Spotify is taking a cue from Pandora Radio for its recommended music feature by incorporating more music DNA. This means recommendations will also include how music sounds and what kind of structure it has to give us better recommendations.
Anyone who's used Spotify has probably grown accustomed to its music recommendations feature. It takes songs and artists that are recommended by people with similar musical tastes as you, including your Facebook friends if you've tied that into your Spotify account. This is called collaborative filtering.
The achilles heel of that is: you might not like music that similar people or friends like, just based on association. This is where Pandora Radio excels, their Music Genome project makes music recommendations based on song structure, rhythym, melody and genre. This makes a nice counter-balance to the social recommendation option -- you also get music that really matches what you like based on tempo, sound, aesthetic, etc.
A University of Ghent Ph.D. student named Sander Dieleman, a Spotify intern this Summer working in New York, is tackling the issue. He's working on a way to recognize acoustic features in music, and factor that into the way that Spotify is recommending music. Sander has published a great blog post on the whole thing, which is a long read but interesting.