Physical restaurant or dark kitchen: which one wins once you read the 2026 numbers

The aggregator commission decides the model, not the rent

Brick-and-mortar restaurant or dark kitchen comes down to contribution margin BY CHANNEL, and that calculation starts by admitting the aggregator takes 18% to 30% of the order value before you touch a cent. Take the case I run into most often: a 9 USD ticket, a 27% commission, a 4% payment gateway fee and 0.80 USD of packaging leave 5.52 USD to pay for food, cook, gas and profit. At a 30% food cost on menu price, 2.70 USD walk out the door, leaving 2.82 USD to cover labor, energy and the shared-kitchen rent. The arithmetic forces the answer: a pure dark kitchen needs food cost below 26% or the channel works for the aggregator. Dining-room rent, the first line everyone subtracts, is the least decisive number in the equation. In the physical restaurant the customer is yours; in the dark kitchen the customer belongs to the aggregator, and that asymmetry explains why commissions never drop over the years.

Who owns the customer in each model?

Rappi knows that person's purchase history, delivery address and discount sensitivity, while you see nothing but a ticket and a pickup time.

Uber Eats closed 2024 with 26.1% of the US delivery market according to Earnest Analytics, iFood holds 87% of e-food bookings in Brazil according to Statista 2024, and in Spain Glovo sits near 31% against Just Eat's 26% according to Ken Research 2025. When two platforms add up to more than half the channel, the intermediary sets the price of access to that database. You are not paying for logistics, you are renting a commercial relationship that never becomes yours. The dark kitchen is not cheaper, it is more VARIABLE, and that difference completely changes which operator should choose it. Swapping fixed rent for a percentage commission means a slow month hurts less because cost falls with sales; an excellent month yields less because every extra order still pays its 18% to 30%.

Cost structure, not cost level

The physical restaurant works the other way: it leverages. When the dining room fills at nine at night, table number forty drops almost entirely into contribution margin because rent, the manager and base payroll were already covered by the earlier tables. An operator growing fast earns more with a dining room; one starting with thin cash and uncertain demand survives better without one. The right question is not which costs less per month, but which sales scenario you plan to live in over the next twenty-four months. Your delivery volume is decided by a ranking you do not control and that can shift within a week without warning. The algorithm at Rappi, Uber Eats, DiDi Food or iFood sorts by conversion, prep time and order acceptance rate, so three straight cancellations from a slow kitchen are enough to sink your position and 40% of the flow with it.

Algorithmic dependence never shows up in the spreadsheet

Contrast that with the Google Business Profile listing of a place with a dining room, which works overnight: somebody searches «restaurant near me», arrives through Maps, eats, leaves a review and that review lifts local ranking cumulatively. A dark kitchen never builds that asset with the same force because it has no verifiable address with foot traffic. The physical venue takes hits too, yet it keeps a demand floor: the people walking past the door. The dark kitchen's cost advantage erodes as competitors pour into the same channel, and the saturation figures are already visible. CANIRAC counted more than 1,200 active dark kitchens in Mexico City in 2025, a 40% jump over 2023, all fighting for the same top slots in the same app during the same lunch and dinner windows. Statista projects the meal delivery segment in Latin America will pass 39 billion dollars by 2027, a figure that pulls in capital and multiplies supply faster than a city's appetite grows.

The market fills up while you decide

Grand View Research reports that 61.7% of cloud kitchen revenue in 2025 comes from the independent segment, meaning operators with no marketing muscle. Entering a saturated channel late with a 30% food cost is the recipe for a quiet shutdown fourteen months in. Place yourself in one of three scenarios before signing any contract. Running a SMALL venue under 40 seats with a ticket below 10 USD, the pure dark kitchen almost never adds up: at 27% commission you need food cost between 24% and 26% and a menu under twelve items, or better yet treat delivery as a secondary channel of the dining room that already pays your fixed costs. Managing a MEDIUM operation of 60 to 120 seats with a 15 to 25 USD ticket, the hybrid wins: the dining room leverages Friday and Saturday peaks while a second virtual brand from the same kitchen absorbs the dead Tuesday shift with no extra rent.

How to read these numbers in YOUR operation?

Directing a GROUP of three venues or more, the dark kitchen works to test new territory for a fraction of what a venue costs, and there the 18% to 30% commission pays for itself through the demand data it returns.

The figures quoted here come from verifiable public sources, and their limits deserve a look before anyone uses them as a forecast. Market share data comes from Earnest Analytics 2024 for the United States, Statista 2024 for Brazil and Ken Research 2025 for Spain; local saturation from CANIRAC 2025; cloud kitchen market structure from Grand View Research 2025. None of them measures your city or your category specifically, and national market shares hide brutal differences between neighborhoods: Uber Eats leads the US average at 26.1% and loses badly in several individual cities. The 18% to 30% commission ranges are what the platforms publish and are negotiable by volume, not a constant.

Where these benchmarks come from and what they don't tell you?

Use these numbers to size the order of magnitude of the risk, never to project your own sales. The decision framework we apply at Masterestaurant starts right there:

Diego F. Parra first calculates contribution margin by channel using the real register data of the last ninety days, and only then compares models. I used to recommend dark kitchens to low-ticket operators because the savings on fixed costs looked obvious, and I was wrong. I never calculated the acquisition cost of a new customer inside the aggregator, which is exactly where the margin evaporates: in-app advertising to reach the top of the list eats 5 to 12 extra points of the ticket in competitive categories, on top of the 18% to 30% commission. An operator doing 400 monthly orders at 9 USD bills 3,600 USD and hands over 1,000 USD or more to the channel before buying the first kilo of chicken.

The mistake that took me years to correct

If tomorrow the aggregator raises its commission two points and a competitor launches a subsidized 30% discount, where do you find the margin to answer? In the physical venue you answer by cutting waste and raising table turnover; in pure delivery, only by raising price or shrinking the portion. Calculate your contribution margin by channel for the last ninety days this week, before signing anything. WHO OWNS THE CUSTOMER. In the physical venue the customer is yours: they arrive via Maps, return because they liked it, and you know their face. In the dark kitchen the customer belongs to the aggregator — Rappi knows their order history, their address and their discount sensitivity, and you see a ticket. That asymmetry explains why commission never drops over time: you are not paying for logistics, you are paying for access to a database that will never be yours. COST STRUCTURE, not cost level.

The four differences that settle the case

The dark kitchen swaps fixed cost for variable cost, which makes it tougher in a bad month and flatter in a good one. The physical restaurant leverages: once the room fills, every extra table drops almost whole to margin because the rent is already paid. An operator growing fast earns more in the physical venue; one starting with uncertainty sleeps better in the dark kitchen. SPEED OF LEARNING. Testing three concepts in a physical venue costs three build-outs. Testing them as virtual brands on one kitchen costs three menu photo shoots and a week of configuration inside the aggregator dashboard, which is the real reason chains use dark kitchens: not to save rent, to iterate the menu fast and keep the winner. MARGIN CEILING. A well-run physical restaurant reaches 12% to 18% net profit with a full room and delivery as a complement.

The four differences that settle the case — in practice

A pure dark kitchen rarely holds above 8% to 12%, because commission is a permanent tax on every peso sold; the only way through that ceiling is moving a share of sales to owned channel, and that demands the brand and the reviews only the physical venue produces.