Predicting music sales is of particular interest for sales managers (e.g. for pricing), inventory management (for CD sales) and server balancing (for music download). In the past years, research therefore proposed several models for music sales prediction. These models have, however, some shortcomings which we want to overcome with a new approach. We suggest using a novel data set that is a byproduct of smartphone apps that help users to identify music. Shazam is probably the most popular of these music identification services for smartphones. This study examines the relationship between Shazam charts and song sales using data from the UK over a period from September 2010 to May 2011. Using seemingly unrelated regression we identify that Shazam charts precede sales charts by about two weeks and serve as good predictors for sales charts. Further, we find that the rock, pop, and hip hop genre, artist?s popularity positively affect song sales.