I spent a few hours the other evening entranced by a relatively new service called Pandora. Type in a favorite artist or song into its intuitive user interface, and in no time you are listening to a seemingly endless list of songs that are musically similar to your entry, meaning in other words, that there is a high probability you will like them too.

It works. Song after song came on, and to virtually every one I found myself nodding in agreement that I liked it, in many cases, I really liked it. Most impressive of all, the music was from artists I had never heard of before, and probably would never have stumbled across on my own.

Since Pandora is another new web venture with lots of buzz and no apparent business model, I automatically assumed that there were some whiz-bang algorithms behind the scenes identifying similar music. On further research, however, I found that Pandora is powered by something called the Music Genome Project, a group of 50 or so music experts laboriously classifying songs by up to 400 musical characteristics each, with every characteristic having a rating from 1 to 5. In short, this remarkable new site is human- powered by a hand-crafted, classification scheme, a musical taxonomy.

This revelation reinforced my belief that organizing information (even information in audio form) adds tremendous value. There are dozens of sites that provide online access to music, but you have to know what you want to hear. Pandora's taxonomy reverses the process: tell it what you know you like, and it tells you everything similar to it. This is a much more rigorous approach than say, Amazon's recommendation system (which says "people who have bought the book you are buying have also bought these books"). The Amazon approach works well, but it has an inherent bias towards things that are already well known and mass market. And that's why Pandora is so exciting.

Apparently, Pandora will allow new and obscure artists to upload their music, have it categorized via the taxonomy, and thus expose it to listeners who might otherwise never come across this music. By organizing serendipity, Pandora has the potential to move beyond being a reference database to an audio buying guide. Type in your criteria (in the form of a song or artist you really like), and Pandora will provide you with a list of other music you are likely to enjoy, which you can presumably then be able to purchase. It's amazing how much you can accomplish with a little organization.

Labels: buyers' guide, buying guide, data publishing, Infocommerce, Pandora, taxonomy