How to start a dark kitchen from scratch: the 40,000 USD mistake and the five alternatives that actually balance the books

Forty-one orders a day before you earn a cent

Building a dark kitchen from scratch with your own buildout only holds above 3,500 monthly orders, and almost nobody runs that math in orders: they run it in money, which is exactly where the error hides. A 45-square-meter hidden kitchen in Bogotá, Mexico City or Medellín carries between 4,200 and 7,800 USD a month in fixed structure —rent, three cooks, utilities, software, licenses— before a single plate leaves the pass, and with a 9 USD average ticket and a 38% contribution margin after platform commission, that demands 1,230 to 2,280 orders a month just to break even. Put in the cook's calendar: forty-one orders a day, every day, including the rainy Tuesday when the app sends nothing and the rent runs anyway. The variable that breaks every spreadsheet that reaches me is platform commission, because it comes off BEFORE you touch the money, not after the raw material cost.

Commission never touches food cost: it takes the selling price

Rappi, iFood, Didi Food and Uber Eats charge between 18% and 30% of order value depending on city, category and whether you use your own couriers or theirs, so a dish with a 30% menu food cost ends up at an effective 41% once the app keeps 27%. That eleven-point jump will not come back by buying chicken better. And the scale of the platform business explains why they will not bend: the platform-to-consumer segment moved USD 96,864 million in 2024 according to Statista, and Delivery Hero reported €12.8 billion in segment revenue for 2024, up 22% year over year. The number that reveals your own buildout was the wrong call shows up in month three, when the virtual brand has no traffic and the algorithm refuses to surface it. A virtual restaurant depends 100% on the app's ranking, and that ranking rewards volume, rating and prep time, three things nobody has fresh out of the gate; the first 60 days get bought with in-app advertising —an extra 8% to 15% of sales— or they stay empty.

When building from scratch falls short for you?

Add those points to a 27% commission and your 38% contribution margin drops under 25% in precisely the quarter when cash is scarcest. If your projection lacks that line from day one, you do not have a projection:

you have a wish with a spreadsheet attached. For the owner already operating a dining room, the right alternative is launching the virtual brand in your own kitchen during dead hours, and the switching cost runs 300 to 900 USD: packaging, menu photography and platform onboarding. Fixed structure is already paid by the dine-in business, so the break-even in orders falls from 1,230 to under 300 a month, ten a day, a figure an existing kitchen absorbs between three and six in the afternoon without hiring anyone. The downside is real and deserves saying out loud: if the virtual brand lands, it will collide with dining-room service at peak, and you will have to decide whom to fail.

Option 1: start inside the kitchen you already run

Profile: operator with his own venue and kitchen occupancy below 60%. Anyone without a venue who wants to test the market should enter through a shared kitchen, where a station costs between 600 and 1,800 USD a month depending on city and shift, against the 4,200 to 7,800 of your own structure. The real advantage is not the saving but the nature of the cost: delivery demand swings up to 40% between the first and third week of the month in Latin American markets, and an hourly kitchen lets you pay little during the slow week, while your own buildout charges the same on the 8th as on the 25th. Against it: you control neither the cleaning, nor the hood schedule, nor the neighbor frying fish beside your dessert. Profile: founder without assets, testing two or three concepts before committing.

Option 3: host kitchen under a co-packing contract

The third route —rarely used and the one I recommend most for brands with a closed recipe book— is contracting an established restaurant to produce your menu for a fee of 12% to 20% of sales, with no rent, no payroll and no equipment. You bring the brand, the technical sheets and quality control; the host brings a kitchen already fired up. Switching cost is near zero in money and high in management: portion weights need auditing every week or the margin evaporates in generous plating. Diego F. Parra has built this scheme inside the Masterestaurant methodology precisely because it turns fixed into pure variable, the only thing that protects a new brand when the algorithm punishes it two months running. Profile: brand with standardized recipes and no operating team. I got this wrong for years by telling owners to look at the initial investment, when the figure that decides is the exit.

The cost of leaving weighs more than the cost of entering

Closing a dark kitchen with a buildout means a rent penalty for the remaining term, hood dismantling and selling used equipment that recovers between 30% and 40% of the original invoice on the secondary market: a 35,000 USD equipment investment comes back as 12,000 if you are lucky and a buyer appears. An alternative without construction closes with an email and thirty days' notice. Run the full counterfactual: if at eight months your brand still sits under 900 monthly orders and you decide to close, your own buildout costs you the sunk capital plus the penalty; a shared kitchen costs you the current month and the recipe travels with you. There is one case where building from scratch is the right decision, and it would be dishonest to skip it: when you already sell more than 3,500 orders a month, negotiated commission below 22% on volume, and hold at least 35% of sales through your own channel —web, WhatsApp, phone— that pays commission to nobody.

When NOT to switch and keep the buildout?

At that volume a 7,800 USD fixed structure dilutes to 2.2 USD per order, while a shared kitchen would keep charging a variable that never drops again.

The mature market confirms it: China has more than 3,200 ghost kitchen facilities according to Coherent Market Insights, and they were built where the volume already existed, not the other way around. Before signing five years, calculate your break-even in daily orders and look at it on a rainy Tuesday. An own build-out converts a variable cost into a fixed one, and in delivery that conversion is dangerous: demand in a ghost kitchen swings roughly 40% between the first and third week of the month across Latin American markets, while rent swings not at all. A shared kitchen by the hour lets you pay little during the slow week; a build-out charges the same on the 8th as on the 25th.

Four differences that decide the cash

Exit cost weighs more than entry cost, and I got this wrong for years by telling operators to look only at the initial investment. Closing a dark kitchen with construction means lease penalty, hood removal, and selling used equipment that recovers 30% to 40% of the invoice on the secondary market. An alternative without construction closes with one email and thirty days of notice. Food cost behaves differently by format: buying for a single brand doing 400 orders a month means retail prices and real food cost drifting to 32%, which is the MAXIMUM allowed and never the target. Buying for three virtual brands on the same line, that identical input drops to 27% on volume. Scale is not vanity here, it is margin. Diego F. Parra insists on a point almost nobody models: the physical menu does NOT disappear because you sell through an app.

Four differences that decide the cash — in practice

If your dark kitchen grows out of a restaurant with a dining room, keep the printed menu for table service —it controls pace, narrative and suggestive selling— and use the QR menu as a complement for delivery, price changes and analytics. The Masterestaurant verdict is BOTH, each with its role, never QR alone.