Physical restaurant or dark kitchen: which one pays off · 2026 decision white paper

Chapter 1 — Brick-and-mortar restaurant or dark kitchen: which one suits you, based on where your margin comes from?

The right format is the one that controls access to your customer, and that rarely matches the one that costs less to open.

A dining room buys visibility once, the day you sign the location, and then amplifies it for free through Google Business Profile, Maps and reviews; a dark kitchen rents that visibility every month at a price the platform sets, with no vote from you. The scale of the intermediary explains why the price never drops: Delivery Hero closed 2024 with GMV of €48.8 billion, up 8% year over year, according to its annual results report. No individual operator negotiates against that. So the correct question is not where do I cook, but who decides tomorrow whether my customer finds me, and how much they will charge me to keep appearing on the first screen of an app that isn't mine. Setting up a kitchen in a shared hub costs a fraction of a full-service location, and that is where the advantage ends.

Chapter 2 — CapEx is one payment; commission is a perpetual rent

The saving is a one-time outlay; the 25% to 30% commission on nearly 100% of your sales gets charged in month one, month twelve and month sixty, and it grows as you grow. Run it with real cash: on annual sales of one million dollars routed entirely through an app, commission alone takes between 250,000 and 300,000 dollars, more than equipping two complete dining rooms. A location pays fixed rent, which inflation erodes in your favor while the check average rises. A dark kitchen pays a PERCENTAGE, which inflation amplifies against you. That asymmetry never shows up in a business plan, because CapEx is visible and variable rent is felt too late. With food cost inside the healthy 28% to 35% band the National Restaurant Association reports, input inflation of 5% is absorbed almost identically by both models: adjust the menu and move on. At 12% the roads split.

Chapter 3 — How each format behaves under input inflation of 5%, 12% and 20%

The dining room holds levers the hidden kitchen simply lacks: it lifts the check with wine, dessert and the long table, it shifts the mix toward high-margin plates, and the guest accepts the adjustment because they are buying an experience. A dark kitchen competes in a list where a rival's price sits one thumb away, and every peso added cuts conversion. At 20%, a 33% food cost becomes 39.6% and the dining room still breathes if it kills waste; the hidden kitchen, carrying 28 points of commission on top, runs at a technical loss unless it renegotiates the entire menu. The real asset you buy or lose in this decision is owned discovery. Follow the counterfactual all the way: tomorrow the platform rewrites its ranking and your kitchen drops from third to twelfth in its category. Nothing closes, nobody gets fired, and yet you lose 60% of volume in a week, because 100% of demand came through that door.

Chapter 4 — Algorithmic dependency is a liability that never reaches the balance sheet

That same day, a location with five-star reviews still fills its Saturday. The channel's scale is deceptive: iFood moved 100 million orders in a single month, August 2024, and works with more than 380,000 partner establishments across 1,500 Brazilian cities, according to its own institutional data. You are not competing against three neighborhood kitchens. You compete against hundreds of thousands of listings inside a search engine tuned for the platform's customer, not yours. The band decides more than the number of locations does, and here is my recommendation by tier. Below 500,000 dollars a year, a small dining room with neighborhood delivery of your own: a 28% commission on that base eats the owner's salary, so a pure hidden kitchen cannot hold. Between 500,000 and 1 million, the hybrid wins: dining room as brand and review anchor, a second virtual brand out of the same kitchen to fill the three-in-the-afternoon valleys.

Chapter 5 — What suits you according to your annual revenue band

From 1 to 5 million, the dark kitchen's real economics appear, because there is volume to negotiate supplies and absorb an operations director. Above 5 million, the hidden kitchen stops being a format and becomes a production unit with its own accounting, its own P&L and its own channel risk. Above 5 million and especially above 10 million a year, the celebrity-chef restaurant or the large-format themed venue plays a different sport, and the debate ends there. Its margin is not born on the plate but in the staging, the reservation booked three months out and the name over the door; pushing that into an app destroys exactly what the guest is paying for. Its own costs confirm it: brigade payroll with sommelier and maître d', design CapEx that in large formats comfortably exceeds 2 million dollars, and a seven-to-eight-year amortization no 18-square-meter kitchen can carry.

Chapter 6 — The high end above 10 million: when the dining room IS the product

A dark kitchen belongs here only as an auxiliary line, for packaged product or a casual sister brand, never as a replacement. When the dining room is the product, the hidden kitchen cannibalizes the promise. My judgment, after years working with owners who arrive at this question holding a lease: for the middle band the correct format is hybrid, with the dining room owning the brand and the hidden kitchen amplifying utilization. The mistake that repeats most is building it backwards, launching the virtual brand first and expecting the dining room to prop it up later. Mexico City has more than 1,200 active dark kitchens, growth of 40% since 2023 according to CANIRAC (2025), and most were born without a brand of their own. At Masterestaurant we call this buying volume by renting identity. Diego F. Parra puts it in a line that irritates some people: if your customer cannot pronounce the name of your kitchen, you do not own a business, you are a supplier to an app.

Chapter 7 — The decision, in three numbers you can pull this week

Before signing anything, measure three things in your own cash register and decide from there. First, what share of current sales comes through owned channels versus platforms: below 30% owned, you are already dependent, and a dark kitchen makes the picture worse. Second, your table check average against your delivery check; if the table beats delivery by more than 40%, your margin lives on experience and the dining room is the machine. Third, your break-even with commission included, not without it. The market will keep growing with or without you: the meal delivery segment in Latin America will surpass USD 39 billion by 2027 according to Statista (2024). That volume is an opportunity for whoever arrives with a brand, and a margin trap for whoever arrives with only a kitchen. The structural difference is not rent, it is who controls discovery. A physical restaurant buys visibility once, when it signs the location, then amplifies it through Google Business Profile, Maps and five-star reviews; a dark kitchen rents visibility monthly at a price the platform sets unilaterally.

Chapter 8 — Four differences that decide the format

With Delivery Hero reporting €48.8 billion in GMV for 2024 in its annual results, this is an intermediary whose bargaining power no single operator matches. CapEx misleads because it is one payment while commission is perpetual rent. Opening a dark kitchen from scratch in a shared hub costs a fraction of a dining-room location, yet 25-30% commission on nearly all sales is charged in month 1 and again in month 120. A dining-room location pays 8-12% rent on sales, and that rent does not grow with volume. Tolerable food cost differs by format, and almost nobody adjusts the menu. The National Restaurant Association places the optimal range between 28% and 35%; in delivery-only operations, with commission and packaging layered on top, 34% food cost pushes EBITDA into negative territory. That number sinks operations that looked healthy on the dining-room spreadsheet. Network density changes the game for capitalized players.

Chapter 9 — Four differences that decide the format — in practice

Blinkit runs roughly 2,100 dark stores and plans 900 more by March 2027 according to Storyboard18 (2025), and that density logic applies to kitchens too: three well-placed ghost kitchens cover a metro area with delivery times a single location never reaches. For a group above USD 5 million in annual revenue, that argument is real; for a single-unit operator, it is noise.