How to start a dark kitchen from scratch: where the model breaks and four honest alternatives

When your own dark kitchen stops being enough?

The number that exposes the problem is net ticket: if your aggregator orders leave you less than 60% of menu price at month close, an owned dark kitchen is no longer working for you.

ActiveMenus calculates that the true all-in cost of third-party delivery runs between 30% and 40% per order once you add commission, packaging, forced promotions and refunds credited to the customer, and on that base a five-year lease stops being leverage and becomes a shackle. Add labor cost, which the U.S. Bureau of Labor Statistics places between 25% and 35% of revenue in food services, and you have fifteen points left to cover ingredients, utilities and profit. It does not add up. The format falls short precisely when the volume promised by the aggregator heat map fails to show up in week twelve and you already own a hood, a grease trap and a notarized contract.

Option 1: shared kitchen rented by the hour or by station

For the owner who has not yet validated a menu, a shared kitchen is the only honest way in, because it turns a fixed cost into a variable one and can be abandoned in thirty days at the price of a deposit. You rent a station with certified ventilation, cold storage and permits already sorted, and you pay by time block instead of signing a lease. Profile: an operator with one or two anchor recipes, under 15,000 USD of capital and no sales history. Switching cost: low, thirty days notice. Effort: medium, since you share the extractor and work inside fixed dispatch windows. The downside is real and worth stating: you do not control the fiscal address long term, you operate beside neighbors competing for the same courier, and you cannot run four virtual brands at once without colliding with the building schedule. What you buy instead is information, learning the rhythm of timed dispatch before risking hard capital.

Option 2: host kitchen inside a restaurant already running

Signing with a restaurant that already has a built kitchen and idle hours solves two problems at once: it removes the infrastructure investment and hands you a team that already plates under pressure. The usual deal pays the host between 15% and 25% of gross sales, a figure that sounds steep until you set it against the 30% to 40% true cost of third-party delivery reported by ActiveMenus, where you also pay rent separately. Profile: an owner with a recognizable brand and a short menu who needs to cover a new zone without capital. Switching cost: very low, usually a month-to-month contract. Real risk: your quality depends on someone else's discipline, and during peak hours the host dispatches their own orders first. Diego F. Parra insists across Masterestaurant diagnostics on locking maximum dispatch times in writing plus a recipe audit clause, because without that you are lending out your reputation.

Option 3: a virtual brand with no new kitchen, inside your own venue

If you already run a restaurant, launching a second virtual brand in the same kitchen usually pays better than any new dark kitchen, because rent, hood and base payroll are already covered and you only carry the variable cost of the dish. The leverage is arithmetic: with labor cost placed by the U.S. Bureau of Labor Statistics between 25% and 35% of revenue, every incremental sale that arrives without a new hire travels almost whole into contribution margin. Profile: an operator whose kitchen occupancy sits below 65% in identified time bands. Switching cost: minimal, some photos, a menu and an aggregator listing. The condition without which none of this holds is the shared ingredient matrix: if the second brand needs proteins or processes your kitchen does not handle, you will double waste and break dining room timing. Start with six items, not twenty. An order arriving through your WhatsApp, your site or your Google listing leaves between 22 and 28 percentage points more margin than the same order coming through an aggregator, and that gap decides whether the business survives year two.

Option 4: an owned channel with pickup point and controlled delivery

The math comes from stripping out commission and cofunded promotions: where ActiveMenus measures a true all-in cost of 30% to 40% per third-party order, an owned channel pays for the gateway, contracted couriers and local ads, and rarely passes 12%. Profile: whoever already handles 400 monthly orders and sits on a phone list they have never used. Effort: high at first, because building your own demand takes four to six months of reviews, photography and replies under five minutes. An honest concession: through that ramp you will still need the aggregator to fill shifts, and pretending otherwise empties registers. Assume volume lands at half of projection and carry the scenario all the way through, because that is where the design flaw shows. At 40 daily orders instead of 80, a 12 USD ticket and 25% commission, you bill roughly 14,400 USD a month and the aggregator keeps 3,600; labor cost, inside the 25% to 35% range published by the U.S.

The counterfactual worth running before you sign

Bureau of Labor Statistics, eats another 4,300; that leaves 6,500 for ingredients, rent, packaging and utilities. With rent at 1,200 USD and food cost at 32%, you close the month about 400 dollars in the red and with no cash for advertising. In a shared kitchen that same scenario closes flat and you walk out in thirty days. The difference is not the money invested, it is REVERSIBILITY, which is why the right order is to validate demand first and sign a lease afterward, never the other way around. Delivery concentration is not trivia, it is the reason your negotiating power with an aggregator is zero. Earnest Analytics measured DoorDash closing 2024 with 60.7% of the U.S. market, Business of Apps reports Zomato and Swiggy together above 95% of online delivery in India, and Mordor Intelligence estimates Meituan and Ele.me past 90% of orders in China.

Where the market sits and what that tells your decision?

When two platforms control nine of every ten orders, commission is not negotiated, it is accepted.

On the format side, Grand View Research calculates that Asia-Pacific held 48.0% of cloud kitchen revenue in 2025, while Global Growth Insights puts Europe at 18.79% of the global dark kitchen market in 2024. Read that as an operating warning: the model scales where urban density and delivery cost allow it, and your neighborhood may not qualify. If your own dark kitchen already bills above break-even and more than 35% of your orders arrive through owned channels, stay where you are and change nothing. Moving to a shared kitchen or a host model would hand back flexibility you no longer need and strip away the one thing that actually compounds: the stable address holding up your Google Business Profile listing, the review history above 4.6, and the WhatsApp number your customers already saved.

When NOT to switch?

Do not switch either if your lease expires in under twelve months, because moving costs, fresh permits and the restart of local ranking will swallow any commission savings.

Staying is the right call when the problem lives in the menu or the ad spend, not in the structure. Audit food cost and listing conversion first, before you move a single hood. The real gap is not money, it is REVERSIBILITY. You walk away from a shared kitchen in thirty days and lose a deposit; an owned dark kitchen on a five-year lease chains you to a polygon that may never carry the order density the aggregator heat map promised. Owning the kitchen means owning the Maps listing, the domain and the WhatsApp number, and that matters more than it looks: an order arriving through your own channel keeps 22 to 28 percentage points more margin than the same order routed through a 25% commission.

Where the comparison actually breaks?

Learning curves run asymmetric here. Running a cloud kitchen with four simultaneous virtual brands demands timed dispatch that a new team simply lacks;

inside a shared kitchen you learn that rhythm on somebody else's fixed cost, and the lesson costs the same. Aggregators do not reward quality, they reward measurable consistency: declared prep time against real prep time, cancellation under 2%, menu availability. A small, tightly run virtual brand beats a large messy kitchen for placement, and that is the one field where the little operator plays with an edge. There is a point where the alternative stops serving you: once you sustain 70 daily orders, the shared kitchen becomes the bottleneck and every extra shift prices badly. That is when the owned dark kitchen earns its keep, backed by proven demand rather than a spreadsheet projection.