>They could do it dutch auction style. Suppose there are five drivers and three rides. The drivers bid per-mile rates of $.75, $1, $1.25, $1.50 and $1.75. The rides go to the three lowest bidding drivers and they all get $1.25, i.e. the highest rate of the drivers who get rides.
The problem with all these variables is that the more "preferences" you introduce into marketplace to matching, the longer your pickup time will be.
The 2nd most frustrating change is going to be the increase in pickup times because drivers will now optimize for rides that are exclusively in city limits.
You can address things like that with predictive matching. You don't actually wait until there are three rides before matching anybody, you just look at the data for this timeslot and expect that statistically there will be three rides this hour, use that rate based on the drivers who are on, and assign the nearest of the three lowest bidding drivers as they come. If a fourth ride appears that hour then you immediately raise the rate to the fourth lowest bidding driver starting with that ride, and so on, and next time have more data and make a better prediction.
The problem with all these variables is that the more "preferences" you introduce into marketplace to matching, the longer your pickup time will be.
The 2nd most frustrating change is going to be the increase in pickup times because drivers will now optimize for rides that are exclusively in city limits.