Record labels deal in risk. They never know if a given artist, song, or album will find traction with the general public. It doesn’t matter how much they spend on marketing and hype, there’s never a guarantee that anything will be a hit.
No wonder labels try to mitigate this risk. The most obvious way is seen in the herd mentality. Once a certain type of music becomes popular, labels chase after artists who sound similar. Think back to the grunge era. After Nirvana caught on, there was a stampede to sign more grunge and grunge-like bands. Today labels are looking for Billie Eilish clones.
A more proactive approach is to construct songs in ways that improve their hit potential. Software programs like Hit Song Science have been used to determine if a given song has a shot at shooting up the charts.
Now Amazon, a company that knows a thing about algorithms, machine learning and AI, is patenting its own hit-predicting software. The first filing was in 2013 but is now being updated with the latest technology.
The approach is interesting. Digital Music News reports that Amazon is creating algorithms that “can identify ‘early adopters’ of obscure media. The idea is to introduce these tastemakers to new obscure media they might like, based on past listening preferences. The algorithm can then ‘determine media items that are likely to become popular in the future based on habits of early adopters.’”
Translation: The software gathers listener data from music obsessives and uses that to predict what songs may cross over into the mainstream. Amazon will presumably use that information to push music discovery through its Amazon Music platform. And the more you use Amazon, the better the system will get.
The good news: This may help make music discovery easier and more efficient. The bad news? More digital smog on you as yet another personal profile is created and exploited. But we now expect that from our new robot overlords, right?