That's not a bad idea. Possibly in the future I'll see if I can incorporate this into the estimates but remember its also again based largely on the location of ones viewers. If you're from the US and your viewing audience is 99% from Canada you'll have a lower payout then the opposite. Same is even more dramatic with other countries. That type of data I'm not able to get.

[Edit] As a general rule of thumb though, the higher up partners are generally making closer to the higher range, and the lower end ones are making closer to the lower end of the range.

Adding some kind of confidence bound or estimated probability distribution might make more sense of the huge ranges. You're already computing the 'height' of partners, so you could potentially incorporate that into the model.

Something like "$100 - $1000, best-guess $500 ± 200"