Andrew Milgram

It was just a scattergram of behavior.

And as a consequence, they were under-earning what we thought they could and should.

By the way, fast forward, later when we set up our own operation, we ran a bunch of experiments with drivers where we said, look, you're likely to make, let's say, $200 a day at that point net.

we'll guarantee your $200.

But we want you to run an experiment with us.

So if you will just follow these patterns of behavior, what we see as the most profitable, we'll guarantee the $200.

And by the way, if you earn more than $200, keep it.

We ran dozens and dozens of experiments.

How many payouts did we make?

Tons of payouts above the 200.

So we paid nothing.

The driver always out-earned when they followed the data-driven decision-making.

I mean, my favorite was what I called the NASCAR loop.

So the data showed that if you picked up at the bottom of Broadway, you were high probability around Columbus Circle, you were going to drop off somewhere north on Broadway.

I was going to make the left turn because the data also showed if you picked up at the top of Broadway, you were going to drop off somewhere midtown around the bottom of Broadway.

So we just run that NASCAR loop, left-hand turns only, and it turns out to be a super profitable circuit.

And there's lots of pockets of opportunity around New York.

The other thing, by the way, is the fleets did a terrible job of telling their drivers when there was a Knicks game.

They did a terrible job of telling them when there was a Rangers game.

Terrible job of telling them when concerts were going to be all at MSG, obviously.