[zaidf]

Uber could know with 100% confidence based on past data that a driver passes by a certain address almost daily. For him to deliver a courier to that address would be significantly less work to driving a few miles only to deliver the coirier(the assumption in your argument).

I still don't know if the financials make sense in my particular example but it does tell you the optimizations Uber can make based on the historical data.

[aetherson]

Well, but no. They can't know with "100% confidence" where a driver will drive in any given day. They probably can't even know with 90% confidence. Taxi drivers don't drive regular routes, they drive where a diverse set of passengers need to go. There are probably a few destinations that a lot of drivers hit almost every shift (most notably the airport), but arbitrary business or residential address 1 to arbitrary business or residential address 2? Probably not.

And if they were to offer some kind of "by the way" service where you say, "I want to get this package across town, you send somebody who will eventually deliver it to my destination because my destination is near a major taxi destination," do people actually want that service? I was under the impression that in-town point-to-point couriering was typically done because you have a tight deadline for when the package needs to arrive -- after all, if it's just a few miles away, and you aren't on a tight schedule, you could just mail it and it would probably arrive the next business day. This doesn't sound like a big business opportunity to me.

Delivery of large quantities of goods on a same-day timeframe, like Amazon wants to do, that's a big business opportunity. But I don't see how Uber's business or Uber's dataset solves that problem.