Setting up a ghost kitchen in 2026: the numbers that decide whether it opens

How big is the market you're actually fighting for

The global food delivery app market moved USD 110 billion in 2024, growing 15.5% (Business of Apps, Food Delivery App Report 2025), and that number explains why so many three-station industrial units get leased on 24-month terms by owners who never counted a single order before signing. Colombia contributed USD 1.18 billion that same year with a projected 7.32% CAGR through 2029 (Statista Market Insights, 2024), Mexico USD 9.22 billion growing 14.66% annually (Statista, 2024) and Brazil close to USD 18.8 billion, the largest in Latin America (Statista, 2024). A market growing at double digits attracts capital and competition in equal measure, so the size of the pie tells you nothing about your slice. The decision these figures trigger together is simple: enter delivery on measured volume inside your own zone, never on a market headline.

Ghost kitchen growth rates, country by country

Ghost kitchens grow faster than the delivery that feeds them, and that gap is why competition inside your 3 km radius doubles in under four years. India goes from USD 552 million in 2023 to USD 1.523 billion by 2030, a 15.6% CAGR (Coherent Market Insights via GlobeNewswire, 2024). The United Arab Emirates moves from USD 430 million in 2025 to USD 1.0826 billion in 2032, 14.1% a year (Coherent Market Insights, 2025). The Middle East and Africa run at 21.9%, from USD 427 million in 2024 to USD 1.074 billion in 2030 (MarkNtel Advisors, 2024). Mexico adds USD 1.1 billion in 2024 at 10.74% through 2033 (IMARC Group, 2024). Operationally that means one thing: if your financial plan assumes year-one competition in year three, it is wrong, because virtual brands per zone multiply at those very CAGRs.

The 45-orders-a-day threshold is arithmetic, not opinion

Opening a ghost kitchen only makes sense once your 3 km delivery radius already generates at least 45 orders a day and your blended aggregator commission sits below 24%. At an average ticket of USD 14, those 45 orders are USD 630 in daily sales, roughly 18,900 a month. Subtract 31% food cost, 19% payroll, 24% commission and USD 2,300 of rent and utilities: about USD 2,700 remain before tax, a 14% operating line that survives a bad month. Drop to 30 orders a day and the same structure hands you a NEGATIVE result, because rent and payroll do not shrink proportionally. The Medellín client who billed USD 11,400 in month one at 27% commission is exactly that second column: good food, bad location arithmetic.

Aggregator commission is equity dressed up as an expense

When you accept 29% of the ticket, the aggregator takes more gross margin than you keep after food cost and packaging, and that is not an operating expense: it is a partner who put in no capital. In a ghost kitchen with a USD 14 ticket, one percentage point of commission equals USD 118 a month per 1,000 orders, so moving from 29% to 23% hands back USD 708 monthly that was already sitting in your P&L. At 1,350 monthly orders — the 45-a-day threshold — those six points are USD 956. Diego F. Parra works it this way with the owners he advises at Masterestaurant: negotiate the percentage before touching menu prices, because the commission point is the only cost you can cut without touching plate quality or anyone's paycheck.

Household density and competitors: the data that picks the site

You can serve the best fried chicken in town and still fail if your zone holds 900 households with real spending power and seven competitors in your category, because the delivery radius outranks product quality. With a realistic 4% monthly penetration on those 900 households and 1.8 orders per household per month, the zone ceiling is 65 monthly orders per competitor when demand splits evenly: not even a quarter of the 1,350 you need. Cheap industrial rent is precisely the signal that nobody with a credit card lives there. Before touring units, count households inside the radius, count active brands in your category within the app and pull the aggregator's peak-hour history for that area; if the zone cannot carry the number, no recipe fixes it.

The rent saving is real, and on its own it saves nothing

A ghost kitchen saves between 60% and 75% on rent versus a dining-room location, and that saving evaporates entirely once blended commission passes 26%. Run the counterfactual: a restaurant with a dining room pays USD 8,000 rent and keeps 100% of the ticket; its hidden version pays USD 2,300 — 71% less — but hands 27% of every sale to the aggregator. On USD 18,900 monthly, that 27% is USD 5,103, meaning the USD 5,700 rent saving is almost entirely consumed by commission and leaves barely USD 597 of structural difference, with no dining room, no tips, no beverage sales at 80% margin. The conclusion this block triggers is uncomfortable and firm: the hidden model is not cheaper, it is more DEPENDENT, and that dependency gets paid on every single order.

Spain and Europe: a mature market shows where the curve ends

Spain closed 2023 at USD 928.22 million in ghost kitchens with a projected CAGR of just 4.5% through 2032 (Expert Market Research, 2024), far below India's 15.6% or the 21.9% of the Middle East and Africa, while Central and Western European delivery moved USD 98.48 billion in 2024 (Statista, 2024). A large market growing at 4.5% has already handed out its positions, and new growth there comes from taking orders away from someone, not from customers discovering the category. China, at USD 40 billion in delivery in 2024 (Coherent Market Insights, 2024), and the alternate cloud kitchen estimate of USD 45.65 billion (MarkNtel Advisors, 2024) confirm the pattern: where the market matures, a virtual brand without a real difference drops off the listing. If your city looks like Spain, do not open one more concept; open one that fills a measurable category gap.

The 3 numbers you should tattoo on yourself

Tattoo 45, 24 and 118, in that order. Forty-five daily orders is the viability floor at a USD 14 ticket: measure it with real aggregator data for your zone across six weeks before signing any lease, and if it falls short, do not build. Twenty-four percent is the blended commission ceiling your structure withstands at 31% food cost and 19% payroll; renegotiate the contract or shift volume to your own channel until you drop below that line, because every point above it eats the rent saving. One hundred eighteen dollars is what a single commission point costs monthly per 1,000 orders: pin it on the office wall as a reminder that a thirty-minute negotiation with the aggregator pays better than three weeks of tweaking portion weights. Start today with the first one: request the hourly order history for your area.

Where the business actually breaks?

Commission is not an expense, it is a partnership stake. When you accept 29% of the ticket, the aggregator keeps more gross margin than you keep after food cost and packaging.

In a ghost kitchen running a 14 USD ticket, every percentage point of commission is 118 USD a month per 1.000 orders. Moving from 29% to 23% is not a negotiation win: it is 708 USD a month you had already earned. The delivery radius outranks product quality. You may fry the best chicken in the city, yet if your polygon holds 900 households with spending capacity and seven direct competitors in your category, the maths refuses to close. Before touring warehouses, count households, count competitors, read the aggregator's peak-hour history for that zone and work out how many orders you can realistically capture at a 6% to 9% share of the category.

Where the business actually breaks — in practice?

Reviews work as a dispatch engine, not as vanity. The algorithms behind Rappi, Uber Eats and DiDi Food weigh rating, acceptance rate and prep time when they order the listing.

A virtual brand climbing from 4,1★ to 4,6★ gains shelf positions and, with them, orders that cost no ad spend at all. That is the cheapest channel in delivery, and almost nobody works it with a process. A hidden kitchen menu gets redesigned, never translated. A dish that travels 25 minutes inside sealed packaging is a different dish. Whatever holds temperature and texture belongs at the top of the digital menu, and whatever arrives poorly has to go even when it sells well in the dining room. I got this wrong for years, arguing that brand identity justified keeping the full menu; the cash said a 14-item menu built to travel beats a 32-item one with half of them weak.