Physical restaurant or dark kitchen: which one wins, and the four alternatives almost nobody prices

What actually decides between dining room and ghost kitchen?

Rent as a share of sales decides it, and nothing else: above 12% you have a location problem, below 8% you hold an asset that would be absurd to give away.

A grill billing 9,400 USD a month while paying 1,500 USD in rent carries a 16% burden that eats almost the entire result, whereas the same site selling 26,000 USD drops to 5.8% and turns cheap. Delivery, which according to Euromonitor International already concentrates roughly 21% of foodservice spending in Latin America —three times its 2019 weight—, does not change that arithmetic: it relocates it. Swapping fixed rent for app commission means replacing a payment you negotiate once a year with another one of 18 to 30% on every ticket, which nobody ever negotiates. Measure your rent percentage this month before looking at anything else. The symptom shows up on Saturday: if your kitchen takes more than 28 minutes to turn a table because the line is handling app orders, the format already broke, and the register confirms it with falling tips and three-star reviews.

When the dining room falls short (and the number that gives it away)?

Two figures deserve a joint reading. First, delivery's share of total sales; once it passes 35%, dine-in service stops being the real priority no matter what you say.

Second, actual table occupancy at 1:30 pm and 8:30 pm: if it never clears 55% at both peaks, you are paying commercial square meters to store empty chairs. With those two signals in hand, keeping everything as it is costs money every week. Without them, touch nothing: an empty room at mid-afternoon is not a diagnosis. An operator with a low average ticket, a young brand and monthly sales under 12,000 USD finds natural ground here. The reason sits in the square meter: industrial space costs between a third and half of prime commercial space, and that gap frees up 900 to 2,200 USD a month currently spent on a storefront.

Pure ghost kitchen: who it fits, and what switching costs

Building one runs 15,000 to 35,000 USD depending on equipment, or 1,200 to 2,800 USD monthly if you rent a module inside an existing hub, and the ramp takes about twelve weeks before the platform algorithm gives you stable visibility. IBISWorld describes a US ghost-kitchen segment already consolidated in size and number of operations, so this is no experiment. One warning: you are born depending on a single channel, and that dependency carries a price. Physically segregating delivery production solves the conflict no app will ever solve for you. I mean a dedicated packing station, an independent pass and, where it fits, a cook assigned during peaks. Minor construction, the hood and the expedition furniture add up to between 8,000 and 18,000 USD, and six to ten weeks go by before the head chef stops cannibalizing the table shift. The profile that wins here bills from 30,000 USD monthly with 35% of sales coming through apps.

Hybrid with a segregated kitchen: keep the room and stop breaking it

By the third month the register shows it: table times return to 18 or 20 minutes and one-star reviews for delays drop off. Anyone who segregates halfway, same pass and same cook, buys nothing; they just sign a construction invoice. You rent out idle capacity —typically 3 pm to 6 pm— to an outside virtual brand and charge per plate produced or a monthly fee of 800 to 2,500 USD depending on city and volume. Alex Canter, founder of Nextbite and fourth generation of Canter's Deli in Los Angeles, built an entire business on this idea: kitchens already exist, they sit idle half the day, and somebody has demand with nowhere to cook it. Your investment is nearly nil, between 1,500 and 4,000 USD in adaptation and training. The risk, however, is reputational: if the guest brand ends up poorly rated, your health rating and your team's morale are what pay for it.

Host kitchen and virtual brands: monetizing the dead hour

This suits a restaurant with an oversized kitchen and split hours. It does not suit one already running at the limit through both peaks, because both will blow up. Zero construction, zero new rent: you create a second brand on the app with the same crew, the same oven and a trimmed menu of six to eight items sharing 80% of inputs with the main one. The outlay runs 400 to 1,200 USD across photography, platform onboarding and packaging, and breakeven typically arrives between month two and month four. iFood reached 100 million orders in a single month during August 2024, with more than 380,000 partner establishments across over 1,500 Brazilian cities, and that volume rewards whoever shows up in more searches. The trap hides right there: if your secondary menu shares product with the main one, it cannibalizes; if it shares a cook at peak hour, dispatch times explode.

Your own virtual brands from your current kitchen: the cheapest arbitrage

Launch the new brand with hours restricted to the slots where your line sits free today. Every commission point you stop paying is pure margin, and that is the least glamorous and most profitable route on this menu of options. A 25% commission on a 22 USD ticket takes 5.50 USD; the same sale placed by phone or your own site, with a courier at a 4 USD fee, leaves an extra 1.50 USD in the register. Multiply by 600 monthly orders and 900 USD appear that nobody was watching. Your own channel also hands back the customer data the app will never give you. That said, it demands consistency: below 200 direct orders a month the fixed courier does not hold, and you crawl back to the app in a worse position. Start with in-store pickup at 10% off, which costs nothing and trains your most loyal customers.

When NOT to change anything (and why staying is usually right)?

With an average ticket above 22 USD and more than half your sales born from people walking in to sit down, closing the dining room would be an expensive and rarely reversible mistake:

the lease is gone, the clientele scatters, and getting back to an equivalent corner costs double. A 24 USD plate served at a table keeps its full contribution to the result, while the same plate handed to an app loses 5 to 7 USD between commission and packaging. That is the paradox governing this whole decision: the format with the highest fixed cost has the healthiest unit economics. Cutting the fixed side by raising the variable side only pays off while the room stays unfilled; once it fills, the math flips. In the Masterestaurant method, Diego F. Parra always starts from the same order: food cost under 32% first, occupancy next, and only then location. Review your rent-to-sales ratio this week.

Four honest alternatives between the dining room and the ghost kitchen

Hybrid with a segregated line. You keep the room but physically split a delivery production line with its own packing station and its own pass. Implementation runs 8 000 to 18 000 USD in light build-out, extraction and expediting furniture. Learning curve: six to ten weeks until the head chef stops letting app tickets cannibalize the table shift. Who it fits: the restaurant already billing 30 000 USD monthly with 35% delivery that currently wrecks Saturday service. The gain shows in month three; the loss, if segregation is cosmetic, is table punctuality, which is the expensive asset. Host kitchen. Rather than relocating, you rent your idle capacity to somebody else's virtual brand and charge per order produced or a monthly fee. Entry cost: zero, on a three to six month contract. Learning curve: two weeks of somebody else's spec sheets plus one trial shift. Who it fits: kitchens idle more than 30% of the window between 3 and 6 p.m.

Four honest alternatives between the dining room and the ghost kitchen — in practice

Alex Canter, founder of Nextbite and fourth-generation operator of Canter's Deli in Los Angeles, has argued publicly for years that the industry's wasted asset is not the brand but the oven sitting cold mid-afternoon, and his published figures point to 20% to 30% added revenue over the base with no structural change. Best risk-return of the four, and the one fewest owners price. Virtual brand inside your own kitchen. A second banner appears, with its own menu and its own pricing, produced in the same space and sold only through Rappi and iFood. Cost: 1 500 to 4 000 USD across photography, spec sheets and platform onboarding. Curve: four weeks. Who it fits: operators with labor headroom and a pantry that carries a second menu without buying five new inputs. One hard rule saves real money here: if the virtual brand needs more than six additional SKUs, it is not a virtual brand, it is a new restaurant in costume, and your inventory carrying cost will prove it within the quarter.

Four honest alternatives between the dining room and the ghost kitchen — key points

Shared kitchen by module. You rent a 12 to 20 m² unit inside an operated hub with extraction, cold rooms and logistics already solved. Latin American pricing: 900 to 2 200 USD monthly depending on city, almost always month-to-month. Curve: one week. Who it fits: testing a brand hypothesis in a new city without signing five years. The limit nobody mentions is ownership — that module is not yours, the hub can raise the fee on ninety days' notice, and your brand has no plan B if that lands in high season.