In your own words:
> We take your Spotify stream count, estimate what that means across all services, take the US portion, and multiply by the US mechanical rate.
As someone who models stuff for part of my living, I understand the spirit of what you are achieving here, but there are too many assumptions / genuine unknowns. Like just in terms of the things you said: you would need the real premium and and supported stream counts over time, the relative market share and stream counts in other services over time, broken down by country, and of course, what "mechanical rate" really means, because there's what Spotify pays versus everything else, and what Spotify, for example, pays, is really fluid.
For other readers: Artists certainly think in terms of pay per stream, but that's not really how Spotify pays, for many payees (the MLC and the other categories). So even though this guy making this website is a professional in the music industry and knows all this, he is so focused on lead gen that he keeps things simple - fine, that's valid - but nonetheless wildly inaccurate.
All that being said, the low-end of the estimate is VERY real, because we sum up an actual number provided by The MLC. For their unmatched data, they provide a range of one dollar sign ($) $0.01 - $0.99, ($$) $1.00 - $10.00, ($$$) $10.01 - $50.00, ($$$$) $50.01 - $250, and ($$$$$) $250.01 and above. Those exact figures get worked into the estimate, and are real hard baselines that we use for the estimate. There is no inflation from us on those figures. When a track has an unbounded (five dollar sign), we then use Spotify counts to estimate where in the range it falls.
I appreciate the thoughtful conversation.