Ghost kitchens: the real math

Frederick Sona

Operator behind a 21 brand virtual restaurant group across several ghost kitchens (2020 to 2022). Full-stack eCommerce and growth leader. About · LinkedIn

The night the numbers did not add up

I was standing over a laptop at the end of a Friday that had cleared roughly ten thousand dollars in gross revenue. That is a strong day for a single restaurant. We were running twenty one brands out of a set of ghost kitchen units, so on the surface the math looked like a real business. Then I scrolled down. After the kitchen operator took their storage and rent, after the delivery platforms took their commissions, after the platform marketing charges came out, after food and packaging cleared, after we paid the cooks and the shift managers who kept the whole thing moving, there was almost nothing left. The bank deposit that would land the next week did not look anything like ten thousand dollars a day. It looked like a lease payment on a hot dog cart with more anxiety.

That was the night I understood that the entire ghost kitchen category, as it was being sold in 2020 and 2021, was solving a real problem for one kind of restaurant operator and quietly extracting most of the upside from every other kind. I want to lay that out honestly, because ghost kitchens are not a scam and they are not dead. For the right operator, in the right posture, they still work. For the operator I was, in the posture I was in, they did not. Both things can be true.

The promise

The pitch was tight and, at the time, hard to argue with. Restaurants were losing dining rooms overnight. Delivery had become the entire business. Real estate was expensive, staffing a full floor was expensive, and the buildout for a real restaurant had always been the biggest fixed cost in the category. A ghost kitchen offered a shared commissary space with pre-built exhaust, plumbing, refrigeration, and delivery pickup logistics. You could rent a unit for a fraction of a real lease. You could run several brands out of one unit because there was no dining room to confuse anyone. You could stand up a new virtual concept in weeks instead of quarters. Delivery platforms would front the customer traffic, which meant you did not need a location that walked past a customer or a brand anyone had heard of.

The whole thing came wrapped in a narrative about brand agility. A hot chicken concept underperforming? Kill it. Test a smash burger concept next week. Run five brands out of the same kitchen because the ovens and the fryers are already there. Every conversation I had with a kitchen operator and every deck I read in the space used some version of the phrase, "the restaurant of the future has no dining room." I heard versions of it from people I respected.

There was a real kernel of truth in every part of that pitch. The problem is that the pitch never spent much time on the cost side of the equation, and the cost side is where the story actually happens.

What the kitchen actually costs

The unit rent by itself was almost never the number that hurt. What hurt was everything the kitchen operator layered on top of the unit. There were storage fees for dry goods, for refrigerated inventory, for freezer inventory, and depending on the operator there were separate fees for shared use of the walk-in and for pallet receiving from vendors. There were sometimes fees for kitchen support staff even when you were not using them. There were peak hour surcharges. There were platform integration fees that showed up when the kitchen operator sat between you and the delivery apps and handled tablet management or order routing.

None of these fees are unreasonable in isolation. A commissary building runs on real overhead. Refrigeration is not free. But when you stack all of them onto a small operator running out of one unit, the effective cost of your kitchen ends up nowhere near the headline rent number. It is common to see the actual cost of occupying a ghost kitchen unit come out at close to the cost of a small traditional lease, with none of the sales benefits a traditional location gets from foot traffic or from a dining room that lets a customer double their check with drinks and dessert.

The kitchen operators I worked with were not villains here. They had built real infrastructure businesses with real overhead, and the fees they charged were the fees their buildings required to break even. That is exactly the point. Their business had to survive, so it charged what it charged, and my business had to absorb the entire stack on top of my food margin.

The platform tax

Every order that came through the door came through DoorDash, Uber Eats, or Grubhub. That is the entire model. Virtual brands do not have walk-in customers by definition. Which means the commission each of those platforms takes on each order is not a marketing expense, it is essentially your entire cost of customer acquisition, and it comes out of the top of every ticket. Platform commissions in that window ran across a wide range depending on the tier of service, but a working assumption of roughly twenty to thirty percent of gross ticket per order is a reasonable directional number for what a small operator sees. On some platforms, on some order types, in some markets, the effective take rate went higher.

That commission is not the end of it. The same platforms sell in-app promotional placement, sponsored listings, discount campaigns, and boost programs. All of those show up as additional line items against the same gross ticket the commission already took a bite from. If you turn the promotions off, your search placement collapses and your order volume drops with it. If you leave the promotions on, your take home per ticket drops. In practice most virtual brands are running some level of promotion continuously, because turning them off is a slow way to disappear from the platform entirely.

Then there is the customer discount you run to stay competitive with the twenty other virtual brands within a two mile radius selling the same category. Ten dollars off a first order. Buy one get one on a specific item. Free delivery over a threshold. Every one of those cuts into what is left after the commission and the promo fee. By the time an order finally posts to your books, the ticket the customer saw is not the ticket you got paid on.

Food and packaging still cost money

One of the quiet lies of the ghost kitchen pitch is the implication that virtual restaurants somehow escape traditional restaurant economics. They do not. A ghost kitchen is still a restaurant. Every ingredient still costs what it costs. Meat is still meat, produce is still produce, cheese is still cheese, and the pricing does not care whether your dining room exists.

Packaging turns out to be worse than a traditional restaurant, not better. A dine-in restaurant serves the food on plates that come back to be washed. A ghost kitchen puts every single item into a delivery container, and the container has to be structurally sound enough that the food arrives looking like the photo the customer ordered from. That means clamshells that hold heat, cup carriers, sauce cups with lids that lock, insulated bags for the driver hand-off, printed sleeves, branded stickers to keep the bag sealed. Every one of those things is a real per order cost. When you are running twenty one brands, each with its own packaging system so the brand feels distinct in the customer's hands, packaging complexity is not a rounding error. It shows up in every single ticket.

Food cost in the ghost kitchen model is typically higher, not lower, than an equivalent traditional operation. You are buying at smaller volume per brand because your volume is spread across many concepts. You are paying for portioning consistency across a lot of menu items. You are eating more waste from cross-brand prep because you have to hit the exact spec of every concept, and a slow night on one brand still means opened inventory that has to be used or discarded. The number that keeps a traditional restaurant honest, cost of goods sold as a percentage of revenue, is stubborn in a ghost kitchen. It does not get smaller because you moved indoors.

Twenty one brands, one line of cooks

The most seductive part of the multi brand pitch is the idea that if you already have a fryer, an oven, and a grill running, adding another concept on top of the same equipment is basically free labor. Twenty one brands from one kitchen. That sounds like a compounding return on the same headcount.

It is not, quite. Every brand you add requires that your cooks know the recipes, the plating, the packaging, and the tolerances for that concept. Cross training is real work. When an order for the smash burger brand and an order for the hot chicken brand and an order for the salad concept all come in inside the same ticket window, the person on the line has to switch context, pull from different prep stations, package into different containers, and drop it all into the right delivery bag with the right ticket. Errors on that hand-off show up as one star reviews on the exact platforms that already have you in a promotion war.

Labor for a twenty one brand operation is not twenty one times cheaper than labor for a one brand operation. It is not even five times cheaper. It is directionally about the same as running a single well run kitchen with a broader menu, plus the additional cost of shift managers with enough attention span to keep the brand identities from bleeding into each other. What you gain in equipment utilization, you give back in staffing complexity.

On top of the labor cost itself, you carry the risk of a slow night for the whole set of brands at once. Every brand runs off the same kitchen. When the neighborhood goes quiet on a Tuesday, you cannot lay off one brand's worth of hours. You are on the hook for the full shift.

What ten thousand dollars a day actually meant

This is the part I want to sit with, because it is the part most operators considering ghost kitchens do not get walked through honestly.

Ten thousand dollars a day in gross revenue is a serious top line. Roughly three hundred thousand a month, well over three million a year, from a kitchen footprint that fits inside a shared building. On a slide deck, that is a real business. In our actual books, once the stack was applied, it was a business that could barely afford to keep running.

Take that ten thousand off the top and let the platform commissions come out first, because platforms get paid before you do. Somewhere around a fifth to a third of that number, depending on the mix of platforms and the tier of service, went to the platforms as commissions. Add the platform promotion spend, the sponsored placement, and the customer-facing discounts and you are already looking at less on your side of the ledger than most people imagine. Food and packaging then come out of what remains, and both of those are more expensive per order than a dine-in equivalent for the reasons above. What is left funds the kitchen operator's fees, the labor on the line, the shift management, the accounting, the insurance, the utilities layered onto the storage bill, and the small handful of overhead expenses that always exist. What lands in the operating account at the end of that stack, on a strong revenue day, is a number you can round to almost nothing.

That is the piece the marketing decks did not include. Ghost kitchens can produce impressive top line numbers because delivery platforms can drive real volume when the operator is willing to pay for placement. The category never had a demand problem in 2020 and 2021. It had a margin problem, and the margin problem was structural, not something an operator could optimize their way out of on the margins.

Owner operator or middle layer

Here is the distinction that mattered most in our case, and it is the distinction I would put on the wall of anyone considering the model.

Ghost kitchens can work when the operator running the kitchen is the same person who owns the concept. When the person on the line at midnight is also the person who set the menu, priced the items, negotiated with the platforms, chose the packaging, and lives on the outcome, the model has a real chance. That operator can cut a menu item that is bleeding margin the same week they notice it. They can decide to pull promotions off a platform that is not paying back. They can cross train themselves and one other person and run lean. They can absorb a slow month personally, because the business is their livelihood and not a line item on a portfolio spreadsheet.

Ghost kitchens are much harder when the operator running the kitchen is one layer removed from the ownership of the concepts. That was our shape. I was the investor operator behind the operator. Every layer of separation between the person making the margin decisions and the person making the food adds friction, information loss, and delayed reaction time. A traditional restaurant can tolerate that structure because dining room revenue and brand loyalty give the operation more slack. A ghost kitchen has almost no slack. The stack of fees on every order is so tight that a two week delay in cutting a losing menu item is enough to swing the month.

The kitchen operators and the platforms have both built their businesses to be exactly indifferent to whether you make money. Their revenue is your gross, not your net. A middle layer operator is asking the model to leave enough behind after two other businesses have already taken their cut. That is a hard ask, and in our case, on a very strong revenue base, it was not enough.

Who ghost kitchens actually work for

I do not want to end this by telling anyone considering a ghost kitchen not to do it. That would be dishonest. There are operators for whom the model is genuinely a fit, and I have watched some of them do well. If you are considering it, here is the honest shape of who wins in the category.

Owner operators with real skin in the game win most often. Someone who is going to be on the line, who owns the concept, who can make same day pricing decisions, and who does not have a payroll layer between them and the fryer. The lower the number of hands between the money and the food, the better the model works.

Operators with a lean menu win more than operators with a sprawling one. One or two concepts with tight prep, disciplined menu engineering, and a food cost target that leaves real room for the platform tax. Running twenty one brands the way we did was in the opposite direction of what actually keeps this model alive.

Operators with a real point of view on marketing spend win, because they can decide when to run promotions and when to hold them, and they can measure it. The default in the category is to let the platforms decide, and the platforms decide in favor of the platforms.

Operators using ghost kitchens as an expansion channel for an already successful brick and mortar brand win, because they get to amortize marketing and brand awareness across a footprint the platform did not create. If the customer already knows the brand from a dining room in town, the platform is just a delivery mechanism, not a customer acquisition channel, and the economics change materially.

Everyone else, including the pattern I ran, is fighting an uphill battle. Not impossible, but uphill.

What I took away

The ghost kitchen years cost more than they made in our case. What I took away from them was not that the concept was fraudulent, because it was not, and not that the operators of these buildings were villains, because they were not either. They were running the businesses they had built, and their businesses were designed to charge for the value they delivered. What I took away was a much more careful eye for the difference between a business that can produce top line revenue and a business that actually gets to keep any of it.

Every category I have looked at since carries some version of the same distinction. There are always platforms happy to sell you traffic. There are always middlemen happy to sell you infrastructure. There are always vendors happy to sell you packaging, software, marketing placement, staffing, and consulting. Each one of them is a legitimate business. Stacked on top of a thin margin operation, they are also, together, the entire margin.

The operator's job is to know exactly how tall the stack is on top of them, and to decide whether the concept can carry that weight. In our case, on the ghost kitchen bet, the concept could not carry the weight and I did not see it clearly enough soon enough. If you are the owner operator with your name on the door, and your menu is tight, and your marketing spend is deliberate, and you are willing to cross train and run lean, ghost kitchens are still a reasonable way to get a food business into the world without a two million dollar buildout. If you are not any of those things, the model will find that out very quickly, and the strong revenue days will not save you.

That is the honest version. I would still put money behind the right ghost kitchen operator today. I would just make very sure the operator was the owner, and that the menu was small, and that the person on the line at midnight understood the entire stack sitting on top of them.