How a ghost kitchen works: traditional method vs Masterestaurant method

Chapter 1 — What a ghost kitchen actually is, and where its revenue comes from

A ghost kitchen is a production unit with no dining room and no storefront, and its revenue arrives through a digital channel — that is the difference almost nobody accounts for properly. The ghost kitchen market keeps growing at a strong double-digit pace year over year, figures that attract capital as easily as they attract improvisation. A place with a storefront sells through location; you pay rent and buy foot traffic. A ghost kitchen sells through RANKING inside an app, and that ranking is either bought with commission or earned with digital reputation, which is neither the same thing nor the same cost. Whoever builds the second using the logic of the first ends up paying twice: commission to the aggregator, plus square meters nobody ever visits.

Chapter 2 — Commission is not a marketing expense: it rewrites your food cost

Copying dining-room prices onto the delivery menu destroys margin through arithmetic, not bad luck. If the aggregator charges 25% commission, you collect 7.50, and those same three dollars now represent 40% of what reaches your register. Ten margin points gone without a single recipe changing. Food inputs have risen sharply since 2019 in the United States, and several large chains answered by raising menu prices well above general inflation for the period. The sector already rebuilt its pricing; the independent delivery operator almost never did. A mirror menu, with its own prices for the digital channel, defends contribution margin.

Chapter 3 — Orders per hour: the variable that decides whether the kitchen survives

Ghost kitchen margin is settled in orders per hour across the same production line, and this is the reading I argue about most with owners convinced their problem is price. One griddle, one fryer and two hands carry fixed installed capacity; if that capacity produces eight orders an hour, every order absorbs the full labor cost of the shift, and with a base wage of 14.20 dollars per hour in U.S. restaurants after a 4% increase in 2024 according to 7shifts, we are talking about labor that only dilutes with volume. Doubling to sixteen orders an hour without hiring anyone cuts that cost in half. Multiple virtual brands on one line are therefore no growth gimmick: they are the natural mechanism for filling demand valleys you pay for anyway.

Chapter 4 — The revenue-band effect: the same model does not perform the same way

Revenue band completely changes which lever moves the result, and confusing them is the costliest diagnostic error. Below 500 thousand dollars a year, a 25% commission eats the owner's own salary: the only real way out is direct ordering, even if it accounts for barely 20% of volume. Between 500 thousand and one million, room appears to negotiate rates with the aggregator and to sustain a second virtual brand on the same line. Above one million, the conversation moves to data: every order leaves a phone number, a delivery zone and a frequency, and that asset monetizes through owned campaigns which, according to Stripo, lift open rates by 26% with personalized messages. Above five million, multi-site hubs and centralized purchasing arrive; above ten million, the dominant variable stops being the kitchen and becomes last-mile logistics.

Chapter 5 — The high end: the celebrity chef's ghost kitchen and its own costs

Above five million dollars a year, the signature ghost kitchen — a chef with a public name, or a large-format themed concept — plays with a cost structure that rarely gets explained. In exchange, costs appear that the small band never meets: image licensing, multi-site quality control, premium packaging and waste from a high standard. The classic mistake in this segment is scaling sites before standardizing the recipe spec, and the outcome is an expensive brand serving irregular product.

Chapter 6 — Owning demand: the aggregator is not your channel, it is your landlord

In the traditional delivery model the customer belongs to the aggregator and you rent their attention every month, which explains why so many operations with rising sales get low valuations when somebody wants to buy them. What gets bought is not the kitchen, replicable with modest investment, but the customer base with demonstrable frequency. That is why the Masterestaurant method applied by Diego F. Parra pushes operators to capture data on every order and build an owned channel in parallel, without breaking from the aggregator while it still delivers discovery. The indicators support that route: Toast reports that in 2025, 28% of AI use cases in restaurants was marketing automation and 27% real-time insights, tools worth little when your customer data lives on somebody else's server that you cannot reach.

Chapter 7 — What happens if commission rises three points next year?

Raise the commission mentally from 25% to 28% and follow the consequence all the way down, because that scenario already happened in several markets and will happen again.

On a ten-dollar ticket with three dollars of food cost, you go from receiving 7.50 to receiving 7.20; effective food cost jumps from 40% to 41.7%. It looks small. On 30 thousand dollars of monthly sales it is 900 dollars taken straight out of the profit line, roughly a cook's salary. There sits the paradox I always resolve the same way: staying with the aggregator makes sense, and building the exit at the same time makes sense too, because total dependence is what sets your price.

Chapter 8 — What to do Monday: three numbers before any decision

Before opening another virtual brand or negotiating with an aggregator, measure three numbers from your own operation across fourteen consecutive days. First, orders per hour during the real peak window, not the daily average, because averages hide the valley that keeps costing you payroll. Second, effective food cost after commission, calculated on what lands in your bank account rather than on the app's price. Third, share of sales through your own channel — the one many owners believe sits at 15% and shows up at 4% once measured. With those three figures the decision becomes obvious and stops being opinion. The sector already showed where efficiency is heading: Chipotle projected opening between 315 and 345 locations in 2025 with more than 80% equipped with a drive-thru, according to Chain Store Age, betting on fast-pickup formats instead of more dining room.

Chapter 9 — The five differences that decide the margin

OWNERSHIP OF DEMAND. Under the traditional model the customer belongs to the aggregator: you rent their attention every month. Under the Masterestaurant method every order leaves data —phone, polygon, frequency— that feeds an owned asset; and that asset, not the kitchen, is what someone actually buys when they buy the operation. BUILT PRICE VS COPIED PRICE. The mirror menu is not a marketing trick, it is margin survival. DENSITY ON THE SAME LINE. A ghost kitchen running one brand wastes installed capacity. With two or three virtual brands sharing mise en place —same base protein, different finish— hourly output rises without adding a single square meter of CapEx, and break-even drops because fixed OpEx spreads across brands.

Chapter 10 — The five differences that decide the margin — in practice

LOCAL SEO AS A CHANNEL, NOT AS DECORATION. A verified Google Business Profile with a precise category, fresh reviews and weekly posts surfaces in immediate-intent searches inside the polygon. That traffic pays no commission, and in mature operations it accounts for 10% to 20% of owned-channel sales. MEASUREMENT PER ORDER, NOT PER MONTH. The traditional dashboard watches monthly revenue; ours watches contribution per order, per brand and per time band. Once you see the hero combo leaving 2.10 USD and the photogenic dish leaving 0.40 USD, menu engineering stops being theory and becomes Tuesday's decision.