3 ms·
In terms of unit economics, UberX is still (mostly) sound -- the rider is paying a fare based roughly on time/distance, and the driver is paid a fare that's als
by kiwidrew 7y ago
In terms of unit economics, UberX is still (mostly) sound -- the rider is paying a fare based roughly on time/distance, and the driver is paid a fare that's also based roughly on time/distance; it's very rare to see a rider paying less than what a driver gets paid. (Even after taking into account the various driver incentives.)
Where the unit economics get completely messed up is Uber Eats. In my local market (Wellington NZ) the eaters are paying a flat delivery fee of $5.99 to $7.99, while the drivers are being paid $5 per pickup and $1.35 per km. This results in drivers earning anywhere from $5 to $20 for each delivery -- and that's without even taking into account any driver promotions. No wonder Uber had to come up with an "excess driver incentives" line item!
Sometimes Eats will have the driver pick up two orders at the same time. This lets Uber collect two delivery fees from the customers while only paying the driver a single pickup fee plus the kms from restaurant to customer A to customer B. In theory this should work in Uber's favour. But their system all-too-frequently will make insane assignments, such as batching together two orders for customers that live in completely opposite directions. There will be situations where the customers have each paid, say, $7.99 for delivery... and the driver ends up earning $30 from the trip!
So this is cutting very deeply into the 35% commission that Uber charges restaurants on the total order value. They're trying to hide the insane unit economics of Eats by spreading the various revenues and costs around different line items...
- hef19898 7y agoBack the day I had a short stint at a bus charter business. There we also tried to reduce costs by bundling travel group's individual itenaries and match that against available busses to reduce empty rides. The result would have been that groups might have different busses on different days. Nice idea, in theory. Turned out that avoiding empty trips in transportation is hugely dependant on volume by time and region. It is also a very hard problem to solve as you need to match availability data from busses (which needs to be collected in a scaleable way), trips (which need to be in the correct regions and times) and then simply come up with an optimization algorithm to sort it all out. Peace of cake... Not. Especially if you want to link that to your pricing already during customer booking instead of after-the-fact discounts. One might suspect that the transportation industry which has the volumes and the cost pain of empty trips might already have solved it if it woupd have been that easy.