Music discovery is more a service to users than it is a sales-driver for content owners. But if it where, the objective would be simple: recommend songs that a user will have the highest probability of enjoying, but does not yet own. Triangulate with friend’s listening patterns, similarity profiles, crowd favorites, music blogs and then cross off what’s already owned (thus not a sales lead)

What kind of recommendations will you get? I would venture to guess that the best sales leads will be remixes, live performances, covers of songs a consumer already likes. Essentially, there’s money in selling the same song/variation to people over and over again! I’m a Dave Matthews fan, and boy have they milked me with their many live performance recordings.

Of DMB’s top 5 songs, I own an average of 6 variations each. The same could said of Pearl Jam and their streak of “bootleg” live CDs

Sure, not everyone is a big fan of these two artists, but we all could have at 2 favorite artists whose music lends well to variation. This is essentially a different take on the Long Tail where over-selling takes the place of over-pricing niche demand.