How to open a ghost kitchen: the before and after of owning your demand engine

Which wins: signing the lease first or validating digital demand first?

Validating digital demand first wins, and the margin is not up for debate: a 3,000 USD experiment against 40,000 USD locked into construction.

The Medellín owner who started with 47,000 USD across hood, walk-in cooler and three months of prepaid rent closed at month eleven with a ceiling of 19 daily orders, while that same menu tested for six weeks out of a kitchen rented by the hour would have cost under 7% of that figure and delivered real ticket size, repeat rate and peak hour before a single brick moved. The improvised launch signs the lease and then prays; the disciplined launch measures and then signs. One figure frames the decision: optimal food cost sits between 28 and 35% according to the National Restaurant Association, a range you can only confirm with real orders, never with a spreadsheet built at your kitchen table.

One strong virtual brand versus two separate brands on the aggregator

Two separate virtual brands beat the single brand, and I got this wrong for years by recommending the opposite so the kitchen would not lose focus. A single brand competes in one listing category; two brands sharing a production line —chicken on one side, bowls on the other— double impressions without moving the 28-35% food cost the National Restaurant Association documents, because the input is identical and what changes is packaging and the listing itself. Cost per order drops because you show up twice in the same hungry user's eight o'clock search. The real risk sits in quality, not in the kitchen: two poorly separated brands cannibalize each other and confuse the algorithm. Separate them by category and by price band, not by a pretty name.

The aggregator is not a sales channel, it is a marketplace with an algorithm

Treating the aggregator as a marketplace with its own algorithm beats treating it as a passive channel, and scale explains why: iFood moved 100 million orders in the single month of August 2024 and runs with more than 380,000 partner establishments across over 1,500 Brazilian cities, according to its own institutional data. At that density nobody finds you by accident. Rappi, Uber Eats, DiDi Food and iFood rank by impression-to-order conversion, prep time you accept and actually meet, merchant-side cancellations and recent rating. The operator who assumes publishing is enough ends up on page three, exactly where the Medellín dark kitchen died. The operator who reads those four signals as daily cash indicators climbs the ranking without spending one extra peso on advertising.

A verified Google Business Profile against relying on the aggregator alone

The verified Google listing wins, and its absence costs more than any paid campaign. At the Medellín dark kitchen the listing read "permanently closed" from week two, and that single status switched off local search for eleven consecutive months while the owner paid full rent. A ghost kitchen has no storefront, no foot traffic, no window display; its storefront is an algorithm-ordered list and a pin on Maps. Relying only on the aggregator hands 100% of demand to a third party that charges commission and can reshuffle rankings tomorrow. With a verified listing, real hours, product photos and fresh reviews, you recover a slice of direct commission-free orders, and that clean margin is what pays for the hood. Diego F. Parra insists at Masterestaurant: the listing first, the bricks afterward.

The Medellín case: 47,000 USD of infrastructure against 19 daily orders

Eleven months and 47,000 USD invested to close at a ceiling of 19 daily orders: that number frames the whole comparison. The build was flawless —certified extraction, walk-in cooler, three months of rent paid up front— and the team delivered. The demand engine failed. At an 8 USD average ticket those 19 orders produced roughly 152 USD of daily gross sales, and with food cost near the top of the 28-35% band the National Restaurant Association reports, daily contribution margin did not even cover rent plus the courier. Break-even for that kitchen demanded close to triple the order count. Nobody ran that arithmetic before signing. When cash flow is the leading cause of financial stress and closure among small businesses, as Inc. reminds us, the mistake was not overspending but failing to measure first.

Contribution margin per order versus margin on the printed menu

Calculating margin PER ORDER beats calculating margin per dish, and in delivery that difference swallows the whole business. A dish at 30% food cost, comfortably inside the 28-35% range the National Restaurant Association documents, looks healthy on the menu and stops being healthy once you add aggregator commission, packaging, complimentary sauces and the promotional discount the algorithm rewards with more impressions. The order, not the dish, is the decision unit. Build a sheet where every real combination —dish, packaging, commission, active promotion— returns contribution in currency, and you will see two or three menu references destroying cash while the rest carry it. Pull those references before you open, not six months later. And remember: payroll, rent and utilities never load onto the dish, they belong in break-even.

What would happen if the aggregator changed its ranking algorithm tomorrow?

Should the aggregator reshuffle its ranking tomorrow, a single-channel operator loses between 40 and 100% of orders within a week, while the one who built owned traffic loses a slice and stays open.

That scenario is anything but hypothetical: the industry moves fast, with Glovo billing more than €1 billion annually in q-commerce according to EU-Startups, Starship robots completing 5.8 million deliveries in 2024 according to Forbes, and Wendy's rolling out ordering AI across 500 to 600 locations by the end of 2025 according to CNBC. Every one of those moves redefines who sits at the top. Here is the trade of the craft: you need the aggregator to launch and you need independence from it to survive. Resolve it by using the aggregator for discovery while pushing repeat customers toward your direct channel.

What to choose based on your owner profile?

If you are an owner with less than 20,000 USD available, do not build your own kitchen yet: rent kitchen hours, publish two virtual brands, measure for six weeks and decide with data.

With 20,000 to 50,000 USD and validated ticket and repeat rate, go ahead with the build, but open with the Google listing verified and contribution margin per order calculated before the first dispatch. If you already run a restaurant with idle kitchen capacity during off-peak hours, your call is different and far more profitable: launch the virtual brand on infrastructure you already pay for, with near-zero investment and food cost inside the 28-35% band the National Restaurant Association sets. In all three cases the order is identical: demand first, bricks later. Open your contribution-per-order sheet today with your three best sellers and check which one is costing you money.

The four differences that decide whether your ghost kitchen survives

The first difference is sequence, and it moves the most money: an improvised launch signs the lease and then thinks about demand, while a method-driven launch validates demand on the aggregator with a virtual brand before a single brick moves. You can test a nine-item menu from an hourly-rented kitchen for six weeks, measure real ticket and repeat rate, and only then decide whether the build-out is worth it. That reversal turns 40,000 USD of risk into a 3,000 USD experiment. Second: the aggregator is not a channel, it is a marketplace with its own algorithm. Rappi, Uber Eats, DiDi Food and iFood rank on impression-to-order conversion, accepted prep time, merchant-side cancellation rate and recent rating. A kitchen that accepts within 60 seconds and dispatches in 18 minutes climbs; one that cancels 7% of its orders for missing ingredients sinks and stays sunk, price cuts notwithstanding. Kitchen operations ARE ranking. Almost nobody works the third difference: you have no façade, but you do have an address, and Google Maps indexes it exactly the same.

The four differences that decide whether your ghost kitchen survives — in practice

A verified Google Business Profile with a precise primary category and fresh reviews captures both the «ghost kitchen near me» search and the branded search —people who tried your burger on Rappi and now look it up by name— and branded search arrives with zero commission attached. In a well-run dark kitchen that traffic is 18% to 24% of mature order volume. The fourth is arithmetic, which forgives nothing. On a 12 USD ticket at 34% food cost, 30% commission and 0.80 USD packaging, contribution margin lands at 1.84 USD per order: you need 78 daily orders just to cover 4,300 USD of monthly fixed costs. Raise the ticket to 17 USD with combos, pull food cost to 29% and blend in 35% direct orders at 6% commission, and margin jumps to 5.40 USD — the same kitchen closes the month at 27 daily orders. What separates 78 from 27 is not the equipment, it is the spreadsheet almost nobody builds before opening.