The Uber of X Will Be Uber
There is a great deal of ruin in a [corporation]
- Adam Smith, approximately
Traditionally, Uber has a very happy New Year indeed. Midnight to five AM on January 1st is one of the peaks of inelastic demand for a ride home, so they usually start the year having just hit a milestone. This year, it’s been downhill since then. A timeline:
January 28th: Uber eliminates surge pricing at JFK after a taxi strike. This led to somewhat confused protests; some tweets claimed that they were cutting prices during the strike, but their price cut was after the strike ended. A garbled version of the story spread quickly on Twitter, making #deleteuber a trending topic.
February 19th: former Uber engineer Susan Fowler Rigetti publishes an explosive blog post about her experience of sexual harassment at Uber and management’s refusal to address it. Three days later the New York Times broadly corroborates the story.
February 23rd: with their usual flair for dramatic, well-timed PR, Google sues Uber for stealing trade secrets. They cite evidence of one Google employee stealing designs before defecting to Uber’s self-driving car subsidiary, and imply that several others did, too.
February 24th: the NYT, citing several Uber employees, says that an Uber self-driving car ran a red light, even though Uber had previously blamed human error.
February 27th: Amit Singhal, Uber’s newly-hired SVP of engineering, resigns after it was revealed that he had left his job at Google due to allegations of sexual harassment.
February 28th: Bloomberg publishes a video of Uber’s CEO arguing with a driver about falling fares. (Kalanick apologized for losing his temper the next day.)
March 3rd: the NYT reveals Uber’s “Greyball” tool, which makes it difficult for regulators to execute sting operations by ordering Ubers. In fairness to Uber, the original purpose of the tool was to prevent drivers from being poached by rival ride-sharing companies, and in even more fairness to Uber, building software to help small business owners not get arrested over a transaction of ambiguous legality is not exactly an undiluted evil.
On the same day, Uber’s VP of product resigns over sexual harassment allegations.
March 4th: Seizing the opportunity, Lyft starts to raise $500m in funding at a $6–7bn valuation.
If it were a public company, how much would Uber’s stock be down since the start of the year? Signet Jewelers just got accused of rampant sexual harassment, and of accounting chicanery, and its shares are down around 12%. On the other hand, Signet is a mature business, and (ostensibly) profitable.
A public company with $5.5bn in revenue and a trailing price/sales ratio of about 12.5x would fall pretty fast on this much bad news — maybe by half or more. But that would still leave Uber as a $~35bn market cap company, roughly tied with Didi Chuxing for most valuable ride-sharing company.
So in a counterfactual world where Uber is public, this hasn’t been a great week. But in that same counterfactual world, no one is making the case that Uber is a zero.
What has to happen to get Uber down to zero?
They lose scale, on both sides of the market. Passengers need to permanently abandon Uber, and drivers do, too.
They lose a huge amount of technical talent.
Crucially, they fail to fix the problems they’ve already promised to fix.
Uber’s senior management has apologized for several of these incidents, and they’ve appointed investigators. They will probably find more problems; additional heads will roll. They will lay down the law, and the wild parties and anything-goes atmosphere will fade into company lore. Uber sounds like it was run as a frat house, and frats don’t seem to scale to thousands of members very easily.
Meanwhile, even if they lose a chunk of their passengers and drivers today, they still have more scale than Lyft or most of their overseas competitors. As long as most drivers take the Uber ride first and most passengers open the Uber app first, their market position relative to competitors will be stable.
Right now, people who work at Uber are implicitly on team Old Uber. But when they replace their SVP of engineering and VP of product, they will be hiring to signal that they’re not the rowdy, rule-breaking Uber of old. This is a transition plenty of companies go through: the person who hears “That’s impossible; you’ll never build an app that can do that” and treats it as a challenge is a valuable employee. If they react the same way to “That’s impossible; if you get caught you’ll go to jail” then you have a problem. But the technical challenges have to come before the legal challenges: one way to not to get in trouble for running a quasi-legal ride-sharing service is to be unable to build a quasi-legal ride-sharing service.
And that means that some of the people who are leaving now would have left anyway, with different timing. Employees 1–50 of a world-changing startup are, statistically, pretty wacky people. Employee #5001 of the same company is going to be tamer. A company with lots of upstanding workers who do a solid job can’t afford the risk of having loose cannons, so those early employees don’t often make it to the IPO. Of Snap’s top seven executives, for example, two are founders but three joined in 2015 or later.
The Uber story is getting tired; journalists will move on to a new target . There’s nothing tech journalists love more than a heel turn, when a beloved company suddenly starts misbehaving and breaking hearts. Even if Uber’s behavior has been worse than Lyft, there are a lot of rocks to look under. (To start: if The Upstarts is to be believed, Uber initially refused to compete with Lyft because they were convinced that Lyft’s ride-sharing business was breaking the law.)
Scale Is An Advantage
If Uber is wounded but not dead, Uber will win. The logistics business is a scale business: your ride count sets a limit on how fast your product can improve, and the quality of the product determines how many rides you can get from each incremental dollar of marketing expense.
Consider the problem of minimizing wait time for food delivery. Uber offers UberEATs, a competitor to GrubHub, Postmates, Eat24, etc. Open the app, send in your order, it arrives in about half an hour. As anyone who has received delivery at work in Manhattan knows, the thirty seconds you spend loitering in the lobby feels a lot longer than the thirty minutes you spent waiting for your chicken tikka masala to get cooked and sent your way. And your delivery guy feels the same way: he gets paid to deliver food, not to wait for someone to show up and collect it. If you take forever, it’s money out of his pocket.
So Uber has a strong incentive to figure out, down to the second, who will arrive in the lobby when. And they can do this! Instead of relying on GPS, Uber can use wifi strength to triangulate their driver’s exact location and speed. And Uber can triangulate among the wifi networks available on your phone to figure out what floor you’re on. Over time, Uber can get a sense for how quickly each building’s elevators move, and how quickly any given user actually gets up to pick up their food.
And there are thousands of micro-optimizations like this:
Users traveling to law firms, investment banks, or Michelin-starred restaurants may be more sensitive to the quality of the car they’re stepping out of, so Uber can tweak the algorithm to favor sedans for those users.
If Uber can always track your location, they can get a sense of how likely every user is to substitute Uber for walking in inclement weather, and they can adjust their surge pricing in advance of rain based on this.
When current time + expected trip length = a round number like 5pm, the passenger is probably desperate not to be late. And some drivers are faster than others. For trips that are close to that band, Uber can tweak their algorithm to emphasize driver skill/aggressiveness over other factors.
Flight landing times and delays are available online for free. Uber can estimate the percentage of passengers for each departure/destination pair who are likely to want an Uber, and how long it will take them to get to the taxi stand, and adjust traffic patterns accordingly.
All of these micro-optimizations are, indeed, micro. But there’s an unlimited supply of tiny tweaks, and whoever has the most riders and drivers can test them fastest. Any one of these might be a .2% improvement, but 500 .2% improvements compound out to 2.7x. Good luck competing with someone who does the same thing you do, just 2.7x better, and is improving all the time. Or, put another way, if you find a million opportunities to save people thirty seconds, you’ve saved humanity an entire year.
The Uber of X
Startups like to pitch themselves as the Uber of something, both because it’s pithy and because Uber’s seed investors made several thousand times their initial investment. (It was apparently 2,000x back when Uber was worth a mere $17bn.)
Any on-demand service needs two things:
A huge number of users who have downloaded the app, entered their payment information, and hopefully opted in to accept push notifications.
A logistics network to deliver whatever it is that’s being delivered on demand.
For all the consumer-facing applications I can easily think of, Uber has #1 and #2.
So there are two kinds of Uber of X: there’s the kind that won’t work (“We’re Uber, but for portraits done in the style of the Old Masters!”) and the kind that will work better for Uber than it will for anyone else (“We’re the Uber for dogs,” “We’re the Uber for cleaning,” “We’re Uber for food, we can sell to anyone with an Uber account, and an Uber courier will pick up your meal thirty seconds after it’s ready,” )
As more vehicles on the road become Uber-controlled, they can move on from micro-optimizations — the Lyft-killers and GrubHub-killers — to macro optimizations. With enough cars on the road, Uber can actually identify traffic jams before they start, and combine rerouting (short term) and repricing (longer term) to prevent them. And while people tend to underpay for it, the benefits of a shorter commute are massive. You don’t need everybody to pay full price for a shorter commute; as housing prices in Manhattan and San Francisco demonstrate, the market-clearing price of a short commute is extraordinarily high. (Granted, there are other perks to living in the most expensive neighborhoods of these cities. But there’s nothing to do in Midtown or FiDi except go to work, and real estate prices in both of those neighborhoods are plenty high compared to other boroughs.)
It’s this kind of strategy that permits, or even requires, Uber to show massive losses. The sooner Uber controls half of the traffic in a city, the sooner they can start managing traffic more efficiently. The more users they have on the Uber app, the bigger their addressable market is for the Uber of X, Y, and Z.
When the Uber S-1 drops in 2018 or so, their domestic quarterly numbers will show a little wiggle in early 2017. The executive biographies will show a suspicious number of senior employees’ tenures starting around that time, too. They might go into some detail, on how they hit hard times. But there have been other hard times; their offices have been raided by the French police, drivers have rioted and sent death threats, and they’ve fought bans or the threat of bans in dozens of cities.
But Uber will still be able to tell the same story they’ve been telling investors from the start: it’s a product users love, with a scale advantage over everyone else. The profits aren’t there yet, but Uber is building new product categories and monopolizing old ones, and copying Uber is even more expensive than building it in the first place. Their competitive moat is a weird one with a weirder backstory, but it’s not going away.