A ghost kitchen is not a cheaper restaurant. It is a different business: production without a dining room, demand without walk-in discovery, and a P&L that lives or dies on delivery marketplaces. Traditional restaurants sell an occasion in a room you control. Ghost models sell a bag at a door you do not. Pick the model that matches the concept, not the one that photographs as “asset light.”

Operators jump to ghost because the buildout quote on a dining room scared them. Sometimes that is the right read. Sometimes it is how you launch a brand nobody can find, with 25–35% marketplace fees and a kitchen that still needs a hood.

Capex: What You Actually Avoid

Ghost formats can skip dining furniture, a lot of HVAC for occupied rooms, restroom counts sized for guests, and front-of-house finishes. That is real money. They do not skip Type I hoods, grease interceptors, refrigeration, fire suppression, or a cookline that can hit ticket times. If the menu fries, the mechanicals still fry you.

Shared kitchens and delivery-only parks trade capex for rent-plus-fees. You may get a bay, hood, and dock. You also get house rules, shared access, and a landlord who can raise rates when the park fills. A second-gen dining room with a working Type I system can be cheaper than a “turnkey” ghost bay once you add deposits, hood upgrades, and the first six months of marketplace spend.

Write two capex columns: ghost and traditional. Include deposits, FF&E, hood, IT, smallwares, and pre-opening labor. The gap is often smaller than a LinkedIn thread claims, especially if you still need a commissary that passes health.

Unit Economics: Fees vs. Four Walls

Traditional: you pay occupancy, FOH labor, and the cost of a room. You keep more of each dollar when a guest sits down or picks up at the counter. Ghost: you cut FOH labor and dining occupancy, then you hand a quarter of the check (sometimes more, after ads and refunds) to marketplaces. Pickup-only ghost helps. Marketplace-only ghost is a tax.

• Delivery mix: if 80% of sales are third-party delivery, model net after commission, packaging, and voids—not gross.
• Ticket times: delivery platforms punish you in ranking when the line misses. You still need a real cookline, not a studio apartment range.
• Packaging: ghost food lives in a bag. Budget packaging as a food-cost cousin, not a rounding error.
• Refunds and “never delivered”: they hit ghost P&Ls harder because you have no host to de-escalate in the room.

Run contribution after marketplace fees. If the ghost model only works at a 40% delivery mix you cannot control, it does not work.

Labor Shape

Ghost kitchens still need a kitchen manager, a prep plan, and coverage for spikes. You save servers, not cooks. If your concept is hospitality-forward, you did not save labor. You deleted the product.

Brand: Discovery Without a Dining Room

A traditional room is a billboard, a habit, and a place to recover from a bad ticket. A ghost brand is a thumbnail in an app. You pay for that shelf with ads, promo pricing, and reviews from guests who never saw a kitchen. Multi-brand ghost stacks can juice kitchen utilization. They also confuse guests and dilute reviews when four “concepts” share a make line.

If the concept needs atmosphere, alcohol service in a room, or a chef’s reputation, ghost is a mismatch. If the concept is a tight menu that travels, with a name you can own in search and on apps, ghost can be the right first box—especially in expensive urban pads where dining rent is the killer.

Do the brand work in restaurant concept development before you pick the four walls. Menu, ticket, and occasion decide the model. The model should not invent a concept that only exists as a logo on a delivery tile.

When a Ghost Kitchen Wins

Ghost wins when three things are true at once: the food travels, the trade area already orders delivery in your category, and dining-room rent or parking would break the traditional P&L. It also wins as a second unit for a proven dine-in brand that wants coverage in a dense delivery radius without another 2,800-square-foot lease.

Ghost loses when you need liquor margins from a bar, when the menu is fryer-plus-hold that dies in a bag, or when you were counting on walk-by discovery to fill a weak marketing plan. It also loses when the only “cheap” bay has the wrong hood or a lease that treats you like a pop-up.

Site still matters. A ghost kitchen in a park with no driver access, or in a trade area that does not order your cuisine, is just a hidden kitchen with rent. Score delivery demand, competitor density on the apps, and drive times the same way you would score a dining room—then pick the model whose economics you can actually operate.

If you cannot explain why the food is better as a bag than as a table, open a restaurant. If you cannot explain why a table is worth the extra capex, go ghost on purpose, not by default.