[Normal_gaussian]

Assuming the normal and other statistical handwaving is very dangerous in this kind of dataset and application.

It will give you a high probability of being near right in a random case but completely destroys the detail. In this case the detail is niche music genres and artists, growing or dying, that contribute greatly to the "flavour and future of society". When we assume too much here we actively suppress minority positions making certain genres and use cases completely non-viable.

I highly doubt elevator music has the audience listening profile of eminem, and thus the current system either over or under pays elevator music producers.

Other areas which are likely extreme are: new EDM artists, classical compositions, play-a-long songs (ie learn guitar), cultural artists (US christian youth, Belarusian native rock).