Estimating crowd size is a settled science. (Probably the only thing you can write about science being “settled.) Making money out of it...well, not so much. Settled, I mean.
Crowd size yes. But how many are supporters vs protesters vs news vs cops?
How do you normalize the data set when one candidate goes to 100 venues and draws 100 people each and a different candidate goes to 50 venues and draws 200 people each?
Do you normalize for a hot / rainy / snow day?
Do you normalize for time of day? Day of week?