Music algorithms should listen to volume
Music apps, like iTunes and Pandora, have found a lot of metrics that correlate with music listening habits. Not long ago, iTunes added skip count (the number of times you hit the next button after hearing the first 10 or 15 seconds of a song).
Here’s one I don’t think any service is taking into account: whether a user raises the volume on a song after it starts playing.
Music software has no way of knowing whether you like one song more than the previous, unless the user chooses to actively rate a track — thumb up in Pandora, 5-star picks in iTunes. Most people, I’d assume, don’t make the effort.
But you might find that many people, when they hear a song they like, will pump up the volume. Music services could listen to that, and take note for future instances of when it’s deciding what to play.
(The services would have to normalize for fluctuating volumes in individual files. Since apps like iTunes can already do that, using a feature called Sound Check, that shouldn’t be too difficult a task.)
Digital services are better when they listen to and apply relevant personal metrics. Nothing new there.
Music software, especially Pandora, have proven that something as abstract as songs tend to follow a predictable pattern. If you like one artist, look at the traits, or genome, of the songs they produce; look at the music tastes preferred by other fans of those songs; and there you have a pretty good idea of who that artist is connected to. Nothing all that new.
Spotify, for example, is pursuing a technology that caters ads based on the mood a listener is in — that’s taken from the tempo and types of songs they’re listening to during any given session.
