Every order that comes through DoorDash or Uber Eats costs you an effective 30–40% of the ticket. The platform keeps the customer's name, phone, and history, so next time you hand over that cut all over again. What stings is that these people already love your food. They just found it through an app that charges you rent on the relationship. Below is the system we've watched operators run, stage by stage.
Stage 0: have somewhere to move them (week 1–2)
The migration goes nowhere without a destination that beats the marketplace on the basics. Minimum bar: a branded ordering app or ordering site with saved addresses, saved cards, and reorder in two taps. If ordering direct is slower than ordering on DoorDash, no discount will save it. On Dots the whole stack launches in about two weeks; whatever platform you're on, don't start printing inserts before checkout is smooth. Test it yourself first: ten orders, phone in hand, standing in the kitchen.
One rule tends to surprise people. Keep your direct menu prices lower than your marketplace prices, where you've probably marked things up 15–20% to survive the commission. That price gap is your best migration argument, and it happens to be an honest one, because a direct order really does cost you less.
Stage 1: the insert, your only ad channel inside the marketplace
Marketplaces block links and phone numbers inside their apps, but the bag is yours. Every marketplace order goes out with a card:
- A single line of value: "Same food, lower prices, 10% cashback."
- A QR code that lands on the app store or checkout, never your homepage.
- A code that expires in 14 days, so it gets used this week or not at all.
Consistency matters more than print quality. A customer who orders three times a month sees three cards, and the conversion adds up. Operators who keep the inserts going report moving a few percent of their marketplace regulars each month. That reads as small week to week, but over two quarters it decides the whole thing, since everyone who switches stops paying the 30% for good.
App, ordering site, cashback, and push — configured and live in about two weeks.
Stage 2: make the first direct order land perfectly
Treat the first direct order as an audition, and three things have to go right. The promised discount applies with no support ticket. The food arrives on time, which means routing these orders through your own delivery management with live courier tracking, because a late first order sends the customer straight back to the marketplace. And the cashback shows up in their account right away, clearly, with a balance they can watch grow.
That last piece is the hook. A discount closes the transaction; a cashback balance opens the next one. $3.20 sitting in your app is a standing reason to order from you instead of scrolling the feed.
Stage 3: the repeat loop, where the system earns its name
Now the customer is in your database, and the tools the marketplace never gave you switch on: a push when their cashback is about to expire, a "your usual Friday order?" nudge at 4 p.m., a win-back message when a twice-a-month regular goes quiet for three weeks. You'll catch that quiet stretch in your order reports long before you'd notice it at the counter. This is retention marketing that costs nothing to send, and it goes a long way toward explaining why branded apps convert up to 35% of visitors while a marketplace listing converts about 3%.
Just don't over-message. One or two pushes a week, each tied to a real reason: a restocked item, an expiring balance, comfort food when the weather turns. You're building a habit, and a habit dies fast if every message reads as noise.
The cadence: what "repeat" means in practice
The system runs on a weekly rhythm, and it's about as plain as it sounds. Monday, check three numbers: inserts sent last week (this should equal your marketplace orders, no exceptions), first-time direct orders, and second orders from earlier converts. Wednesday, run one push campaign to the existing base, tied to a real reason. Every day, cards go in bags and staff say the one-sentence script when they hand the order over. That's the whole operating load, under an hour a week once the habit sets.
The curve will feel slow for the first month and then bend. A convert in week two brings their next order in week four and their fifth by month three, while new converts keep stacking on top. Operators who run the numbers at day 90 usually find direct share grew faster than any single week would have suggested. If nothing has moved by day 30, the cause is almost always one of two things: inserts going out unevenly, or a checkout that's slower than the marketplace's.
What not to do
Don't quit the marketplaces the day your app launches. They still handle discovery better than anything you can buy, bringing you new customers who've never heard of you. Treat it as a division of labor: marketplaces acquire, your channel retains. And don't break marketplace terms by messaging their customers off-platform with scraped data. The bag insert works precisely because it's your bag and your kitchen, not theirs.
Start this week with an audit that costs nothing: count how many marketplace orders left your kitchen yesterday without an insert in the bag. Take that number, multiply it by 30 days, by your average ticket, and by 30%, and you have the size of the leak this system plugs.
We'll set up the destination — app, site, cashback, push — and show the migration numbers of real restaurants.