The digital age has already condemned record company A&R men to spend as much time hanging out on MySpace as they do in sticky floored bars. Now a piece of Israeli software might have them perusing statistics to find their next breakthrough act.
The program is claimed to be able to spot upcoming pop artists weeks or months before they hit the big time by watching people share music on peer-to-peer file-sharing networks.
But rather than spending its time listening to poor quality demos, the software tries to identify local spikes in interest in a particular artist, says Yuval Shavitt, a computer scientist at Tel Aviv University.
This could indicate a local artist making a splash before the buzz spreads to a national level.
Atlanta-area rappers Shop Boyz were largely unknown in early 2007, when Shavitt says his software flagged them for success.
Weeks later they had signed with Universal. Their single Party Like a Rockstar made the Billboard rap charts later that year, and became 2007's most popular ring tone in the US.
Shavitt's predictions were based on information from a commercial service called Skyrider that tracks searches on the Gnutella peer-to-peer file sharing network. Because the network reveals the IP address of some users, Shavitt was able to track many of the search terms back to their geographical locations.
Searches for the Shop Boyz jumped from 0 to about 200 a day within two weeks, with most coming from the Atlanta area. That signal was extracted from the daily background of 20 million searches on average, 20,000 of them for Madonna.
Shavitt says that, on average, 30% of the acts his system flags up succeed at a national level, and that it could be useful to record companies. A similar approach could detect new talents on centralised music or video sharing systems, for example YouTube, he adds.
Shavitt's work shows it is possible to develop software "that might help the music business predict and then help foster some artists that are likely to be big successes," says James Marsden, from the University of Connecticut.
Marsden, who develops systems that can spot useful information and the efficiency of markets, says: "I think it's got a nice business application to it."
A paper on Shavitt's technique was presented at the ACM International Conference on Knowledge Discovery and Data Mining earlier this year.