Read your DoorDash or Uber Eats agreement and it's spelled out plainly: people who order through the platform belong to the platform, not to you. Their name, phone, email, address, and order history stay on the marketplace's side. What reaches you is an order ticket: the items, a first name, sometimes a masked phone number that expires once the food is delivered. None of this is accidental. The diner relationship is the marketplace's most valuable asset, and it guards that far more carefully than it guards your margin.
What they see vs. what you see
The platform knows Sarah orders pad thai every other Thursday, spends $34 on average, and went quiet three weeks ago. It also knows she orders from the Thai place two blocks over, which is the part that should bother you. You get "S." and a masked number. That gap does more than feel unfair; the platform charges you for it. When Sarah opens the app, it decides whether she sees your listing or a competitor's sponsored one, then sells you back the visibility you earned by cooking good food.
The price tag on one invisible regular
Put numbers on it. A regular who orders twice a month at a $30 ticket is worth $720 a year in gross sales. When she drifts away on a marketplace, you never see it and can't respond, because she was never in your database to begin with. On your own channel it's routine: a lapsed-regular alert fires, you send one push showing her cashback balance, and it costs nothing. Operators say win-back campaigns recover a real share of lapsed customers. Take a conservative one in five, and the math per hundred lapsed regulars comes out like this:
And that's a single campaign type. Layer on birthday offers, "back in stock" pushes, and weather-driven promos, and the database turns into the most productive marketing asset you have. Which is precisely why the platforms keep it. Look at what the recovery costs, too: no ad spend, no agency, one notification you write in a minute. The whole mechanism is a list you own and a reason to message it.
We'll estimate it from your marketplace volume and show what a win-back program returns.
Second-order costs: the ones that surface later
Open a second location, launch a new menu, or roll out a family meal deal, and with no list of your own every announcement starts from zero or gets paid for through ads. You can't bring your customers with you, because you never had their contact details.
A buyer values a restaurant with a 10,000-customer database very differently from one whose "customer base" is a marketplace listing that can be switched off tomorrow. One is an asset on the balance sheet; the other is a rental agreement. That alone makes the business harder to sell.
The platform also studies your market and acts on what it learns. Aggregated demand data — what sells, where, at what price — feeds its decisions about promotions, dark kitchens, and private-label brands in your zone. You supplied the data and get no say in how it's used.
"But the marketplace gives me reports" — what you actually get
Marketplace dashboards hand you aggregates: order counts, gross sales, ratings, maybe a repeat-customer percentage. What they hold back is everything you could act on at the level of a single person: who those repeat customers are, how to reach them, when they slip away. Being told "40% of your orders are repeats" is not the same as being able to message those exact people an offer. One is a talking point; the other pays the rent. Some platforms even sell restaurants aggregated "insights" packages on their higher commission tiers, which reads oddly once you see it clearly: you're buying back a blurred summary of data your own kitchen produced.
The masked phone numbers are worth dwelling on. They exist so that even the handoff at the door, with your food and your customer right in front of you, can't become a lasting connection. The relay number goes dead the moment the order closes. Whatever you make of the policy, treat it as a measure of how much that connection is worth to the platform.
Rebuilding the asset, one order at a time
You can't export your history out of the marketplace; that door is shut. But every order you take from here on is a chance to capture the relationship yourself. The mechanics are simple. A branded app or ordering website writes a real customer record at every checkout. A cashback program gives customers a reason to identify themselves each time they order. And analytics turn those raw records into RFM segments, so you can see who's new, who's loyal, and who's starting to slip. After that, push and win-back campaigns run on their own. Across the 3M+ orders processed on Dots, this is what sits behind apps that convert up to 35% of visitors: a recognized customer with a saved card and a balance to spend just orders more often.
Getting marketplace customers into that database through bag inserts is its own article, the insert-and-repeat system, but the first move costs nothing. Starting today, treat every order that leaves your kitchen without creating a customer record as stock you gave away for free. Count how many you had yesterday. That number is your leak, measured in lost relationships.
App, site, cashback, and customer analytics — your database from order one, live in ~2 weeks.