Getting taken for a ride by dynamic pricing

The New York Times had an interesting article on Uber, a car service firm operating in several cities (Disruptions: Taxi Supply and Demand, Priced by the Mile, Jan 8). Uber allows users to order a car through their smartphone and have everything billed to a credit card on record. They also aim for quality service. Their website promises that your ride will come within five to ten minutes. That all sounds lovely, but how can deliver on that service goal when demand will be high. Think New Year’s Eve. How do you handle the spike in demand after the ball drops and folks want to go home? This quote from the Times gives an idea of how they did it:

On New Year’s Eve, Dan Whaley, a tech entrepreneur in San Francisco, got into a black Town Car and was driven one mile to a holiday party. The ride cost him $27. At the end of the night out, Mr. Whaley took a Town Car home from the party. This time, the exact same ride cost $135.

That’s right. Uber was using dynamic pricing, applying a multiplier to their basic rates depending on how much demand there was. Here is what customers saw when they ordered a car.

The Times article emphasizes that many customers were rather ticked by the seemingly exorbitant rates; because Uber prices by the mile and you don’t tell them where your going until the driver arrives, some amount of sticker shock at the end of your trip may be unavoidable.

But the Times actually sells short just what the firm is doing. Yes, dynamically raising the price chokes off some demand but it also brings more capacity to the market.As far as I can tell from their site, Uber doesn’t own any cars or hire any drivers. Rather they are working as a middleman matching customers with customers with independent car services in the city. Rising rates then also gets more drivers to commit to Uber customers.  Here is how their blog explained the plan before New Year’s Eve.

In the last several days, we’ve received a ton of questions about how reliable Uber will be on New Year’s Eve. “Will there be a car available?” “Can I count on you guys for the New Year’s festivities?”

On New Year’s Eve, yes that’s tonight  , we are aiming to provide a reliable ride to anybody who needs one, no matter how crazy demand is or what is going on in the city. We won’t be perfect, but we’ll be damn close; and as you might expect, there’s a price to that kind of reliability and convenience during such a massive spike in demand. We’re rolling out what we call Surge Pricing in order to achieve such a high level of service on NYE. …

We are able to get a far greater number of drivers on the system when Surge Pricing is in effect – it’s basic economics. Higher prices encourages more supply to come online. It gets some drivers out to work on NYE. It keeps other drivers from going to alternatives like renting their car out for the night, or trying their luck at hustling rides on the street. Higher prices means more cars, means more rides, means more people getting around the city efficiently, safely AND in style 

(Note the emoticons are theirs, not mine.)

So how did this all work? This is what they recount in their blog after things hit the fan and users complained about the cost of Twitter etc.

The higher the price would go, the more people would choose not to use it, but there was an almost unlimited amount of demand to suck up the limited supply. At UberMissionControl, it felt like there was no sensitivity to price, but what was really going on was a massive torrent of desperate demand with a very small fraction (still huge # of people) willing to pay anything for a ride.

To our dismay, the pricing multiplier kept going up. The math was doing its job—you could start to see the utilization figures getting some slack, but then another wave of demand would hit, and continue the price surge. At some point the east coast cities started breaking 6x multipliers—we accepted defeat at that point—the unbending demand breaking our will. We would bring cities down to 3x, only to see conversion go up, supply go down, cars get saturated, and “zeroes” popping everywhere (zeroes is an internal term we use when an app opens and there are no available cars). The surge algorithms would bring the prices back up, and we would again take prices down again.  The numbers beared out what we were trying to accomplish. Uber provided 60% more rides than our biggest day ever with the average fare at 1.75x (75% greater) than normal.

The whole experience was at once exhilarating and a bit defeating. We knew to keep cars available, we had to let the price go where it needed to. But the higher the price, the more vulnerable we were to a customer support nightmare.

In thinking about this, I have to admit that as a consumer if I were blindsided by this kind of pricing, I would annoyed.  And let’s face it, consumers will be blindsided. Who after midnight on New Year’s Eve remembers how much the car ride over cost? Further who can multiply that by 6.25 after a couple of glasses of champagne?

But from an operations economics perspective, I think this is totally awesome. On the one hand, it is pretty standard dynamic pricing akin to what airlines and car rental companies do, but the relationship to the drivers adds such an interesting twist. For airlines, capacity is basically fixed. If United always flies a 737 on some route, they are unlikely to change planes as demand picks up and only the price adjusts. Car rental firms have a little more flexibility on capacity and if an uptick in demand is recognized early enough, more vehicles can be sent into the market. Importantly, the vehicles have no choice. The people doing the pricing also control the capacity and can choose to add supply.

For Uber, though, they do not control the capacity directly and have to keep drivers interested. Further, drivers can game the system. A given driver might be willing to take a ride at four times over the normal rate but if the premium was six 20 minutes ago, why not wait to see what happens with the price? It is an intriguing complication to the problem.