Physical restaurant or dark kitchen: which one suits you, with the numbers that actually decide

Step 1: measure where your demand comes from before touching the lease

Before comparing rents, split your last 90 days of revenue into three channels and assign each one a percentage: foot traffic, local discovery on Google Maps and aggregator delivery. The deliverable is a sheet with three figures adding up to 100%, cross-checked against your Google Business Profile panel (discovery searches, calls, route requests) and the monthly commission report from Rappi, iFood or Glovo. When more than 45% of your sales come from walk-ins or local discovery, moving into a hidden kitchen amputates half the business and no rent saving covers that. Verification: the three figures must reconcile with your reported sales for the period within a 3% gap. Without that sheet you are not deciding anything, you are gambling with the year's cash. Delivery contribution margin is calculated after commission, packaging and the aggregator's mandatory promotion, not just food cost. Take your average ticket, subtract food cost —which on a healthy dish never exceeds 32%—, then the aggregator commission (18% to 30% of the ticket depending on platform and country) and add 4 to 7 points of disposable packaging.

Step 2: calculate real contribution margin per channel, commission included

What remains is your REAL contribution per order. Run the same exercise for a dine-in order: no commission, no packaging, same food cost. The gap usually sits around 20 percentage points, and that gap is what pays the rent on a physical location. The deliverable is a two-column table showing contribution in currency per order and per channel. Verify it by multiplying contribution by the channel's monthly volume: the result should land close to your accounting gross profit. In a hidden kitchen the sales ceiling is set by the production line, not by the dining room. A well-designed 24 m² line tops out between 380 and 520 orders a day at peak, and once you hit that you cannot grow without opening another unit, because there are no tables to absorb demand at a higher ticket. A restaurant with a dining room has two independent levers: table turnover and average ticket, plus a kitchen that still serves delivery.

Step 3: project each format's capacity ceiling, because they are different animals

Multiply the order ceiling by the contribution per order from the previous step and you get each format's monthly profit ceiling. The deliverable is a single figure per scenario. Honest verification: if the cheap format's ceiling fails to cover your break-even plus a 25% cushion, that format is out even when the rent looks like a bargain. Google Business Profile treats a business with a visitable address differently from a service-area listing, and that difference decides this case. A listing with an address competes on proximity in the Maps local pack; a delivery-only one loses that advantage, which weighs more than any other local ranking factor. Audit today: review count, average rating, date of the last review, response rate and photos posted in the last 60 days. If your listing has fewer than 40 reviews, the latest one is over three months old and you never reply, the digital asset supposedly defending your location is already dead and the argument for the physical format collapses.

Step 4: audit your digital asset and decide whether you can sustain it

At Masterestaurant, Diego F. Parra puts this step ahead of the real estate one: first you measure where demand comes from, then you pick the format. Assume you move into the hidden kitchen and six months later the park operator raises the fee or changes its commercial model. What do you have left? Nothing transferable: your reviews live on the aggregator, not on your brand; your traffic was rented from the park operator; and going back to a physical location means starting the Google listing from zero, with the maturation cycle that demands. The reverse works better, because a business with a mature listing and neighborhood regulars launches its delivery vertical with a known brand. Write both exit routes in a paragraph each, with the reversal cost in money and in months. The deliverable is that document, signed by you. If one route has no writable exit, the decision is not reversible, and irreversible decisions demand more evidence than a rent discount.

The four mistakes that sink this decision, and how to dodge them

The costliest mistake is treating this as a real estate problem when it is a channel problem: who brings you the customer and what they charge for it. The second is comparing gross rent against gross rent while ignoring the aggregator commission, which in practice is variable rent of 18% to 30% of the ticket with no contract fixing it. The third is assuming reviews move with you: 900 reviews on an aggregator are worth zero on Google. The fourth, quieter one, is projecting delivery volume from market growth figures —Statista puts Latin American meal delivery above 39 billion dollars by 2027, and Mexico City already runs more than 1,200 dark kitchens according to CANIRAC, up 40% since 2023— when a market that grows and saturates at the same time raises your acquisition cost instead of lowering it. Fix each mistake with your own number, never the sector's.

What renting your channel really costs: the fine print of aggregator dependence?

Depending on an aggregator means a third party sets your acquisition cost and can change it tomorrow. Concentration explains why:

in Spain, Glovo holds roughly 31% of the market and Just Eat 26% according to Ken Research 2025, so two players govern more than half the channel. In Brazil, iFood reports over 380,000 partner establishments across more than 1,500 cities and hit 100 million orders in a single month of 2024 according to its own published data. You do not negotiate with that scale, you accept it. A physical location buys traffic ONCE, through location and years of a mature listing; the hidden kitchen rents it monthly and the landlord adjusts the fee whenever its commercial model calls for it. None of that disqualifies the dark format —it disqualifies using it as your only channel. The rule I hold: never let a single aggregator pass 40% of your sales.

Closing checklist: how to know the decision landed right

Your decision is settled when you can show six documents and none of them embarrasses you. One, the demand-by-channel sheet with three figures reconciled against sales. Two, the contribution table per order and per channel, commission and packaging included. Three, each format's monthly profit ceiling against your break-even plus 25%. Four, the Google listing audit with reviews, recency and response rate. Five, both exit routes with cost in money and in months. Six, the written dependence limit per aggregator. If any of the six is blank, hold off on signing the contract. And one final test that never fails: ask your accountant how much net profit the chosen format leaves in month twelve. If that number takes longer than ten minutes to appear, the model does not exist yet. A physical location buys traffic ONCE, through its address; a ghost kitchen rents it every month, and the landlord charges 18% to 30% of the check without prior notice when it adjusts its commercial model.

The differences that decide the cash, not the ones argued on LinkedIn

Google Business Profile treats a visitable establishment differently from a delivery-only one: the service-area listing loses the proximity advantage inside the Maps local pack, which is the heaviest filter in local ranking. Aggregator reviews never transfer. Pile up 900 reviews on Rappi, then open a dining room, and you start at zero on Google; the reverse holds too, though a business with a mature listing launches its delivery vertical with a known brand. In a dark kitchen the kitchen caps sales, not the room: a well-designed 24 m² line tops out between 380 and 520 orders on a peak day, and growth stops there unless you open another node. A dining room recovers margin through suggestive selling, pairings and dessert; in delivery the check rises through combo packaging, and that lever ceilings around 14%-18%. An aggregator algorithm change swings a ghost kitchen's sales ±35% inside seven days. The same change moves 12% of a dine-in restaurant's total, because the rest walks through the door.