Ghost kitchen business model: the mistakes that turn it into a trap, and the alternatives that actually leave cash

When does the pure ghost kitchen stop working?

The pure ghost kitchen falls short the day your contribution margin per dish drops below 62% before commission, and the number that gives it away sits in the aggregator's biweekly settlement, not in your income statement.

A 42-square-meter kitchen on a second floor, three brands inside Rappi, an average ticket of 11 dollars and 21,400 dollars billed per month: the owner swore he was making money, until we pulled the commission out of the sales line and 5,600 monthly dollars showed up that had never been his. That gap between the gross you record and the net the platform pays you IS the profit. With commissions running 18% to 30% and packaging at 0.45 to 1.20 dollars per order, a dish with 62% gross margin lands near 30% real, and payroll, rent and utilities still come out of there. Whoever removes the dining room also removes the only demand that charges no toll, and almost nobody budgets for that.

The founding mistake: removing the dining room moves the problem, it doesn't delete it

The pedestrian walking by, the guest who books a table, the neighbor with your number saved: none of them charges 28% to bring you an order. In exchange you inherit an entire funnel governed by an algorithm you cannot audit, which reshuffles results whenever it suits, and which can raise your commission with a Tuesday email. I got this wrong for years, recommending dark kitchens to operators sitting at 36% food cost with a low ticket, seduced by the cheap-rent argument when they were coming from 9,000 dollars a month on a street-level space. None of those projects made it past month eighteen. The survivors shared two things: a pickup window, however tiny, and real Google reviews. The hybrid format —a production kitchen plus a two-meter counter for pickup— offers the best effort-to-return ratio for the operator already billing above 15,000 dollars a month through aggregators.

Option 1 · Hybrid with a pickup window

The logic is ownership, not rental: in the pure ghost kitchen you rent customers order by order, forever, and somebody else sets that rent; with a window you buy the customer once, through Maps and reviews, and serve them again with no toll. The lever is measured: every additional star in your review rating moves between 5% and 9% of revenue, according to Harvard Business School (Michael Luca, Reviews, Reputation, and Revenue). Switching cost: façade work and signage, a permit to serve the public, and one extra shift during peak hours. Expected return: 12 to 25 points of direct, commission-free sales. Concentrating the operation into ONE brand with a proprietary ordering channel beats the three-virtual-brands play whenever the average ticket is low and the menus share ingredients. Three listings inside the same app split your reviews, split the order history the algorithm rewards, and triple the work of photography, listings and pricing, without adding a single new customer to the business.

Option 2 · One brand with your own ordering channel instead of three virtual brands

With one brand you stack up rating, and that is where direct email finally pays: the average email open rate was 25.1% in 2023 according to Omnisend, and personalized messages lift opens another 26% according to Stripo. Ideal profile: owner-operator with fewer than three kitchen employees and a base of a thousand contacts or more. Switching effort: cheap in money, expensive in discipline, because it means killing brands you love. Before changing the format, change the menu, because the same kitchen with a different mix rewrites the whole financial verdict. The food cost ceiling stays at 32% per dish —with no payroll or rent loaded onto the plate, since those belong to break-even— but in the pure ghost kitchen, with commission at 28%, every extra point weighs 39% more on the margin than in a physical restaurant serving the same menu. So the way out is pushing volume toward what is born wide: 46% of surveyed operators in the United States name alcohol among the highest-margin menu categories, according to Technomic for Nation's Restaurant News.

Option 3 · Menu engineering toward high-margin categories

If your license bars alcohol, the equivalents are prepared beverages, plated desserts and assembled combos. Switching cost: two weeks of recosting and new photography. Renting kitchen time in a commissary suits the operator who has not yet validated demand, and it is the only option that turns a fixed cost into a variable one without touching product quality. Opening a QSR or food truck in the United States cost under 150,000 dollars in 2024 according to Square; an hourly station runs orders of magnitude below that, and you pay only for the blocks you actually bill. The profile is clear: a founder with a proven recipe, no sales history and no capital for twelve months of rent. The downside is real and worth saying out loud: you do not control the oven schedule, you share the rhythm with other operators, and you cannot scale production the day demand shows up all at once.

Option 4 · Shared kitchen by the hour before signing your own lease

It proves a concept; it does not grow one. An operator who absorbs the commission without recosting loses margin twice, because the input side moved too. In Colombia menu prices rose 9.8% from February 2025 onward to sustain 98,000 jobs, according to ACODRES; in the United States food inputs are up 35% against 2019, with labor up the same 35%, according to the National Restaurant Association. Now work through what happens if the aggregator lifts commission from 22% to 28% on an 11-dollar ticket: that is 0.66 dollars less per order, and across 1,900 monthly orders it becomes 1,254 dollars taken out of profit, not out of sales. If your contribution margin sat at 62%, you hold. If it sat at 54%, that Tuesday email closes the business within a quarter. Recost BEFORE, never after. Staying in a pure ghost kitchen is the right call when contribution margin clears 62% before commission, the ticket runs above 18 dollars, and the kitchen already works above 70% of installed capacity during peak blocks.

When NOT to change format?

There the commission is a healthy acquisition cost, and building a storefront only adds rent, permits and a shift you do not need. Do not switch mid-high-season either, nor with a rotating crew:

each departure costs 150% of the salary in replacement, according to StaffedUp, and remodeling with new staff is two crises at once. The Masterestaurant method I work through with owners always starts in the same place: Diego F. Parra asks for the aggregator settlements of the last ninety days, not the POS report. Measure three months of real net before moving a single brick. The difference is not rent, it is WHO owns the demand. In a pure ghost kitchen you rent guests by the order, forever, and somebody else sets the rent. In the hybrid you buy a guest once, through Maps and reviews, then serve them again with no toll attached.

Where the model actually breaks?

Food cost is measured the same way in both formats —32% ceiling per dish, with labor and rent kept out of the plate— but one extra point does not weigh the same:

inside a pure ghost kitchen at 28% commission, each food cost point hits margin roughly 39% harder than it would in a physical restaurant running the same menu. Packaging is the invisible cost nobody budgets. At 0.45 to 1.20 dollars per order against an 11-dollar ticket, that is four to ten margin points which simply do not exist in a dining room, and in the pure model you pay them on every single order. Visibility works backwards from what most owners assume. A street unit puts its façade to work free, 24 hours a day; a hidden kitchen that misses the aggregator's first screen and the local Maps pack does not exist at all, and getting back on screen costs money every month.

Where the model actually breaks — in practice?

A virtual brand riding an existing kitchen changes the whole equation, because rent, hood and head chef are already covered by the core business;

delivery enters at marginal cost there, and a 28% commission hurts far less.