In-house ghost brands: the real cost and the line items nobody quotes you

What does it cost to launch a ghost brand inside your restaurant in 2026?

Launching a virtual brand inside a restaurant that already operates costs between 1,800 and 12,000 USD per brand as of August 2026, and the bracket depends almost entirely on how many new inventory items it forces you to buy.

That range covers product photography, platform onboarding and verification, packaging design, labeling and the adjustment of your hot line; it excludes first-month working capital and launch promotions, which usually add another 600 to 2,400 USD. The owner of a Mediterranean kitchen in Guadalajara billing 41,000 USD monthly in the dining room kept the grill cold from three to six in the afternoon: fifteen dead hours a week, already paid for in rent, electricity and the split-shift cook's wage. That gap, not the logo, is the asset. And no agency ever quotes it to you.

What each investment bracket includes, dollar by dollar?

The 1,800 to 3,500 USD bracket buys a virtual brand built on the inventory you already hold:

six to ten dishes derived from your current mise en place, catalog photography, listing on two platforms, generic packaging with a printed label and a review of line timings. From 3,500 to 7,000 USD you get proprietary visual identity, custom-printed packaging with minimum runs of 3,000 to 5,000 units, studio photography for twenty items and three or four exclusive SKUs. The top bracket, 7,000 to 12,000 USD, shows up when the concept demands extra equipment — a dedicated fryer runs around 2,200 USD, a small blast chiller passes 4,000 — plus twelve or more proprietary items and a month of paid placement inside the app. That last bracket is the one that almost never pays for itself. Profitability of a virtual brand is set by the food cost of the virtual menu plus the platform commission, not by the upfront outlay.

The number that decides profit or loss is not the investment

Sector-optimal food cost sits between 28% and 35% according to the National Restaurant Association, and at Masterestaurant we hold 32% as a hard ceiling per dish; above that, no delivery volume rescues the operation. Add a commission that in 2026 swings between 18% and 30% depending on country and service tier, and the arithmetic closes itself: 32% food plus 27% commission leaves 41% for payroll, packaging, rent and profit. With packaging at 6% and imputed labor at 14%, your real margin drops to 21% pre-tax. If virtual food cost drifts to 38%, that 21% becomes 15% and every new order pushes your cash closer to the bone. Every exclusive item you add to the virtual menu carries a holding cost that almost nobody books: cold room space, waste from slow rotation and one extra purchase order per month. With zero new SKUs, break-even arrives in six to ten weeks and the brand is incremental margin on fixed costs you already pay.

Shared inventory outranks the brand

With twelve proprietary items, break-even stretches to seven months, and by then half those items have expired in the walk-in. The difference between a ghost brand that contributes 4,000 USD of annual margin and one that burns 9,000 does not live in the concept: it lives in the SKU count. My working rule is blunt and I do not negotiate it: three new references maximum per virtual brand, and only if they turn at least twice a week. Four variables explain nearly all the price dispersion in these projects. Custom packaging comes first: moving from a generic label to your own printed box adds 1,400 to 3,000 USD because of printer minimums, roughly 35% of the jump between brackets. Second is dedicated equipment, which lifts the budget 2,000 to 5,500 USD when the concept will not fit your existing line. Third, the count of exclusive SKUs, which loads about 180 to 320 USD per item between initial purchase and first-quarter waste.

Four factors that move the price, and how much each one weighs

Fourth, photography: twenty studio dishes cost 900 to 2,100 USD, and bad photos sink in-app conversion by 15% to 25%. Geography matters too, because in markets with 132,000 food establishments and 41% formality (Acodrés 2025), printers and studios quote on a different scale. Commission gets deducted before you set the selling price; it is never absorbed after the fact. If your dining-room dish sells at 12 USD with a 30% food cost, listing it at that same 12 USD on an app charging 27% drives effective food cost to 41%, and you have just turned your best dish into the one draining the most cash. The correct platform price is 15.60 USD, the dining-room price divided by 0.77, and the customer pays it because they compare against other restaurants inside the app, not against your printed menu. An owner in Bogotá corrected those prices across 32 items and recovered 2,900 USD a month without selling one extra order.

Platform pricing is calculated BEFORE, never afterward

The obvious objection is that raising prices kills volume; in his case orders fell 6% and gross margin rose 19 points. I take that trade every time. There are four negotiating levers that work, and I use them in this order. First: demand an itemized quote and strip printed packaging out of the launch, because a 0.04 USD adhesive label on a kraft box performs the same for a quarter and saves you the printer minimums. Second: negotiate commission against committed volume; a brand guaranteeing 400 monthly orders usually earns 3 to 5 points off, which on 6,000 USD billed is 240 USD clean per month. Third: ask the agency for milestone payments — 30% at listing, 40% after the first operating month, 30% against break-even — and watch how many walk away. Fourth: launch on a single platform. Temporary exclusivity buys better listing position, and in 2026 that visibility is worth more than coverage.

Line time is the cost your P&L never shows you

I got this wrong for years, and I will say it plainly: I used to recommend judging these projects by projected order counts, the number everyone wants to see, when the variable that actually governs the outcome is line time. A delivery order landing at 20:40, at the peak of dining-room service, does not cost what the commission says; it costs the ninety seconds of a cook who stopped plating two table dishes carrying a 68% margin. None of that surfaces in any income statement until your dining-room rating slides from 4.6 to 4.1 and Friday occupancy drops 11%. Suppose you switch the virtual brand off between 19:30 and 22:00: you lose 30% of the night's orders, you keep the dining room intact and weighted margin climbs. That scheduled shutdown is the single most profitable decision in the whole project, and almost nobody makes it.

The four differences that decide the money

Shared inventory is the economic lever, not the brand. Every exclusive reference your virtual menu adds carries a carrying cost almost nobody books: walk-in space, spoilage from slow turns, one more purchase order per month. At zero new SKUs, an in-house ghost brand is essentially incremental margin on fixed costs already paid; at fourteen, you built a second restaurant with the first one's accounting, which is the worst of both positions. Platform commission gets netted out BEFORE you set the price, never after. If your dining-room dish sells at 12 USD with 30% food cost, publishing it at 12 USD on an app charging 27% pushes effective food cost to 41%. The correct platform price is 15.60 USD, and the customer accepts it because they compare against other delivery listings, not against your printed menu. Owners who resist that price gap are usually protecting a brand consistency their customer never perceives, since these are two channels with two cost structures.

The four differences that decide the money — in practice

Line time is the scarce resource, not the kitchen. An idle flat-top at four in the afternoon is worth gold; that same flat-top at nine at night, with twelve tickets queued, is worth less than nothing. That is why the Masterestaurant method cuts the virtual brand by dining-room occupancy rather than by fixed schedule: past 75% of tables, the brand pauses in the app. It sounds like leaving money on the table, and it does leave roughly 12% of potential orders, but it protects dining-room average check, which is where your high margin lives. Validating costs less than correcting. One trial month with a six-dish menu, a single platform and decent photography runs 1,800 USD and hands you an answer no market study will sell you. If at six weeks the virtual brand's contribution margin per kitchen hour does not clear 40% of what the dining room produces in that same daypart, the answer is to shut it down, and shutting it down there saves you the 9,000 USD of the full tier.

The four differences that decide the money — key points

That is what a Restaurant Model Canvas does when it is used properly: it turns a hunch into an experiment with an expiry date.