I recently summarized Seth Goldstein's epic series of essays on the tantalizing prospects of applications and business models powered by user-generated metadata. Well, just a few days earlier I also posted elsewhere, about Audioscrobbler, which is...wait for it...an (actual) application and (potential) business model powered by user-generated metadata. Took me a few days to make the connection, but I finally got there.
In their own words, Audioscrobbler...
...builds a profile of your musical taste using a plugin for your media player (Winamp, iTunes, XMMS etc..). Plugins send the name of every song you play to the Audioscrobbler server, which updates your musical profile with the new song. Every person with a plugin has their own page on this site which shows their listening statistics. The system automatically matches you to people with a similar music taste, and generates personalised recommendations.
I kept iTunes running on shuffle for a week to build up a profile and see what happens. Here's a look at my Top Artists at 1,011 songs. So let's apply Goldstein's aaaaanalytical framework to Audioscrobbler:
- Automata: Spontaneously generated metadata in sufficiently useful amounts? Check. (Although as my friend Hunter pointed out, can you trust it? Part of the exhibitionistic motivation in sharing this info is demonstrating how cool you are by virtue of your impeccable taste in music--but judging by actual sales, there's a lot of bad taste out there. Just what are you getting when you're looking not at actual, unmediated behavior but at what people want you to see? The system's also vulnerable to more aggressive gaming: At this writing, eight of the top ten songs across Audioscrobbler's entire network are by System of A Down. I like Nu Metal as much as the next person (read: not that much), but this strikes me as somewhat, ah, corrupted market data.)
- Algorithm: Means by which individuals and their content can be aggregated, classified and rated? Means by which that output can be fed back into the development of new content? Well, semi-check. Audioscrobbler provides much of the former now, although the system's a work in progress, and they clearly intend to provide the latter at some point in the future, although I don't think that's happening just yet.
- API: Interface through which individuals generating content can identify and communicate with each other? Check. Here's my Audioscrobbler network of 50 Musical Neighbors. Small world story: Josh Benton, neighbor #8, is a journalist in Dallas whose personal website is one of the first blogs I ever read regularly. Never met or talked to the guy, although I commented on his site a few times. Haven't even thought about him in a few years, and here he shows up in my Audioscrobbler backyard. Interesting. Interface through which they can communicate with any secondary content-generators? Not yet, but it wouldn't be hard, especially since all your individual data is available as a feed under a Creative Commons license.
- Alchemy: Can value be added, i.e. can you commercialize the data? Big check, contingent on a satisfactory resolution to the corruption problem noted above. I think it would be incredibly valuable to bands, labels, promoters, magazines--anyone in the musical supply chain--to be able to crunch this data...if it's actually a reliable indicator of people's listening and purchasing behavior.
- Arbitrage: Can they buy low and sell high? Big check, again, contingent on its reliability. You can't buy much lower than having it given to you for free.
Cool stuff, and even if Audioscrobbler fails to take off, I'm convinced that this framework is the right way to assess applications like this.
