How a Ghost Kitchen Works: Before vs After with Masterestaurant

What a ghost kitchen is — and what it is not?

With no dining room and no front-of-house staff, a ghost kitchen turns 100% of its square footage into pure production for digital platforms such as Rappi, Uber Eats, or DoorDash:

no servers, no table-service register, just the cook line and dispatch. London saw the model born in 2013, with Deliveroo Kitchen, and the pandemic mainstreamed it: registered dark kitchens in Mexico, Colombia, and Argentina tripled (3x) between 2020 and 2023, and Statista puts the global market above USD 65 billion in 2025. What it is NOT matters as much as what it is: not a restaurant with a pickup window, not a generic shared catering kitchen, not an informal operation running without recipe cost cards. The opening error is conflating the two models, something Diego F. Parra corrects often through the method now applied across more than 8,400 restaurants in 43 countries: carrying 18%–25% dining-room costs without the in-person sales to justify them destroys margin before the first order ships.

How the order and production flow works?

In 28 to 35 minutes, door to door, in dense urban areas:

that is how fast the entire ghost kitchen cycle closes, from the order hitting the platform tablet to a production ticket firing on the cook line and leaving with a third-party rider the operator never manages directly. Cold, hot, and packaging organize the stations, always in one direction to prevent cross-contamination, and that flow discipline cuts dispatch errors by up to 30%. Well designed, a 30–60 m² kitchen staffed by 3–5 cooks dispatches 80–120 orders a day at peak, assembling each dish in under 8 minutes. Here is the blind spot most operators never calculate before opening: below 40 orders a day the model loses money even with controlled food cost, because rent (8%–10% of sales), utilities, and permits do not shrink when sales drop. Break-even, on average, falls between 55 and 70 daily orders.

Cost structure: where it wins and where it bleeds

Payroll is where the advantage of skipping a dining room shows up first: a sit-down location dedicates 18%–25% of sales to servers, host, and captain, while a dark kitchen at the same volume spends only 4%–8%, because that headcount simply does not exist. Those 14 to 17 percentage points become extra profit or a buffer for platform commissions, which run 15%–30% per order. But food cost is where operators actually bleed: without recipe cost cards, they underestimate protein trim loss (8%–14%) and the supplier price swings the USDA tracks in its food price index. Payroll, rent, and utilities belong to break-even, never to the plate; food cost, in the Masterestaurant method, is the ONLY direct cost of the dish, and the model turns profitable when it stays in the 22%–28% range (32% is a ceiling, never a goal), platform commission stays under 25%, and rent stays below 8%–10% of net sales.

Cost structure: where it wins and where it bleeds — in practice

All three levers need control at once. Launching a ghost kitchen costs between USD 8,000 and USD 25,000, depending on country, city, and available equipment, against the USD 80,000–200,000 a full-format restaurant with a dining room requires: a barrier to entry 6 to 10 times lower. Kitchen equipment absorbs 40%–50% of the total, facility build-out another 20%–25%, 60-day working capital 20%–25%, and health permits plus licenses the remaining 5%–8%. I have seen it in dozens of launches: the budget inflates when an operator with no sales history oversizes the kitchen, installing industrial grills for 200 daily covers and locking up capital in idle capacity that takes 14 to 18 months to recover. Size for 70% of projected month-3 volume, and scale equipment only once the daily order average clears 85% of installed capacity for two straight weeks: that is the Masterestaurant rule of thumb, validated across 43 countries.

Menu engineering: the most underrated weapon

A ghost kitchen menu is not the same animal as a dining-room menu, and confusing the two is, for Diego F. Parra of Masterestaurant, one of the costliest mistakes in the business. A delivery menu needs dishes that travel well, with stable temperature and preserved texture over 20 to 30 minutes, a maximum food cost of 28%, and prep time under 8 minutes during peak hours. Monthly menu engineering maps profitability and popularity on a 2×2 matrix, and that mapping is what eliminates dishes generating only ingredient waste while pushing price on the workhorses. Twelve to eighteen items, not forty: that is how the best-performing dark kitchen operators run their menus in practice, because fewer SKUs mean less trim waste, more concentrated purchasing, and supplier prices 8%–12% better per unit volume. Datassential and Circana back this with data: tight delivery menus convert up to 20% more per session by cutting on-screen decision friction.

Cost control with the Masterestaurant method

Standard costing by recipe card, monthly menu engineering, and portion control down to the gram: with those three levers, the Masterestaurant method drops a ghost kitchen's food cost from the initial 38%–45% to the 22%–28% range within 60 to 90 days of disciplined operation. The recipe card records actual cost per portion, including prep trim loss and supplier price variance, and without it the operator prices by intuition and loses 6 to 11 margin points to untracked portion drift. Every week, a physical inventory check against theoretical consumption exposes the leaks: in kitchens implementing this control for the first time, the gap between theoretical and actual inventory typically runs 5%–9% of sales, money lost to unrecorded generous portions, hidden spoilage, or product loss. Masterestaurant's CASH tool automates that reconciliation and flags any category above 32%, the red line Diego F. Parra sets for food cost per dish.

Delivery platforms: commissions, visibility, and dependency risk

Between 15% and 30% per order, depending on country, city, and membership tier: that is what delivery platforms charge, a fixed cost that is unavoidable from day one. Avoiding them is not the smart strategy; using them as a customer-acquisition channel while building a proprietary base is. Generating 100% of sales through a single platform leaves a ghost kitchen critically exposed, because one algorithm change, one commission hike, or a penalty from low ratings can wipe out 80% of revenue within 48 hours. This is where AI applied to restaurants pays off: models that forecast demand by time slot and adjust menu availability in real time lift average check by 6%–12%, per McKinsey's reporting on foodservice digitalization. No more than 60%–70% dependence on any single platform, and 3%–5% of net sales invested in owned-channel development (WhatsApp Business, customer database, direct-order discounts) before month 6: that is the healthy benchmark.

When a ghost kitchen makes sense — and when it does not?

When a restaurant owner already running a location wants to scale without signing a new 10-year lease, or when an entrepreneur wants to enter the food business with limited capital, a ghost kitchen makes sense.

It stops making sense when the area lacks rider density (fewer than 15 active riders per km² at peak hours pushes delivery times above 45 minutes and wrecks ratings), when the menu needs high-complexity techniques that do not survive transport, or when the operator lacks cost discipline. Between 2022 and 2025 food delivery in Latin America grew 38% (Statista, 2025), and ghost kitchens now capture about 14% of that market with annual expansion rates above 22%, a figure the National Restaurant Association corroborates in its industry research. The model works only when the unit economics validate the operation before the doors open, and that is why Masterestaurant sits the operator down with recipe cards and break-even math before the first order ships.

The differences that matter for your bottom line

In labor is where the difference shows up first: a dine-in restaurant dedicates 18%–25% of sales to front-of-house staff, while a ghost kitchen of the same volume spends only 4%–8%, because there are no servers, hosts, or captains. Those 14 to 17 percentage points end up as either additional profit or a buffer for platform commissions, which run 15%–30% per order. Whether a ghost kitchen turns out to be a business or a trap gets decided on the food-cost battleground. Without a recipe-level costing system, operators routinely underestimate food waste (8%–14% for proteins) and supplier price fluctuations. Standardized recipe cards, monthly menu engineering, and portion control with exact weights: with the Masterestaurant method, food cost drops from the initial 40%+ to the 22%–28% range within 60 to 90 days. Rappi, Uber Eats, and DoorDash charge between 20% and 32% of order value depending on city and contract terms, and that is the structural risk baked into the model: platform dependency.

The differences that matter for your bottom line — in practice

Without a direct channel (WhatsApp Business plus its own payment link), a ghost kitchen is trapped, because if the platform raises commissions or drops the operator's ranking, revenue can fall within 48 hours. Masterestaurant recommends the direct channel represent at least 20% of sales before month 4. Few operators exploit a real ghost kitchen advantage: menu iteration speed. Without printing physical menus or training waitstaff, you can test a new dish in 72 hours, measure its digital conversion rate, and remove or scale it with real data, something a traditional restaurant takes 3 to 6 weeks to do once design, printing, and team training are counted. Diego F. Parra weaponizes that agility with weekly menu engineering: iterating the mix every 7 days, not every quarter, lifts contribution margin by 3 to 6 points in the first quarter.