What I want from a music discovery service

I think it's rare to fall in love with a piece of music on first listen. Sure there are track that are that good. But some really good track take a few listens to really appreciate. I hypothesise that between 3 and 10 listens are required to get to know a piece of music to be able to assess whether you like it (the repetition number). I also hypothesise that that 3 to 10 listens must occur over a relatively short period of time, such as a month (exposure window).

So, I want a music discovery service that deliberately repeats tracks to get the exposure necessary to allow assessment. The service would need to record, and hence predict, the number of hours of music each listener listens to on average each day in order to hit the repetition number within the exposure window.

This service would be open to all artists from major labels to indie self-publishers, although the users would be able to select a popularity focus (for example if they're interested in discovering the undiscovered). Artist would be able to upload their tracks, applying tags to give a rough genre description (genres are boxes if you're put in only one, but they're a tool for discovery if you're put in many).

First-time users would be required to sign-up, and select genres of interest, and then begin listening. When they find a track that they like, they star it (and share it), which moves it out of the exposure playlist and into a liked list. The parameters for selecting music for the users exposure playlist would be updated to take account of their preferences, matched against the preferences of other users.

Users would also be able to listen to a playlist of the liked music, starring from these moves the track to a really liked category.

Users would also be able to force a track to be dropped from their exposure playlist or could hit a keep this in for a while button to ensure they get a few more listens before it ends its tenure in the exposure playlist.

The service would store a record all tracks that had previously been in the exposure playlist were not liked, hence ensuring that these are not added back to the exposure playlist. However, this catalogue would be searchable, by date listened to, time, artist, etc to allow users to find music that they only realised they liked a while after. Alternatively, users could choose to listen to a playlist generated from all their previously listened to exposure tracks that they didn't like.

Controls the user could be given over their exposure playlist:

• Popularity of the track (max / min sales, likes or listens)
• Popularity of the artist (max / min sales, likes or listens)
• Genre tags (must have / mustn't have)
• Year
• Repetition number
• Exposure window