How to start a dark kitchen from scratch: the 2026 numbers that decide whether it survives or not
How to start a dark kitchen from scratch in 2026 comes down to four numbers, not to a brand idea: platform commission between 18 % and 30 % of the ticket, food cost kept under 32 %, packaging running 5 % to 8 % of sales, and a break-even that rarely falls below 900 to 1,100 monthly orders per brand. The mistake that ruins most openings is costing the dish against a dine-in menu price and then selling it inside Rappi or iFood, where commission eats the margin before the first payroll run. The right method flips the sequence: set the in-app price first, design the recipe that fits that price second, and only then sign the kitchen lease.
A 38-square-meter kitchen in Bogotá, two virtual brands sharing one line, 640 orders in month one and a 9,400-dollar loss. The owner had done nothing foolish: decent equipment, a cook who knew the craft, paid placement inside the app. His single mistake was costing his dishes against street-restaurant pricing and then selling them through a channel that keeps close to a third of every invoice.
That is the pattern. Dark kitchens rarely fail on the food; they fail on the arithmetic of a channel nobody read before signing. And in 2026, with prepared-food delivery still growing at double digits across Latin America while commissions hold firm, that arithmetic is tighter than it was in 2021, when anyone could launch a virtual brand in a week and raw volume covered the costing errors.
What follows are the numbers I reach for whenever someone asks how to start a dark kitchen from scratch, grouped by decision: what it costs to open, what the channel takes, what margin survives and how many orders are needed. Every figure carries the decision it triggers. If a number does not change what you do on Monday, it did not belong here.
Side-by-side comparison
| Common opening mistake | Masterestaurant method | |
|---|---|---|
| Costing sequence | ✕Recipe first, price later: 34 % food cost on street pricing becomes 46 % inside the app | ✓In-app price first, recipe built to fit: 32 % food cost ceiling calculated net of commission |
| Delivery commission | ✕15 % assumed from memory; the actual contract runs 18 %-30 % plus 3 %-5 % payment processing | ✓Worst contractual case modeled at 30 % and validated against the first four weekly settlements |
| Virtual brands at launch | ✕3 to 5 brands on day one, 60 SKUs, saturated line and 22-minute prep times | ✓1 brand, 8 to 12 SKUs, prep time under 12 minutes before the second brand opens |
| Opening investment | ✕45,000 to 80,000 USD replicating the kitchen of a dining room that does not exist | ✓18,000 to 35,000 USD in production equipment, packaging and four months of working capital |
| Packaging | ✕Treated as a minor consumable, estimated at 2 %, never measured; lands at 9 % of sales | ✓Its own costing line, 5 % to 8 % audited per order, tested through a 25-minute transport run |
| Break-even | ✕Never calculated; gross sales on the Rappi dashboard get celebrated instead | ✓900 to 1,100 monthly orders per brand, calculated before the lease and reviewed every week |
| Menu and printed piece | ✕Digital menu only because there is no dining room, nothing physical inside the bag | ✓Digital menu tuned for the app PLUS a printed card in the bag that tells the brand story and drives direct repeat orders |
What does it really cost to open a dark kitchen in 2026?
Opening a dark kitchen from scratch costs between $25,000 and $60,000 in Latin America, and the independent segment generates 61.7% of cloud kitchen market revenue, not the large chains (Grand View Research, Cloud Kitchen Market 2025).
That figure matters because it dismantles the assumption that you need an investment fund behind you: the solo operator is still the majority. What moves the opening number is not square footage but the extraction hood, which in a 35 to 40 square meter kitchen takes $6,000 to $9,000 in one shot, and the lease deposit, which in high order-density zones demands three months up front. Mexico City now has more than 1,200 active dark kitchens, 40% more than in 2023 (CANIRAC, 2025). The decision this figure triggers is simple: if you do not have $30,000 available without touching working capital, you are not ready to open.
Platform commission takes 18% to 30% of every invoice
A third of your ticket is not yours, and that number decides whether the business exists at all. Platforms charge between 18% and 30% per transaction depending on the plan, and channel concentration makes negotiation worse: iFood controls 87% of e-food bookings in Brazil (Statista, 2024), Uber Eats closed 2024 with 26.1% of the U.S. market (Earnest Analytics, 2024), and in Spain Glovo sits near 31% against Just Eat's 26% (Ken Research, 2025). When one player handles almost nine of every ten orders in a country, commission stops being negotiable and becomes a tax. Which is why you do not set the selling price: it is set by what the customer pays inside the app while comparing eight other restaurants on the same screen. Treat commission as input data, never as an expense you will review later. In a dark kitchen, target food cost sits below 32% of the selling price, and that 32% is the MAXIMUM tolerable figure, not the goal.
Food cost under 32%: the ceiling nobody should touch
The reasoning is pure arithmetic: if the channel keeps 25 points and the product eats another 32, you have 43 points left to cover packaging, kitchen payroll, rent, ordering software and profit. Diego F. Parra, restaurant consultant and founder of Masterestaurant, keeps hammering a point that almost everyone inverts: in a virtual brand the recipe is the variable and the price is the fixed input, because the customer already decided what a rotisserie chicken is worth before opening the app. Costing backwards contaminates the portion size, the supplier and the margin. If your spec sheet returns 38% food cost against the channel price, do not raise the price: redesign the dish or pull it from the catalog. Packaging eats between 5% and 8% of sales in a dark kitchen, and it is the line that shows up as zero most often in the spec sheets I review. A tamper-sealed container, the thermal bag, the brand label and the cutlery add $0.45 to $1.20 per order; on an average $12 ticket that lands between 3.7% and 10%.
Packaging: the 5% to 8% of sales almost nobody costs
When someone tells me their margin evaporated with no explanation, packaging is usually half the hole. Consider the reverse scenario: drop packaging from 8% to 5.5% with a single-piece container and remove default cutlery, which 70% of customers never use at home, and you recover 2.5 margin points without touching the recipe or the price. Those 2.5 points, across 1,000 monthly orders at $12, come out to $300 clean every month. Almost no dark kitchen crosses break-even below 900 to 1,100 monthly orders, and that range is the figure that hurts most once you run the numbers before signing a lease. The math holds no mystery: with a $12 ticket and a 35% contribution margin after commission, each order leaves $4.20; a fixed structure of $4,200 a month covering rent, two cooks, utilities and software demands 1,000 orders just to reach zero.
Break-even: between 900 and 1,100 orders a month
The Chapinero case that opened this piece sold 640 orders in its first month and lost $9,400: demand was not missing, 360 orders were, and eight points of food cost were in excess. Before you open, divide your monthly fixed cost by the margin per order and ask whether your zone generates that volume. If the answer feels uncertain, the zone is the problem. Running two or three virtual brands from one kitchen works only when they share at least 60% of their inputs, and outside that condition it multiplies waste instead of margin. The logic is inventory, not marketing: two brands with separate pantries double SKUs, double spoilage and force you to rotate product that does not rotate. With Latin American meal delivery projected above $39 billion by 2027 (Statista, 2024), the temptation to launch brands and capture more screen space is strong, and there sits the paradox: more brands raise visibility inside the app, yet every new brand splits the same kitchen team across more simultaneous tickets and stretches dispatch time, the very factor platforms punish in their ranking.
Two brands on one stove: when it adds and when it sinks you
The rule I use is blunt and it works: launch a second brand only once the first holds 1,200 monthly orders. Kitchen automation still does not solve the problem of a small dark kitchen, whatever the headlines claim. North America holds 40.8% of the global food robotics market (Grand View Research, Food Robotics Market), Starship robots completed 5.8 million autonomous deliveries in 2024 (Forbes, 2025), and White Castle passed 100 drive-thrus running voice AI, with a 90% order completion rate and roughly 60 seconds per order (Restaurant Dive, 2024). All of that is true, and none of it lowers food cost in a 38 square meter kitchen in Bogotá. What does pay off today is the software layer: a standardized recipe book, inventory control by spec sheet, and a daily read of margin per dish. The technology worth buying in 2026 has no arms; it has a cost sheet that updates itself when chicken prices move.
The 3 figures you should tattoo on yourself
Three numbers decide whether your dark kitchen survives, and none of them can be fixed after opening. First, commission: 18% to 30% of the ticket, in markets where a single player handles 87% of orders (Statista, 2024); concrete action, cost every dish against the NET price the platform settles with you, never against the price the customer sees. Second, food cost below 32%; concrete action, pull from the catalog this week any dish above that ceiling and do not compensate by raising the in-app price. Third, break-even at 900 to 1,100 monthly orders; concrete action, before signing the lease divide your fixed cost by contribution margin per order and confirm your zone, in a city with 1,200 active dark kitchens like Mexico City (CANIRAC, 2025), holds demand for that volume. Run that division today, on paper, before you look at a single location. The first difference is SEQUENCE.
Three differences between a profitable dark kitchen and one that closes
In a physical restaurant you design the dish, cost it and price it; in a virtual brand the price is imposed by what the customer will pay inside the app while comparing eight other restaurants on the same screen, so price is the input and the recipe is the variable. Flipping that order is the mistake almost every other one descends from, because it contaminates costing, portion size and even supplier choice. The second is the NATURE of the channel. Rappi, iFood and their peers are not one more sales channel: they are an intermediary that charges 18 % to 30 % per transaction and keeps the customer data. A restaurant with a dining room that gives up 25 margin points in one channel offsets it across its tables; a dark kitchen has no tables. All of its volume runs through there, which makes commission a cost structure, not a marketing expense.
Three differences between a profitable dark kitchen and one that closes — in practice
The third is WORKING CAPITAL. Platforms settle every one or two weeks while protein and fresh-produce suppliers bill at 8 or 15 days and payroll does not wait; the gap between what you sold and what you collected is what kills kitchens that were technically profitable. Diego F. Parra hammers this point in every Masterestaurant diagnosis: a dark kitchen needs four months of cash in the bank before opening, and whoever lacks it is funding the platform's growth out of their own pocket.
Mistake versus method, criterion by criterion
Seven mistakes that drain the cash in 90 daysWhat repeats
- Costing the dish against a physical restaurant price and then publishing it on Rappi with the same tag, which turns a 31 % food cost into a real 44 % the moment commission lands.
- Launching four virtual brands at once because the platform account manager promised more visibility, when what multiplies is dead inventory and prep time.
- Signing a 24-month lease before collecting a single week of demand data for the delivery polygon.
- Forgetting that payment processing bills separately from commission, an extra 3 % to 5 % that never makes it into the opening spreadsheet.
- Buying dine-in equipment: display cases, hot tables, an industrial dishwasher that will never run at the pace it was built for.
- Judging success by gross sales on the app dashboard instead of by the biweekly settlement, which is the only number that reaches the bank.
- Handing over the bag with no printed brand piece inside, donating to the platform a customer relationship that your food paid for.
The right method, in the order it gets executedMasterestaurant
- Set the selling price INSIDE the platform before writing the first recipe, then build the dish downward until food cost sits at 32 % or below.
- Model commission at its worst contractual case, 30 %, plus 4 % processing: if the business works there, it works under any better deal.
- Launch with one brand and eight to twelve items, all executable in under 12 minutes by a single cook during peak hours.
- Rent monthly or share an existing kitchen through the first quarter; the long lease gets signed once polygon demand has been measured.
- Audit packaging as a cost line with its own 5 % to 8 % target, and test every container through 25 minutes of real transport before buying volume.
- Calculate break-even in monthly orders rather than currency, and post it on the kitchen wall where the team sees it daily.
- Put a printed card in every bag with the brand story and a direct reorder channel: it is the only way to stop renting your customer from the app.
Side-by-side comparison
| Common opening mistake | Masterestaurant method | |
|---|---|---|
| Costing sequence | ✕Recipe first, price later: 34 % food cost on street pricing becomes 46 % inside the app | ✓In-app price first, recipe built to fit: 32 % food cost ceiling calculated net of commission |
| Delivery commission | ✕15 % assumed from memory; the actual contract runs 18 %-30 % plus 3 %-5 % payment processing | ✓Worst contractual case modeled at 30 % and validated against the first four weekly settlements |
| Virtual brands at launch | ✕3 to 5 brands on day one, 60 SKUs, saturated line and 22-minute prep times | ✓1 brand, 8 to 12 SKUs, prep time under 12 minutes before the second brand opens |
| Opening investment | ✕45,000 to 80,000 USD replicating the kitchen of a dining room that does not exist | ✓18,000 to 35,000 USD in production equipment, packaging and four months of working capital |
| Packaging | ✕Treated as a minor consumable, estimated at 2 %, never measured; lands at 9 % of sales | ✓Its own costing line, 5 % to 8 % audited per order, tested through a 25-minute transport run |
| Break-even | ✕Never calculated; gross sales on the Rappi dashboard get celebrated instead | ✓900 to 1,100 monthly orders per brand, calculated before the lease and reviewed every week |
| Menu and printed piece | ✕Digital menu only because there is no dining room, nothing physical inside the bag | ✓Digital menu tuned for the app PLUS a printed card in the bag that tells the brand story and drives direct repeat orders |
The 2026 numbers that rule a dark kitchen, and the decision each one triggers
“We shut down two of the three virtual brands and raised the average in-app ticket from 21,000 to 26,500 pesos. We lost 140 orders a month, but margin per order went from 8 % to 24 % and the biweekly settlement stopped scaring us: from a 9,400-dollar loss in the first quarter we moved to 6,200 in profit in the second, same 38-square-meter kitchen, one cook fewer on the night shift.”
How to start a dark kitchen from scratch in four steps, in this order
Open Rappi and iFood in the polygon you intend to serve, look at what customers actually pay in your category and choose your in-app selling price. That number is an input, not a result. Subtract the worst contractual commission, 30 %, plus 4 % processing, and what remains is your real net price. Calculate 32 % food cost against that net. If the recipe in your head does not fit there, the recipe changes; the price does not.
Launch with a single virtual brand and eight to twelve items that share mise en place. Time every dish with one cook during peak hours: anything over twelve minutes leaves the menu or gets redesigned. The temptation to open three brands on the same day always comes from the platform's account manager, and I understand the argument —more search surface, more orders— but it multiplies inventory, waste and dispatch errors precisely when you still have no process.
Add rent, utilities, fixed payroll and equipment depreciation; divide that total by contribution margin per order, which is net price minus food cost and packaging. The result is monthly orders, almost always 900 to 1,100 per brand. Post that figure on the kitchen wall and check it every Monday against the week's real orders. A team that knows how many orders remain before costs are covered works differently from one that only watches the ticket screen.
Put a printed card in every bag with the brand story, the cook's name and a direct reorder channel. I got this wrong for years: I believed that without a dining room paper was pointless, and what was pointless was my faith that the platform would hand the customer back. It does not. Keep both pieces always, the digital menu inside the app —which updates prices and feeds analytics— and the printed piece inside the order, the one asset that belongs to you.
And with AI?
Optimize channels, pricing and unit economics of your dark kitchen. Diego F. Parra is an expert in AI applied to restaurants.
Free tools to apply this now
Ecosystem tools to build the dark kitchen on numbers
None of the four steps above survives on memory or a napkin; they survive on a sheet where net price, food cost and break-even recalculate themselves every time the platform moves its commission. These three Masterestaurant tools cover that work and are built for delivery logic rather than dining-room logic.
Diego F. Parra — International consultant, expert in creating and scaling restaurants and in AI applied to restaurants, foodtech and HORECA. Methodology applied in 8.400+ restaurants across 43 countries · Expert in Artificial Intelligence applied to restaurants, hospitality and food businesses · 20+ years in restaurants, catering, large events and business growth · Author of 3 ISBN-registered books: «Triunfar o morir en el intento» (2013) and «De esclavo a dueño» (2023) · International keynote speaker for the HORECA sector.
Frequently asked questions about starting a dark kitchen from scratch
How much does it cost to start a dark kitchen from scratch in 2026?
How much does it cost to start a dark kitchen from scratch in 2026?
Between 18,000 and 35,000 dollars for a single-brand operation: production equipment, minimal build-out, initial packaging and four months of working capital. Openings in the 45,000 to 80,000 dollar range are almost always replicating the kitchen of a dining room that does not exist here and never will.
How much do Rappi and iFood keep per order?
How much do Rappi and iFood keep per order?
Negotiated commission runs 18 % to 30 % of order value depending on category, exclusivity and tenure, and payment processing sits on top at another 3 % to 5 %. Always model 30 % plus 4 %: if the business holds under that scenario, any better deal is extra profit.
Is a dark kitchen better than a physical restaurant?
Is a dark kitchen better than a physical restaurant?
It depends on where you want to fight for margin. A physical restaurant pays expensive rent but controls the experience and keeps 100 % of the ticket; a dark kitchen halves the opening investment but hands 22 to 34 margin points to the channel. If your concept cannot absorb losing 30 points, do not take it to pure delivery.
How many virtual brands should a new dark kitchen run?
How many virtual brands should a new dark kitchen run?
One, through the first quarter. The second brand opens once the first sustains its 1,100 monthly orders with prep times under twelve minutes and waste under control. Launching three on day one to increase sales on Rappi is the fastest route to dead inventory and orders canceled for delay.
Sector data 2026 (official sources)
Verifiable industry benchmarks from official, non-commercial sources (government, industry associations, market research) - not competitors.
| Metric | Benchmark 2026 | Source |
|---|---|---|
| Mayor mercado de delivery (China) 2026 | USD 539.87 mil millones de ingresos en China en 2026 | Statista 2026 |
| Delivery en línea América Latina 2027 | Segmento meal delivery superará USD 39 mil millones en 2027 | Statista 2024 |
| Mercado delivery en línea América Latina 2024 | USD 12,917.3 millones en 2024; CAGR 8.6% (2025-2030) | Grand View Research 2025 |
| Modelo plataforma-a-consumidor en LatAm | 80.07% de participación de ingresos en 2024 | Grand View Research 2025 |
| Usuarios de delivery en línea LatAm 2026 | 147.0 millones de usuarios en 2026 | Statista 2024 |
| Mercado delivery y dark kitchens España | Aprox. USD 5 mil millones | Ken Research 2025 |
Related content
Put numbers on your dark kitchen before you sign the lease
The Masterestaurant method starts with real costing net of commission and ends with a break-even in orders your team understands. Start with the costing tool and bring your in-app prices.