Delivery commissions that kill your margin: which alternative is left for your restaurant in 2026

When the delivery app stops paying off?

The app stops paying off the day repeat orders cross 40% of your volume, because from that point on you are paying a discovery fee for a customer who already knew you.

That is the number that gives it away, and almost nobody checks it: pull the statement, count how many orders come from returning users, multiply that subtotal by your commission rate. With plans running from 15% to 30% depending on the tier you signed (CloudKitchens, 2024), a restaurant billing 62 million pesos a month with 55% repeat business is handing roughly nine million to the platform for introducing people it already had. Add the 15% co-funded discount the app switches on during promotions plus the payment-gateway hold, and you understand why only 41 million landed. This is not a sales problem. It is a pipe with the wrong diameter. On a 30% food cost —and 32% is our ceiling, never the target— a 28% commission becomes the heaviest line item after ingredients, sitting above kitchen payroll in most of the restaurants I review.

Commission is your second-largest cost, not a marketing expense

Line the arithmetic up and it turns brutal: of every 100,000 pesos sold through the app, 30,000 go to product, 28,000 to commission, and you keep 42,000 to cover rent, gas, packaging, the cook who plated the dish and whatever profit you hoped for. Diego F. Parra keeps hammering an uncomfortable point at Masterestaurant: that percentage is charged BEFORE you have paid a single peso of anything else, so it is not an elegant variable cost, it is a silent partner with first claim on the till. And the partner never plated a thing. Renegotiating works for one specific profile: the operator with three or more locations, solid volume and an assigned account executive, who can shift the plan from 30% down to 25% (CloudKitchens, 2024). Switching cost: essentially zero in cash, two or three weeks of emails and one meeting. The trouble is the arithmetic.

Option 1: renegotiate the rate (and why it yields so little)

Five points on the app channel, when the app carries 45% of your sales, return barely two points on total revenue, and those points usually arrive tied to exclusivity or to co-funding promotions, which sends the savings back out the rear door the following quarter. For the owner of a single location without volume, bargaining power against a platform holding thousands of restaurants in your city is, honestly, nonexistent. Treat it as hygiene, not as strategy. This is where most operators go wrong, raising prices flat, say 15% across the whole menu, which punishes the dishes that already performed and keeps bleeding on the ones that never did. The right move works item by item: a burger at 26% food cost absorbs a 12% platform markup without scaring anyone, while shrimp at 38% food cost needs 25% or simply should not sit on the digital menu at all. Profile: any restaurant with recipe cards for its fifteen best sellers.

Option 2: price by contribution margin, dish by dish

Switching cost: eight to twelve hours of costing work, zero investment. The effect lands immediately and depends on nobody signing anything for you. One warning from the field: never load payroll, rent or utilities onto the plate to justify the markup, because those belong to break-even, and folding them into costing leaves you with prices no customer pays. Treat the platform as a shop window rather than your entire restaurant. Publish eight to twelve items, the ones with the strongest contribution margin and the best tolerance for transport, and keep the rest for your own channel and the dining room. That fixes three things at once: average ticket on the app rises, waste from items selling twice a week drops, and prep time shortens, which is precisely what the platform rewards inside its ranking algorithm. Profile: kitchens carrying more than thirty menu items and running tight at peak. Switching cost: one afternoon of uncomfortable decisions, because there is always an owner's favourite dish that never made the numbers.

Option 3: short menu on the app, full kitchen at home

User penetration in meal delivery reaches 29.2% in 2026 according to Statista, so the shop window does matter; what does not matter is having it fully stocked. Here is the lever that genuinely moves the needle, and it is also the slowest. A properly worked Google Business Profile —weekly dish photos, exact hours, menu uploaded, every review answered— plus a direct ordering button through WhatsApp or your own site lets you shift repeat business into a channel where cost per transaction falls to 3% gateway plus the rider. The evidence backs it: 58% of customers prefer ordering through the restaurant's own app or website, according to NCR Voyix (Restaurant Dive, 2024). Profile: an owner willing to hold discipline for six months without seeing results in month one. Switching cost: between 400 and 900 dollars to set up, plus two hours a week from someone who answers reviews and posts.

Option 4: your own digital engine with Google Business Profile and direct ordering

Move 60% of your repeat business there and you recover fourteen to eighteen margin points across the whole business. No account executive will ever sign that. Picture the full scenario, because owners put it to me every month. You switch Rappi off on Monday. You lose 45% of your orders overnight, and of those you recover —optimistically, with a customer database in hand— about half within the first ninety days, provided you kept the phone numbers and worked the Google profile. The rest leaves with the platform, because that customer was never yours: they were an app user who one Tuesday ordered whatever sat closest. Sales fall 22% and PROFIT rises, since the orders left standing no longer pay commission. That is the paradox to settle before touching anything: more revenue with less profit is exactly the trap the Medellín owner sat in for fourteen months, celebrating the wrong kind of growth.

What would happen if you switched the app off tomorrow?

The right sequence never starts by switching off. Build the owned channel, measure migrated repeat business across two quarters, and only then decide whether the app drops to shop-window duty or goes dark.

Stay exactly where you are if you opened less than six months ago, if your brand carries no recognition in the neighbourhood, or if you run a dark kitchen with no dining room. In those three cases the platform sells you something you cannot buy any other way: cold traffic, people who have no idea you exist. Paying 28% for discovery when you hold no other source of discovery is not a mistake, it is the entry fee, and the global ghost kitchen market projected at 142.5 billion dollars by 2029 according to Research and Markets was built on precisely that logic. The signal to start migrating shows up when repeat orders cross 40% and your Google reviews pass one hundred.

When you should NOT change anything?

Before that point, building an owned channel eats management hours that pay better inside the kitchen. I got this wrong for years, recommending owned channels to restaurants that still had nobody to call.

The traditional operator negotiates the percentage; we change the mix. Cutting commission from 30% to 27% returns three points on that channel; moving 60% of reorders to an owned channel returns fourteen to eighteen points on the total, and no account executive will ever sign that for you. The traditional operator raises prices evenly, which punishes dishes that already performed while still bleeding on the rest. We differentiate by contribution margin, dish by dish, because a burger at 26% food cost and shrimp at 38% cannot carry the same platform uplift. The traditional operator reads Rappi, iFood or DiDi as a sales channel. For us it is a MEDIA channel, comparable to geotargeted advertising: you pay to appear in front of someone who did not know you, which makes sense the first time and stops making sense on that same customer's seventh order.

Five differences that decide the margin

The traditional operator ignores that the platform algorithm rewards availability, prep time and cancellation rate rather than the quality of the cooking. We work those three metrics deliberately because they lift ranking without co-funding a single discount, which is the only cheap way to increase sales on Rappi. The traditional operator holds not a single phone number from delivery customers. We build an owned database from day one, using packaging as the ad medium, and by month eight that asset is worth more than the listing position, because it survives any change in app rules.