Making Music


I’ve been impressed and entranced by the music service Pandora since I first ran across it several online lifetimes ago in 2007.

Two things particularly impressed me about Pandora. First, unlike services such as Spotify that allow you to access music you already know about, Pandora was the first large scale attempt to offer music discovery. Enter the artist or tracks you like most, and Pandora would find more music that was similar. Normally you would expect to learn that Pandora is powered by cutting-edge algorithms.

In fact, Pandora is powered by humans. Music school graduates. Many dozens of them, all methodically classifying individual songs against a master taxonomy of over 400 characteristics. It’s an expensive approach, but it’s organized and returns consistently high quality results. And while Pandora continues to struggle from a profitability standpoint, nobody argues with the quality of its service.

But what if you could create a Pandora-like service without the high labor costs? That’s what a company called 8Tracks set out to do.

Rather than having a paid staff categorize music, 8 Tracks went the social media route. Everyone was invited in essence to become a DJ, and upload their own song lists to the 8Tracks site. These playlists were organized via tags, so users could discover music based on mood or musical style, for example. If users like particular playlists, they can follow the people who uploaded them in order to see all their new playlists right away.

8Tracks is unquestionably providing a music discovery service, just like Pandora. But it’s a fundamentally different experience. Pandora is dependable, seamless and efficient. 8Tracks is hit-and-miss, time-consuming and requires lots of user interaction.

There’s room for both services in the vast music market and indeed, both services have many enthusiastic adherents. Yet by looking at both services side-by-side, you can see the strengths and weaknesses of user-generated content very clearly.

Music is entertainment. There’s no risk or consequence if you don’t discover a certain song by a certain artist. But when you move into the realm of business information, that dynamic changes. Suddenly, getting the right answer starts to matter a lot. That’s where user-generated content can come up short. Users generate whatever content they want, whenever the want, for as long as they want. You have little control. User-generated content works best where there is a massive volume of content (think Yelp or TripAdvisor) and the correct answers will win out, or in situations where there is no alternative information source, making your content the best that is available. But when the quality of your content matters, social approaches to content creation can yield decidedly off-key results.