Selling on Delivery Apps: Commission, App Pricing and Margin in 2026

What commission do restaurants pay on delivery apps?

The most common band runs from 20 to 24.9 % of the order: 28 % of restaurants using third-party delivery pay an average commission in that range, according to the survey the National Restaurant Association submitted to the FTC in May 2026.

That figure is where any pricing sheet should start, because a dish costed for the dining room has no room to hand over nearly a quarter of its price without its contribution collapsing. My reading, after reviewing platform contracts in very different kitchens, is that owners rarely know which band they sit in: they remember the advertised commission of the basic plan and forget the charges for promotions, card payments or placement that pile up in the payout. Before you touch the menu, ask for the breakdown of your last four payouts and calculate the EFFECTIVE commission on the subtotal, which is the one you are actually charged.

The customer fee costs you too

In New York, consumers paid the apps an average of 7.26 USD per delivery in the first quarter of 2024, according to the data the Department of Consumer and Worker Protection collects from the platforms, and that fee never shows up in your payout but it does weigh on the customer's decision. When the dish already costs more to cover the commission and the buyer also sees a service fee and a delivery fee, conversion drops, and the platform then suggests a promotion that you pay for. That closes the loop that destroys the most margin in the channel: inflated price, hesitant customer, discount funded by the restaurant. The way out is menu engineering, with mid-to-high ticket dishes that absorb the surcharge without looking expensive and bundles that raise order value, because the customer fee weighs less the larger the subtotal it is spread across.

Why is delivery no longer an optional channel?

It is no longer optional because it weighs more than before: 70 % of limited-service operators with delivery say it now represents more of their sales than in 2019, according to the NRA's 2025 Off-Premises Trends report.

The broader picture points the same way, since the association estimates that 3 out of 4 restaurant orders in the U.S. are eaten off-premises. Here lies a tension worth resolving coldly, because the channel grows while margin per order narrows, and many owners react by switching the app off or, the other way around, pushing volume with discounts. Neither reaction survives a quarter. What works is treating off-premises sales as a business unit with its own P&L, separate from the dining room, where every order is measured by contribution rather than revenue, and where a dish that does not pay its share leaves the app menu even if it is the best seller in the dining room.

How many delivery apps should you be on?

As many as your kitchen can fulfill without errors, and almost never all of them from day one, even though 63 % of U.S.

limited-service operators offer delivery on three or more platforms (National Restaurant Association, 2026). Being on three apps multiplies the tablets at the pass, the menus to keep in sync and the payouts to reconcile, and every platform charges with its own fee structure. My recommendation is firm: join one, measure contribution per order for a full month, and only then open the second, with the app price already tested. If the second platform demands a different commission, the price is recalculated for it, because copying the menu from one app to another repeats the mistake of copying the dining-room menu to the first one, only now with two contracts working against you.

How to read these numbers in your operation?

For a small single-location restaurant, these figures call for caution: the app is a complementary channel, and a badly costed order takes up the same station at peak hour as a table paying full price.

For a mid-sized kitchen or a dark kitchen with two or three virtual brands, the app IS the dining room; there is no reference price to copy, so each brand needs its own formula, its own costed packaging and a short menu of its own. In a group with several locations the risk changes scale, because a poorly negotiated commission in the upper band multiplies across every unit, and the real lever becomes the contract and centralized menu engineering. My old advice to launch with the full menu for visibility was a mistake I took too long to correct. Today, under the MASTERESTAURANT method, all three profiles start the same way: a few dishes that travel well, each with its own price.

Source methodology and its limits

The figures in this piece come from two kinds of source, and it pays to know what each one measures before using it to set prices. The National Restaurant Association combines a survey of U.S. operators using third-party delivery, fielded in May 2026 and submitted to the FTC, with its Off-Premises Trends report and its analysis of industry costs. New York's Department of Consumer and Worker Protection publishes every quarter what the platforms report to it about orders, commissions and fees in the city. Neither describes a Latin American market, and a survey captures what owners declare, not their audited financial statements. That is why I treat them as a sense of scale and never as the rate in your contract. Commissions and fees change without notice with each plan, so confirm the figure in the official document before you move a single price.

The rule Diego F. Parra applies before turning on a promotion

A promotion is switched on only when the app price already leaves positive contribution with the discount included, and that is the rule Diego F. Parra puts first in any kitchen selling through platforms. For example, if a dish listed at 16 USD leaves 5 USD of contribution after commission and packaging, a 15 % promotion takes 2.40 USD and the order stays at 2.60 USD, still positive; with a price copied from the dining room, the same promotion sinks it below zero. One cost almost nobody puts on the sheet is deferred payment. The platform pays out days later, while you have already paid for protein, packaging and the week's payroll, and cash flow suffers even when the P&L looks healthy. At Masterestaurant that gap is planned in the thirteen-week cash flow, and the promotion is switched off as soon as contribution per order falls below target.

How to read these numbers in YOUR operation: three scenarios and what the customer sees?

Small restaurant, one location with a dining room. The app is a complementary channel and the risk is clogging the line at peak, since a badly priced app order takes the same station as a full-price table.

For years I told owners to list the whole menu for visibility, and I was wrong: start with six to ten dishes that travel well, each with its own price, and no promotions until you have measured a full month of contribution per order. Mid-size kitchen or ghost kitchen with two or three virtual brands. Here the app IS the dining room, so there is no dine-in price to copy and the mistake changes shape: owners price against competitors in the listing. Each brand needs its own formula, costed packaging and dark kitchen software that consolidates orders from every app on one screen, because a lost ticket on a stray tablet costs more than that order's commission.

How to read these numbers in YOUR operation: three scenarios and what the customer sees — in practice?

Group with several units or shared virtual kitchens. At this scale delivery is no longer marginal: 70% of U.S.

limited-service operators with delivery say it weighs more in their sales than in 2019, per the NRA's 2025 Off-Premises report. Negotiation is what changes, because with volume commission plans are discussed and app prices are set by region and daypart, never with one table for the whole network. What the customer sees on screen. Commission is not the only toll: New York consumers paid the apps an average $7.26 per delivery in early 2024 (DCWP), so a dish that costs more on the app competes against a fee that already inflates the order. Raising the price works when the photo, the name and the portion justify the ticket, and fails when the app dish is the dining-room dish with a different number.