Live or Die by the Quality of Your Metadata
Update (April 23, 2006): This weekend La La pushed a new catalog which considerably improves the metadata situation. There is room for improvement (still no Spoombung or Sarin Israel Nes Ziona), but I’m far happier than before.
I’ve been using La La a lot lately. La La is a great, simple idea: create a CD bartering marketplace, exploiting the web as an extremely efficient machine for matching those who want with those who have, and borrowing a streamlined postal delivery mechanism from NetFlix. All you really need is a database full of CD catalog metadata, a neato AJAXified interface, and a bunch of preprinted CD mailing cases, and you can sit back and let the $1.50 per trade pour in.
But that CD catalog metadata had better be good. Every CD missing from it is a trade that will never happen. Every CD filed under the wrong artist is a trade that will never happen. Every box set listed as a single CD, every truncated album title… you get the idea. Even incorrect cover art could lead to confusion and inhibit trading. Bad metadata is especially destructive for the long tail trades–if only one person out there in La La land has the CD I want, and they have trouble finding a way to list their CD because it’s listed in La La’s catalog with the wrong info… well, it’s goodbye La La and back to Amoeba for me.
Unfortunately for La La, their metadata sucks. They have all the problems listed above and more. Missing CDs I could possibly forgive–maybe they’re due to legal doubts around imports–but mispellings, uncorrelated artist name variations, and truncated album titles I can’t. My advice to anyone thinking of starting a similar venture: don’t try to build your own metadata catalog, and don’t buy one from some crappy closed metadata vendor. Closed vendors can’t scale to the long tail. You need metadata from an open system: in the case of music, something like MusicBrainz or Discogs (I think Discogs is slightly better). The closed metadata company’s drones won’t get around to cataloging the latest Spoombung album or a limited edition Muslimgauze CD. Hardcore music fans, on the other hand, will–especially if it means they’ll be able to trade them.
Services like La La exist by virtue of their metadata. If they aren’t careful, they’ll cease to exist because of their metadata too.
April 13th, 2006 at 7:26 pm
and they didn’t even ship me a starter kit… La la is about as flakey as it gets, it seems. Their support FAQ made me laugh at first, then cry and then want to send them a mailbomb. (of course, I’m kidding).
April 14th, 2006 at 8:37 am
my experience hasn’t been that bad–i got a starter kit a few days after signing up, and my trades have all been smooth.
they are still in private beta, so i’m willing to cut them some slack. they definitely need to hire someone from SoI to help them organize their forums, FAQ, support, etc… if they don’t die from their metadata issues, they will die from poor community management.
April 17th, 2006 at 9:39 am
I checked out the two sites you linked to (MusicBrainz and Discogs), and they both seem to be missing UPC/EAN data for albums. Without correct and complete UPC data, metadata like this is fairly useless for e-commerce (or trading) purposes.
Do you know of any open-source projects that a) collect UPC data, and b) have a Web API or data download? If not, I may start one myself.
April 17th, 2006 at 2:55 pm
I don’t know that I agree that UPC data is critical for trading. MusicBrainz and Discogs record metadata at the level of individual expressions (albums or singles), while UPCs are specific to individual manifestations or physical embodiments of those expressions. While there are certainly some applications that require manifestation-specific metadata, I don’t think trading is necessarily one of them–most of the time, as long as it has the same versions of the same songs on it, people don’t care which specific pressing they have.
That said, I agree that it would be nice if either of those efforts had UPC/EAN codes as well, where available (there are a lot of indie andbootleg releases that don’t get assigned such codes).
Also, MusicBrainz does have an API: http://musicbrainz.org/client_howto.html