Masterestaurant analysis of dark kitchen unit economics 2026: physical restaurant or dark kitchen, which one pays off

Finding 1 — Brick-and-mortar or dark kitchen: which one pays off?

The model that pays off is the one leaving you more CONTRIBUTION MARGIN per dish after the aggregator takes its cut, and that number rarely matches the owner's gut feeling.

An owner wrote to me in January with a two-page plan to close his 90-seat dining room and sell only through apps, and when I asked how much he kept per dish after commission, he had no answer; nobody does until they run the math. Markets did move: cloud kitchens grow at 12,6% a year according to Grand View Research (Cloud Kitchen Market), dark kitchens at 12,7% between 2025 and 2033 according to Global Growth Insights, with a projection of 248,10 billion dollars by 2035 according to Precedence Research. Market growth and your kitchen's profitability are two different things, and confusing them gets expensive fast. Closing the dining room erases nothing, it turns costs into variable ones, and that is the trap I keep finding in the conversion plans that reach my desk.

Finding 2 — A dark kitchen doesn't remove costs: it relocates them

Out goes the dining-room rent and the front-of-house payroll —fixed costs you dilute as sales climb— and in comes the channel commission, charged as a share of every order and therefore never diluted. With food 35% more expensive than in 2019 and labor cost also 35% higher according to the National Restaurant Association (2024), a hidden kitchen starting with a thin margin gets WORSE as it sells more, because each extra order drags its own commission along. The physical location has the opposite problem: reaching break-even hurts, yet once crossed, every additional peso lands almost whole. Two opposite slopes, not two versions of the same business. Run break-even for both models before signing anything, because they cross at different moments and with different consequences. Assume 120 pesos of sales per dish with 35 in product: in the dining room, with no commission, 85 remain to cover fixed costs; through an app, with a typical cut of a quarter of the sale, 55 remain.

Finding 3 — Break-even moves in opposite directions

So you need to sell well over a third more in volume just to restore the same peso of contribution, and with menu prices at large United States chains up 42% between 2020 and 2025 against 22% general inflation according to One Haus, raising prices to absorb the commission is no longer a free move. The Masterestaurant framework sorts this as prime cost, contribution, break-even and territory risk, in that order. A pure dark kitchen hands the customer relationship to the aggregator, and no income statement ever shows that patrimonial risk. iFood closed 2024 with 55 million active customers according to its own institutional data, an enormous base belonging to iFood, not to your kitchen; the day the algorithm drops you two positions in the listing, sales fall without you having changed a single dish. A physical location protects something the balance sheet omits: your own measurable reputation. Each additional star in review ratings is worth between 5% and 9% of revenue according to Michael Luca's research at Harvard Business School, and that lever stays in your hands.

Finding 4 — Territory risk: when the customer isn't yours

Ask yourself what happens if the aggregator raises its commission three points tomorrow. If the answer is «I close», the model already decided for you. Dark kitchens win as a SECOND brand feeding on idle capacity in a kitchen you already pay for, and lose when they are the only brand. I got this wrong for years, recommending dining-room closures over rent: rent was the most visible line in the lease and the least decisive in the result. If your kitchen runs at 60% during the afternoon shift, a virtual brand filling that gap adds no rent and no head chef, only product and commission, and there the contribution margin does survive the channel's discount. Ghost kitchens grow 11,65% a year between 2022 and 2032 according to Statista via Toast, and that growth goes mostly to operators whose infrastructure is already paid. Whoever builds a new kitchen purely to sell through apps pays twice: the investment and the commission.

Finding 5 — Your own channel breaks the tie

When both models show similar margins, the winner is whoever owns a direct channel, because every point of direct sales arrives with no commission attached. A personalized email lifts open rates 26% according to Stripo (Restaurant Email Marketing Statistics 2025), a creator post raises next-week bookings 30% according to Marketing LTB, and a complete digital offer of menu, ordering and payment lifts the ticket between 20% and 30% according to Sunday (QR Code Ordering 2025). Add them up: levers the aggregator never invoices. They work just as well in the dining room, where kiosks lift the ticket between 8% and 15% versus the counter according to QSR Magazine (2024). Diego F. Parra frames it this way in Masterestaurant audits: first measure what share of sales flows through a channel you control, then argue about format. Start with the numbers, never with the format, and the whole matter resolves in two weeks of real work.

Finding 6 — The right order to decide within two weeks

Take your last 90 days, split sales by channel and calculate the contribution margin of each dish in the dining room and in the app, using the exact commission written in your contract rather than the one you remember. Then compute break-even for both scenarios and study the slope: how much peso 101 above break-even contributes in each. If your kitchen is already paid for and has dead hours, launch a virtual brand without touching the dining room. If app contribution cannot cover your fixed costs even at double the volume, closing the room is a deferred sentence, not a saving. And if you close anyway, negotiate the commission first: three points less rewrites the entire arithmetic. A dark kitchen does not remove costs, it RELOCATES them: dining-room rent and front-of-house payroll leave, channel commission arrives, and commission is a variable cost proportional to revenue.

Finding 7 — The differences that actually change the outcome

Fixed costs dilute as you grow; variable costs never dilute. That is why a ghost kitchen with a thin contribution margin gets worse as it sells more, and it is the most repeated error in the conversion plans that reach my desk. Break-even moves in opposite directions. The physical location needs more revenue to clear it, but every peso above the line contributes heavily. The dark kitchen clears it sooner and contributes little above it. With food inputs 35% more expensive than in 2019 according to the National Restaurant Association (2024), that difference in slope decides whether the year closes with EBITDA or with an explanation. Average check is the dining room's non-negotiable advantage. At the table there is a physical menu, a server, a suggested dessert; menu psychology techniques lift check 15% or more without touching prices according to NeatMenu (Menu Psychology 2026). Inside an app you compete with a photo and a price, sorted by an algorithm that does not work for you.

Finding 8 — The differences that actually change the outcome — in practice

Reputation behaves differently under each model. Harvard Business School (Michael Luca) measured that each additional star moves 5% to 9% of revenue; in the dining room you repair a bad experience at the table, guest in front of you, while in delivery you find out once the rating has already dropped and the ranking has already punished you. Channel growth is real but bounded, and it deserves to be read without enthusiasm: 12.6% a year in cloud kitchens according to Grand View Research and 11.65% in ghost kitchens according to Statista/Toast (via OysterLink). That is a market growing at double digits, not one guaranteeing unit profitability. Market growth and your kitchen's margin are two things operators confuse constantly. Price defense already happened at the top of the market and almost nobody replicated it below: chains raised menu prices 42% between 2020 and 2025 against 22% general inflation according to One Haus. If you skipped menu engineering per channel during that window, the aggregator is charging commission on a price that stayed in 2021.