Earlier quoted context omitted.
It's to analyse where riders actually go after being dropped off; for example understanding that if you want to go to building XYZ you have to get off here. Collecting data for self driving cars probably.
Yeah, but why does Uber need/care about that information? If I ask to be dropped off at X and walk a couple blocks to Y, Uber shouldn't care. I can understand the argument that maybe going to certain buildings you have to be dropped off at a specific spot, but that seems like a minor gain for something that's more than just a minor intrusion.
The data is used for (or can be used for) a number of practical and important purposes. These are some examples of what we do as well as what we could do with such data:
- Fighting fraud
- Improving "suggested pickup" locations
- Improvements around POOL (we'd rather suggest a location that's easy to pick up from than one that's unsafe)
- Figuring out what side of the street you end up on (we can do a better job of choosing a route if we know you're able to get out of the car safely)
- Optimizing pickups and dropoffs around events or certain times of the day or week (Caltrain riders: imagine if we suggested pickups at 5th and Townsend instead of 4th and King during rush hour. Much easier to find your driver and prevents congestion)
- Dropping riders off closer to the "correct" entrance to their building
- Dropping riders off in a location that's easier for the driver to depart from without making the rider walk too far (e.g., instead of getting trapped in a weird road or parking lot)
- Analyze where riders ask drivers to make stops mid-trip and make better experiences around that.
As far as trust and safety goes, we can avoid charging you if your driver was looking for you at the wrong location and gave up. We can better see where you were actually dropped off at, if your driver lies and leaves the app running after you've left the car. Leave your phone in the driver's car, and we can potentially do more than just put you in touch with them.
Individual users' data is very closely guarded internally. It's immensely difficult to look at user data without specific access. Overwhelmingly, this data is queried in aggregate and fed into machine learning systems. The risk of abuse is exceptionally low.