Ranjan Roy has a great article on Substack about DoorDash and “pizza arbitrage”: Roy’s friend was first annoyed to discover that DoorDash was providing delivery services for his nondelivery pizzeria: taking web orders without his knowledge, phoning in for takeout and sending a DoorDash delivery worker to pay and pick up the food, and often delivering to a customer who would be annoyed that the pizza arrived cold. And then he was surprised to see DoorDash was selling his $24 pizzas for only $16. This meant he had an arbitrage opportunity: Order his own pizzas at $16, sell them to DoorDash for $24 each, and pocket the difference. This worked even better if he didn’t put real pizzas in the delivery boxes. But how on earth was DoorDash ever supposed to make money selling his pizzas at a loss?
Delivery via smartphone is one of those venture-funded sectors where business executives appear to have taken seriously the old joke about “losing money on every transaction but making it up on volume.” Normal rules of capitalism about maximizing profits do not apply. This has led to a strange situation where restaurants feel squeezed by the fees charged by delivery services (when, unlike Roy’s friend, they participate voluntarily on a delivery platform) and yet the delivery services themselves manage to keep losing money. Why is this even happening?
A mental model that a lot of people have for these businesses is that they are waiting to establish a dominant market position, at which point they can raise prices to a level where they will be profitable. That is, in the future, restaurants and customers will pay even more in delivery fees, and DoorDash will make money. The problem with this view is that “the future” never seems to come. Uber has been providing rides for ten years. When does the “profit” phase of its business show up?
I think the missing element for profitability is different: productivity. The hope with a lot of business models that bring app intermediation to a preexisting element of the economy like ride services or food delivery is that technology will make workers more productive. You can see instances where this is obviously true: a Peloton instructor who teaches a class to tens of thousands of people is more productive than a SoulCycle instructor who can only teach about 60 people at a time. But with a lot of apps, the promised boost to productivity never materializes. The worker still has to render personal service to one customer at a time, and the app doesn’t do much to reduce the worker’s downtime or help him or her complete the task faster. As such, the productivity boost that is needed to make the financial model pencil — paying the worker a high enough hourly rate while charging a fee the customer is willing to pay and still having a positive profit margin — does not materialize.
Consider a few examples. In the traditional model, restaurants use their own employees to deliver food. DoorDash and its competitors offer a different approach: DoorDash contracts with the delivery person, sending him or her to whatever restaurant has orders at any moment. In theory, this should lead to better matching of labor to work: Restaurants don’t get backed up with too many orders, because DoorDash can send over extra staff as necessary; the restaurant also never has to pay a worker to sit around and not deliver food. But there are offsetting disadvantages to this outsourced model. A restaurant-employed delivery person knows the menu and can tell quickly whether a bag appears to contain what is listed on a receipt. He has a rhythm with the staff he’s picking up from. He knows the neighborhood and knows the addresses of frequent customers. He has the right equipment — if he’s delivering pizza, he has an insulated bag so the pizza is hot when it gets to the customer. A third-party delivery person is more likely to screw these things up: slower, less accurate, lower-quality delivery. At the very least, this mutes the productivity gains from better staff matching; it could offset them entirely.
Or consider Uber. The application is supposed to reduce drivers’ downtime by efficiently matching riders to drivers. In places where you would ordinarily have to phone for a taxi, the app technology probably does provide real, large advantages. In practice, there are a lot of places where technological matching is less efficient than in-person matching, especially airports, where inefficient Uber pickup has led to increasingly clogged access roads as drivers search for specific passengers instead of picking up whoever they see at the front of the line. Unfortunately, these places where the app is at a disadvantage compared to traditional taxi hails tend to be the places where demand for transportation services is highest: A key driver of Manhattan’s increased traffic congestion over the last decade is passenger-less Uber drivers, moving around looking for fares. And if Uber’s technology is so good at matching drivers and riders and capturing the variation in customers’ willingness to pay across time, why has the company had to repeatedly resort to paying drivers more than it collects from customers in order to ensure that enough of them are around at peak times?
An oddity of the on-demand economy is the assumption that people should be buying a lot more personal-assistance-type services than they used to. That’s what these apps do: let you hire people in small blocks of time to drive you around, pick up your food, clean your apartment, and so on. This is not a totally new model. Long before the internet, a lot of people who didn’t have the money for a full-time personal assistant would hire someone for a couple hours a week, especially to clean their homes. And as Roy notes, pizza delivery is a long-standing business model that many business owners have made money on for decades, before DoorDash and its competitors decided to lose money doing it. But does technology provide a reason that this part of the economy should be greatly reshaped and expanded — should every restaurant offer delivery now, because technology makes that financially viable in a way it was not before? Should we be hiring people to drive us places that we would have ordinarily driven ourselves? The “yes” answer relies on the idea that technology has significantly changed the value proposition for such services, either improving their quality so that we are willing to pay more, or improving workers’ productivity such that the services can be provided more quickly and therefore cost less.
But what if the main reason the value proposition for these services has changed is that a third party is weirdly willing to lose money on the transactions? That doesn’t seem like a sustainable situation — and yet it has been sustained for years at this point. If it ends, if investors in app-based service companies start demanding profits, then we should expect the size of the personal-service part of the economy to contract. Some restaurants that came to rely on app-based delivery may find it makes sense to take delivery in-house. But others may find delivery isn’t worth it if they actually have to employ the delivery person. And then customers, revealed to be unwilling to pay the true economic cost of having their food delivered, may have to go pick it up.
