Virtual restaurant business model: real alternatives when commission eats the margin

The arithmetic that defines the model before the stove is lit

A virtual restaurant's business model works when the average ticket clears 28 USD and real food cost stays under 28%; outside that band, aggregator commission —22% to 30% on Rappi and iFood— plus packaging, which eats 4% to 7% of the sale, leaves a 3% to 5% operating margin that any protein price hike wipes out. An operator in Bogotá showed me his March P&L: 42 million pesos sold, zero tables, zero servers, profit of 1.9 million. Sales were fine and the kitchen delivered, yet 26% went to commission, 6% to packaging he picked because it looked good, and 31% to raw material, because he never costed the recipe against the price the customer sees in the app. This does not break in the kitchen. It breaks in the last mile. Removing the dining room transfers cost to a third party rather than eliminating it, and that confusion sinks more virtual kitchens than any product problem.

Lower fixed costs is not more margin: it is cost transferred

The format's promise holds in the half that matters: without a dining room, initial investment drops 60% to 70% compared with a sit-down location, and break-even arrives sooner because you carry no server payroll and no rent on square meters whose only job is seating people. Fine so far. Trouble starts when the owner reads that saving as available margin instead of as space the platform will occupy, charging on every single order. The global delivery app market moved 110 billion dollars in 2024, growing 15.5% according to Business of Apps, and that money comes from somewhere: from the commission you pay and from the customer the app considers its own, not yours. Living off the aggregator alone stops working once the platform's internal advertising passes 5% of gross sales, and that figure gives you away well before the income statement does. Three things shifted the board in 2026.

When the original option falls short?

Aggregators raised their commission range and pushed paid visibility plans, so the real cost of selling on Rappi is no longer commission alone but commission plus ad spend.

Serious operators started building their own delivery channel through WhatsApp and web ordering, where acquisition cost collapses once the customer has already tried the product. And menu and pricing AI stopped being an experiment: a recommendation engine now decides what the user sees, and whoever misreads that ranking competes blind. Pull your last fortnight's settlement. If commission plus ads plus packaging clears 35%, the channel is no longer a channel: it is a majority partner. Building your own channel suits the operator already running 400 recurring monthly orders with a minimum base of captured phone numbers; below that volume, you pay for structure to sell very little.

Option 1 · Your own delivery channel via WhatsApp and web ordering

Switching costs little money and a great deal of discipline: 60 to 150 USD a month for a web ordering platform, a half-time person answering WhatsApp, and either an in-house rider or a fleet contract, which in Colombia runs 2 to 4 USD per delivery against the 26% commission that on a 30 USD ticket equals 7.80 USD. The arithmetic is brutally favorable to the owned channel whenever the customer repeats. The real brake is not technological: nobody in the kitchen wants to answer messages at eight at night. Unless you assign that job to someone with a first and last name, the owned channel dies in week three. Multi-brand serves the operator whose kitchen sits underused in specific windows and whose dispatch times stay under 25 minutes at peak; for anyone else it simply multiplies the chaos.

Option 2 · Multi-brand on a single kitchen, with measured capacity

Most of the dark kitchen closures the trade talks about were never demand problems but structural ones: people who stacked four virtual brands on one production line without installed capacity for the eight o'clock peak, ended up dispatching in 45 minutes, saw ratings fall below 4.2 and vanished from the ranking. Platforms do not punish selling little. They punish failing to deliver. Entry cost is low —photography, listing and menu per brand, maybe 300 to 800 USD— and that is exactly the trap: the expensive part arrives later, in cross waste and in one cook reading three tickets from three different brands at once. Renting a station in a third-party cloud kitchen fits whoever wants to test a market without tying up capital, and not whoever already bills steadily and needs absolute process control.

Option 3 · Shared kitchen inside a specialized operator

The market sits near 83.5 billion dollars in 2026 with a projected 9.7% CAGR through 2034 according to Fortune Business Insights, and in Mexico specifically it moved 1.1 billion dollars in 2024 growing 10.74% a year per IMARC Group, so station supply exists and competes. You pay fixed rent plus a variable slice, typically 8% to 15% of sales, layered on top of aggregator commission. Add it up: 26% app, 12% shared kitchen, 6% packaging. That leaves 56 cents of every dollar for product, labor and profit. At a 30% food cost, your operating profit lives between 4% and 8%, and there is no room for error there. In the digital channel, price is built from the number the customer sees, working backwards, not from plate cost forward as every brick-and-mortar manual teaches. Take the final price in the app, subtract commission, subtract packaging, and only then look at what remains for raw material.

Option 4 · Pricing built backwards from the app

If that remainder does not allow a food cost of 32% or less —and 32% is the ceiling, not the target— the dish does not go on the digital menu. Full stop. That inverted calculation separates the kitchens that survive from those that discover after eight months that their three best sellers were losing money on every dispatch. It is the cheapest lever of all: no investment, just an afternoon with a calculator and the nerve to pull your top-selling dish. At Masterestaurant, Diego F. Parra insists on running that exercise before negotiating commission, because a badly costed menu is not fixed by shaving two points off the aggregator. Stay exactly where you are if your operating margin clears 12%, your rating holds above 4.6 and your kitchen dispatches in under 22 minutes at peak: moving that system destroys value for the sake of fashion.

When NOT to change anything?

Colombia moved 1.18 billion dollars in online delivery during 2024 and grows 7.32% annually through 2029 according to Statista Market Insights, which means the channel will keep feeding volume to whoever operates clean, expensive as it is.

Aggregator dependence is a risk, agreed, though a risk managed with healthy margins beats an improvised owned channel that will steal attention from the operation. Before changing anything, do this tomorrow: open the last four weeks of settlements, calculate commission plus ads plus packaging over gross sales, and compare it against your operating margin. That single number tells you whether you own a business or work for an app. Price gets built backwards. In a physical restaurant you start from plate cost and add margin; in the digital channel you must start from the final price the customer sees in the app, subtract commission, subtract packaging, and only then look at what remains for raw material.

What separates a profitable virtual restaurant from one that only bills?

If that remainder does not allow a food cost of 32% or less, the dish does not go on the digital menu. Full stop.

That inverted calculation is the difference between kitchens that survive and kitchens that discover at month eight that their three best sellers were losing money. The digital menu and the printed menu serve different jobs and neither replaces the other. When an operator has a dining room, Masterestaurant ALWAYS recommends keeping the physical menu alongside the QR menu: print controls the experience —service pace, dish narrative, suggestive selling, hospitality— while the QR and digital menu handle delivery, accessibility, live price changes and analytics on what the customer actually looks at. In the virtual arm of the business the digital menu rules; in the dining room the printed card rules. Both, each in its role. Dependence gets measured, not guessed. A healthy operator in 2026 takes 45% to 60% of deliveries through direct channels —WhatsApp with a catalogue, web ordering, phone— with the rest split across two aggregators.

What separates a profitable virtual restaurant from one that only bills — in practice?

If Rappi represents 90% of your revenue, you do not own a business: you hold a revocable concession. And concessions get revoked without notice when the platform reshuffles its ranking algorithm or a competitor outbids you for visibility in your own zone.

Installed capacity caps how many brands you can carry. Every extra virtual brand on the same kitchen adds SKUs, adds mise en place, adds picking error and stretches dispatch time. The rule I use: a second brand only when the first runs below 65% peak occupancy and shares at least 70% of its inputs. Outside that condition, the second brand does not multiply sales; it cannibalises the quality of the first and drags both ratings down.