Uber says it beat Didi on efficiency, not subsidies

The China fight cost Uber "tens of millions of dollars a month," Kalanick says. Travis Kalanick, the co-founder and former chief executive of Uber, says the point was never to outspend Didi, the dominant Chinese ride company. The point was to subsidize more cleverly, and use the subsidy to become so efficient that Uber could make money on a fare where Didi still lost money.

An efficiency edge, in his account, starts with small things. If it is easier to download the app, sign up, enter a credit card, and get a car to show up without a failed pickup, support costs fall and more trips get completed. If the driver app steers drivers toward where riders actually are, pickup times fall too.

That is where the compounding starts. Shorter pickups mean less unpaid waiting for drivers. If drivers are trying to earn a certain amount per hour, less dead time means the whole system can charge less and still pay them. A cheaper, faster service gets bigger faster, which puts more cars near every rider, which makes pickups faster and cheaper again. That loop is the network effect.

Once the loop is running, Kalanick says, the bigger and better operator can profit at a price where the rival must still pay subsidies. He compares Uber's focus on such small details to the playbooks of John D. Rockefeller, founder of Standard Oil, and Andrew Carnegie, the steel magnate. In Kalanick's telling of a story from the Rockefeller biography Titan, the point was one drop of solder saved on each oil barrel, after trying 48 drops and getting leaks and finding 49 would hold, worth $2,500 in year one and hundreds of thousands of dollars at scale.

There is a catch. Subsidies stop working at scale. Kalanick says you cannot pay $2 a ride on 10 billion rides a year because no one will fund $20 billion in losses. That gives a well funded small rival, perhaps one tenth the size, an opening. Its total subsidy bill is roughly one tenth as large, so it can cut prices and force the big company to burn ten times as much to answer, or seven times as much if the big company is more efficient.

So Kalanick describes how he would test efficiency rather than assuming it. One method, he says, would be to let a rival grow to an even 50-50 share, then see if Uber could hold half the market at a higher price. If it could, its system was truly more efficient. A big company could also attack asymmetrically by offering drivers from the rival, say, $1,000 for completing 100 trips in the first week. To keep those drivers, the rival had to raise pay across its entire driver base.

In China, that local math became global. Kalanick says Chinese sovereign wealth funds poured billions of dollars into Uber competitors in other regions, forcing Uber to subsidize on many fronts at once and drain its resources.