Dark kitchen vs traditional restaurant: the demand engine decides, not the kitchen

When a pure dark kitchen stops being enough?

A pure ghost kitchen stops working the day aggregator commissions pass 25% of your sales, because beyond that no rent savings can cover them.

The P&L I reviewed from a Medellín operator shows it plainly: 92 million pesos billed in July, food cost at 29%, payroll held at 24%, and an operating result of 1.8% because commissions and paid ads together took 31% of sales. The number that gives it away is not the margin, it is the proportion: when more than 80% of your orders arrive through a platform that sets the commission and keeps the buyer's data, you stopped running a restaurant and started subleasing kitchen capacity. That is the moment to look at alternatives, and not one month earlier. Pure ghost kitchens work for the operator who already has proven demand and simply needs extra production capacity, usually a brand with a digital sales history that wants to cover another zone without duplicating a dining room.

Pure dark kitchen: who it fits, and what entry costs

Opening investment drops between 40% and 60% against a storefront location, and break-even usually lands between month four and month seven, versus twelve to eighteen months for a traditional restaurant. Switching cost is low in money and high in dependency: signing a shared warehouse costs little, leaving costs even less, yet throughout that contract your entire customer acquisition sits in someone else's hands. It fits you if you sell more than 60 orders a day from day one. If you plan to BUILD demand from zero in there, the format will charge you twice for every order. A street-facing dining room still wins when demand does not exist yet, because it receives a flow of customers that no marketing department ever paid for. A Google Business Profile with a visitable address, a real façade, opening hours and fresh reviews is the cheapest acquisition asset in this trade, and «near me» searches concentrate a purchase intent no campaign buys at that price.

Traditional restaurant: the traffic nobody invoices

The profile is clear: new brand, medium or high average check, an offer that needs physical experience to justify itself. Entry cost hurts —build-out, deposits, servers, three to six months operating in the red— and exit cost hurts more, since a commercial lease rarely breaks for free. Diego F. Parra puts it this way in Masterestaurant audits: if you are going to pay for every customer, that customer had better come back on their own. The hybrid solves the underlying paradox: a small twenty to thirty seat room that sustains a verified Google Maps listing, built over a kitchen sized to produce three or four times that capacity in delivery. You pay rent for square meters that genuinely generate organic discovery, and you amortize those meters with digital volume that does not depend on the aggregator alone. In rough numbers, investment lands 20% to 30% above a dark kitchen and 30% to 40% below a full traditional restaurant, with a trimmed service payroll because the dining room is not the center of the business.

The hybrid: small room, oversized kitchen

This suits the owner whose brand is already known in the neighborhood and who wants to scale without giving away margin. Switching effort is medium: it means renegotiating layout, and sometimes moving. Launching a virtual brand inside your current kitchen is the cheapest alternative available, and almost nobody weighs it before signing a warehouse lease. You already pay the rent, you already own the equipment, your kitchen payroll is already covered: the second brand only adds variable cost in ingredients and some packaging, so break-even can show up in weeks rather than quarters. The hard condition is capacity, not creativity: if your kitchen already runs above 70% occupancy during peak hours, an extra brand will wreck dispatch times and the ratings of both. It serves the operator with a kitchen that sits loose midweek and demand concentrated on weekends. The real risk is focus, and that one never shows up on any spreadsheet.

Your own ordering channel: trading commission for work

An owned channel trades a commission of 22% to 30% for a steady retention effort that almost nobody sustains. The arithmetic favors it brutally: with an average commission of 27% and five points of rent saved, every extra peso of digital sales in a ghost kitchen costs you 22 cents more than in the dining room, and that gap is precisely the budget that funds your own channel. The crossover point where a dark kitchen stops being cheap tends to appear between USD 55,000 and 70,000 in monthly sales. Scale on the other side explains the laziness: iFood moved 100 million orders in the single month of August 2024 (Statista), with over 380,000 partner establishments across 1,500 Brazilian cities, according to iFood. Competing with that from your own site looks absurd, and still the returning customer is the only asset that stays. If your bottleneck is the cost of taking and delivering the order, automation is already a measurable alternative rather than a trade-show promise.

Automation and robots: the alternative that cuts cost per order

White Castle's voice AI reaches a 90% order completion rate at roughly 60 seconds per order, according to SoundHound as reported by Restaurant Dive, and by the end of 2024 it ran in more than 100 drive-thrus. Wendy's projected FreshAI across 500 to 600 US locations by the close of 2025 (CNBC). On the logistics side, Starship robots completed 5.8 million deliveries in 2024 according to Forbes, and Serve Robotics passed 50,000 commercial deliveries in Los Angeles per its Form 8-K filed with the SEC. The profile that benefits runs high volume with a short menu. For a thirty-order-a-day operation, the payback simply is not there yet. Sometimes staying where you are is the right call, and I would rather say so than sell you a move. If your traditional restaurant leaves 12% or more in operating margin and 60% of your sales come through the dining room, switching formats strips away the asset holding the business up to solve a problem you do not have.

When NOT to change format?

If your dark kitchen already crossed USD 55,000 a month yet cash flow holds and ratings sit above 4.6, do not rush either:

cash flow is the leading cause of financial stress and closure among small businesses, according to Inc., and a badly financed move bankrupts profitable operations. The question that settles everything is uncomfortable: is your problem production or demand? Measure what share of this month's sales came from repeat customers; that number decides, not the format. COMMISSION versus RENT. A ghost kitchen trades rent for commission, and that swap only pays below a certain volume: with 27% average commission and 5 points of rent saved, every extra dollar of digital sales costs you 22 cents more than in the room. The crossover usually shows up between USD 55,000 and 70,000 in monthly sales, and that is where the dark kitchen stops being cheap.

Five differences that change the outcome

DISCOVERY. Google Maps rewards what it can verify: visitable address, storefront, hours, photos. A dark kitchen inside a shared warehouse rarely clears verification with any strength, and without a strong profile it loses «near me» search traffic, which in local dining carries more purchase intent than any campaign. CUSTOMER OWNERSHIP. In the dining room, the returning guest is yours. On the aggregator, the algorithm decides whether it shows you again, and the moment you pause geotargeted ads your ranking drops within the same shift. A business that cannot recontact its base has no base. MENU ELASTICITY. A ghost kitchen lets you stack two to four virtual brands on one production line and spread fixed costs across all of them, which a dining room with printed menus and décor expectations cannot do. Executed well, it lifts sales per square metre by 18% to 35%. MEASURABLE REPUTATION. Your dining room's five-star reviews live on Google and work everywhere, delivery included.

Five differences that change the outcome — in practice

In-app ratings on Rappi or Uber Eats never leave the app, and a slide from 4.8 to 4.4 inside the platform can erase half your impressions while your public reputation has not moved a millimetre.