you come in the first exam you take qualifies you for the first basic pop-rock genres but if you want to do jazz you have to demonstrate that you have the chops to do it. Most of our analysts can do most of the genres.
With technology like Pandora and with widespread changes in listening habits, has the notion of genre in music become irrelevant?
For us it’s certainly not core to who we are. We don’t define music by genre. Genre is really just a collection of genes to us. And it’s fluid, so you flow from rock to acoustic rock to folk seamlessly in Pandora. You can go back and forth between them. So we don’t really bucket music that way. I think genre has served a useful purpose because it’s a shortcut, it’s a semantic shortcut to help you decide which part of the record shop you should go to. But it’s a double-edged sword because you’ve got to make a decision about which bucket you’re going to be in. If you’re an artist, then as soon as you do that you’re excluding yourself from potentially a whole group of people who have chosen an adjacent genre. So I think it’s got benefits and drawbacks.
Does Pandora serve to deepen tastes or to broaden tastes?
Well, I think that what it does is that it exposes you to music that you’ve never heard before. Fundamentally, sort of by definition, we’re recommending music that’s similar. So we’re not going to take you from a pop tune to a reggae song. That’s not really what the Genome is set up to do. But we will take you from music you know to music you don’t know, so I don’t know if you call that broaden or deepen but I think that’s what we do.
From observing people’s tastes in music in a “wisdom of crowds” sort of way, what have you learned about what people like most and like the least? Any peculiarities surfaced?
It’s funny. I haven’t really looked at lists like that. There’s an awful lot of data we have sitting in a database somewhere that we haven’t even looked at yet. Collaborative feedback does play a role in Pandora and I think that it has a role to play. I don’t think that it’s enough by itself. When you give a thumb to a song, you’re not only helping to fine tune the classification just for you your thumb gets added to that thumb’s overall rating. We then take a look at the sort of aggregate feedback that the general audience is giving to each song within the context of each station on which it’s played. If an audience overwhelmingly thumbs a song down, it’s kind of like telling us that the Music Genome Project just doesn’t have this one quite right. So we start playing that song a little bit less for everybody that plays that station. And vise versa if it gets thumbed-up a lot.
Thanks for sharing your thoughts with us, Tim.