> Uber has been very poor at optimizing logistics here, which is why they lost. I talked with drivers who were pissed at having to drive 30 mins for a 15 min de

> Uber has been very poor at optimizing logistics here, which is why they lost. I talked with drivers who were pissed at having to drive 30 mins for a 15 min delivery, and not be told where it was. But it could be a mix of paying drivers too little, resulting in too few people and long waiting times, resulting in people buying from the competition. No, it's the logistical problem not making sense. The reason they have to drive so far for a 15 minute delivery is because there's nobody closer. The reason they don't tell the drivers is because they wouldn't want to do some of these drives if they had all of the information. Uber is likely making pretty good guesses at this problem, but it's a very hard problem that frequently won't have a feasible solution that is profitable. You have a driver at A. You get an order for B. Do you send the driver to B or wait for an order to A? Hmm crap there's nobody else nearby. Shit we just got an order for B that's going in the opposite direction. Should we bundle them up? The key problem is the ratio of drivers to demand. If the ratio is very high, the optimization is easy but the drivers spend a lot of time not doing orders and so they waste time and earn a shitty wage. Wait time is high. If the ratio is very low, the optimization is very difficult and requires a lot of guesswork. And by guesswork that's queuing theory and maybe some time series forecasting combined with real time optimization algorithms. Wait times for drivers are lower but delivery times get a lot worse and the labor is more expensive. The only time it "works" is when everyone is sweating. There's just enough drivers to meet demand and get a satisfying wage. And all of that depends on the price for sending drivers on mostly continuous routes is palatable to the customer, which it likely isn't because frankly that sucks and will always have a lot of variance for the drivers. Put another way. If a driver wants to earn X per hour, if they spend 10 minutes waiting per hour, they need to be paid 20% more than X per hour for the hours they actually work (6/5). If they spend 20 minutes waiting per hour, they need to be paid 50% more than X per hour for the hours they actually work (6/4). The uncertainty and difficulty of the optimization problem mean that they're going to have these breaks. So even if your mental model says "I would pay someone $X per hour to do deliveries" you need to put on a pretty hefty percentage for the burden of the gig model. And then of course uber is pretty bloated and unprofitable.