DoorDash Is Buying Instacart. Its CEO Said Delivery Is Near Perfect From Its Own Shelves and Low 90s From the Store's. The Store Makes the Sale.
· CX Pulse
DoorDash's CEO put a number on delivery accuracy: near perfect from its own warehouses, low 90s from a store's shelf. The store's name is on the bag.
DoorDash agreed on Thursday to buy Instacart for about $17 billion. Every headline carried the price. The sentence worth reading came three hours later, on DoorDash's own earnings call, when an analyst asked Tony Xu why now. "Customers today are still being asked to pay a premium for imperfect delivery," he said. Then he put a number on imperfect.
Orders DoorDash fills from its own warehouses, the DashMarts, are "near perfect." Orders its couriers pick inside a third-party store run "only in the low 90s in terms of our accuracy." That's the CEO of DoorDash telling analysts that as many as one order in ten picked off somebody else's shelf arrives with something wrong, and that the fix he found was to stop using somebody else's shelf.
The deal buys the other answer. Instacart has spent fourteen years sending shoppers into grocers' aisles, and Xu's own release describes the job: "Someone has to pick a ripe avocado, find a substitute you'd choose yourself, and get the ice cream to you before it melts." Nothing changes for a customer yet. The two apps run separately until the deal closes, which DoorDash expects at the end of 2027 or the first half of 2028, after shareholders and regulators have had their say.
Whose name is on the bag
One more sentence from the release: "On either app, customers pick a store in their own community, and the store makes the sale." That's the structure of the whole business. The platform owns the courier, the substitution rules and the support queue. The store owns the sale, the receipt and the customer's memory of what arrived.
So the accuracy gap Xu described has an address. When the ice cream arrives melted or the avocado arrives hard, the customer remembers the store's name, and sometimes drives past it the next day. The store never saw the order picked and never heard the complaint. The one number that would tell the store how often this happens was said out loud to investors, and almost never to the store.
Xu was direct about where the value of the merger sits. Most of it, he told analysts, is "almost like not visible to the consumer." Denser courier networks, one tech stack, shared ads. The customer will see one membership instead of two, which DoorDash estimates saves members $100 million a year. What the customer won't see is whether the bag got better.
At scale: the number exists, so ask for it
If you run customer experience at a grocer or a retailer that sells through either app, the accuracy rate by fulfilment path is a number the platform keeps. It was precise enough to say on a call. Ask for yours, split by picked-in-store, picked-from-platform-inventory, and delivered-from-your-own-site. Put it on the same page as your contact reasons for "missing item," "wrong substitute" and "arrived damaged," because those contacts are the same number read from the customer's side.
Then look at who decides the substitute. On most platforms the courier chooses, inside rules the platform wrote, from stock the platform believes you have. Your out-of-stock data is now an input to somebody else's accuracy rate, and a stale shelf file produces a wrong order with your name on it. Ask how often your inventory feed is read, and what the app shows the shopper when it's wrong.
Last, the membership. One fee for two apps means a refund or a credit for a bad order will be decided under one company's policy for an order the other app placed. Write down what your own make-good is, so the customer who calls you instead of the app gets an answer that doesn't depend on which logo they tapped.
Smaller operation: the shelf file is the order
For a restaurant, a corner store or a florist on either platform, the low 90s is the share of your orders that reach the customer wrong, and every one of them carries your name. The input you control is the menu or the stock list the app reads. An item you sold out of at noon and marked at four produces four hours of substitutions you never chose.
Mark the 86 list before the rush, not after. Read your platform's complaint reasons once a month and count how many are about the order and how many are about the drive. And set your own make-good for a wrong order before the first customer walks in holding the bag, because the platform's credit lands in the app, and the customer's opinion lands on you.
Xu said the product isn't there yet, and that customers are paying a premium for it anyway. He is buying the company that has picked more groceries in other people's stores than anyone else. Until it closes, and probably after, the accuracy of a delivery is measured by the platform, paid for by the customer, and remembered against the store.
Sources: DoorDash announcement, Form 8-K exhibit, October 29, 2026 · DoorDash Q3 2026 earnings call transcript · DoorDash newsroom