Guides
How to study ghost kitchen operations: 25 signals by operating model
A method for reading ghost kitchen operations from the outside: 25 signals grouped by operating model, how to run the study, and what transfers to your own radius.
What to take away
- Ghost kitchens to watch in 2027 include CloudKitchens, Reef Technology, Virtual Dining Concepts, Franklin Junction, Kitchen United, C3, Rebel Foods, and Deliveroo Editions.
- CloudKitchens, founded by Travis Kalanick, rents kitchen bays by the month in Los Angeles, Chicago, and London.
- Reef Technology puts delivery kitchens in parking lots and shipping containers, with locations in Miami and Houston.
- Virtual Dining Concepts runs MrBeast Burger and Guy Fieri's Flavortown Kitchen out of other restaurants' kitchens.
- Franklin Junction licenses Nathan's Famous and Fazoli's to host kitchens.
- Kitchen United closed several delivery food halls in 2023 and shifted to a software model.
- The named operators are examples. What is worth watching is the twenty-five signals below, which show how any ghost kitchen operation, anywhere, can be read from the outside.
- The signals are grouped by operating model, because the models fail differently.
- Watch for what transfers to your radius and your commission tier, and discard the rest.
- The most useful operators to study are the ones you can order from and rate.
Operators are told to watch the leaders. In delivery-only food, leaders are often private, often reorganizing, and often described in secondhand accounts that cannot be checked. What you can check is the operation from the outside: order from it, read its listings, see where it cooks.
The twenty-five signals below are that method; they are the substance behind the framework in the expansion and market guide.
Signals by operating model, with named operators
Ghost Kitchen Operating Models
Facility operator
- Core signal
- Peak-slot availability
- Compliance split
- License approvals
- Rating risk
- Driver handoff design
- Tenant fit
- Listing turnover
Virtual-brand licensor
- Core signal
- Prep overlap
- Compliance split
- Commission terms
- Rating risk
- Rating spread
- Tenant fit
- Packaging per dish
Restaurant idle capacity
- Core signal
- Live vs peak hours
- Compliance split
- Refund reasons
- Rating risk
- Refund reasons
- Tenant fit
- Photographs in containers
Delivery-first multi-brand
- Core signal
- Brands per address
- Compliance split
- Attention limits
- Rating risk
- Brand count
- Tenant fit
- Shared ingredients
Signals by operating model
| Model | Signal | What it tells you |
|---|---|---|
| Facility operator (rents bays or hours) | 1. Peak-slot availability | Whether the building is full at the hours that matter |
| 2. Driver handoff design at the door | How much of the tenants' rating risk the building absorbs | |
| 3. What the license says its approvals cover | The split between facility and operator compliance | |
| 4. Turnover of tenant listings on the platforms | Whether the model works for tenants or only for the landlord | |
| Virtual-brand licensor (sells a brand to kitchens) | 5. Prep overlap with a host kitchen's menu | Whether the brand is a menu or a second restaurant |
| 6. Commission and fee terms in the agreement | Where the licensor's revenue comes from | |
| 7. Rating spread across host kitchens | Whether the brand travels or depends on one kitchen | |
| 8. Packaging specified per dish | Whether the transit test was done by the licensor or left to hosts | |
| Restaurant running delivery brands from idle capacity | 9. Hours the brands are live versus the restaurant's own peak | Whether the brands fill idle hours or compete with the dining room |
| 10. Listing photographs shot in containers | Whether the brand was designed for the bag | |
| 11. Refund reasons visible in ratings | Where the handoff breaks | |
| Delivery-first multi-brand operator | 12. Number of brands per address on the platforms | Tablet load at peak |
| 13. Shared ingredients visible across the brands' menus | Prep discipline | |
| 14. Radius quoted per brand | Whether transit tolerance is designed per dish | |
| 15. Own-channel presence and its price relative to the platforms | Commission strategy | |
| Platform-affiliated or platform-run kitchens | 16. Sort position of affiliated listings | A question to raise with the platform's merchant documentation, not to assume |
| 17. Terms offered to independent operators in the same building | Whether the affiliation changes the economics | |
| Office and event specialists | 18. Quote structure, if published | Whether staging and delivery are separate lines |
| 19. Weekday-only hours | A radius chosen for office density | |
| Co-packing and shared production | 20. Batch scheduling against a client's peak | Whether production displaces the client's orders |
| 21. Labeling and allergen documentation | Compliance maturity | |
| Any model, anywhere | 22. Rating distribution, not the average | Where a listing is about to trip a threshold |
| 23. Menu length at peak | Whether the operator ranks dishes by constrained minutes | |
| 24. How fast an unavailable item disappears across platforms | Whether menu management is solved | |
| 25. What the bag looks like when it arrives | Everything |
Public examples for each model show how the signals play out. CloudKitchens rents kitchen bays by the month in Los Angeles, Chicago, and London. Reef Technology places delivery kitchens in parking lots and shipping containers in Miami and Houston. Deliveroo Editions provides kitchen space in London and other cities.
Virtual Dining Concepts runs MrBeast Burger and Guy Fieri's Flavortown Kitchen from host kitchens. Franklin Junction licenses Nathan's Famous and Fazoli's to restaurants with spare capacity.
C3, founded by Sam Nazarian, operates Umami Burger and acquired Nextbite in 2022. Rebel Foods, based in India, runs Faasos, Behrouz Biryani, and Ovenstory Pizza from more than 400 kitchens. DoorDash Kitchens operates shared kitchen spaces in Redwood City and Brooklyn.
These are public examples, not endorsements. Each one can be studied with the signals above.
How to run the study
How to run the study
- Order at peak, note handoff and bag
- Read all listings side by side
- Watch the street at peak
- Read low-rating reasons
- Write which signals transfer
- Order from the operation at its peak, on the platform it seems to favor, and note handoff time, bag condition, seal, insert, and what the refund process looks like if anything is wrong.
- Read every one of its listings side by sidebrands per address, shared ingredients, radius, photographs, hours.
- Look at where it cooks, from the street, at peakdriver traffic, the door, the handoff point.
- Read the ratings' distribution and the reasons in the low ones.
- Write down which signals transfer to your radius and tier, and which depend on the operator's facility, capital, or platform relationship that you do not have.
What does not transfer
Scale on one platform in one city says nothing about your radius. A facility operator's economics depend on real estate you do not own. A platform's affiliated kitchens may be treated under terms you are not offered. The one thing that always transfers is signal 25, because a bag that arrives as designed is the whole business, everywhere.
The Small Business Administration's Business Guide frames growth as a set of general decisions about market, funding, and management. Reading other operators is research for those decisions, not a template.
The people and the rules behind every signal
Every operation above answers to the health authority for its address, found through the FDA's directory of state retail food codes and agencies, and an operator abroad answers to its own.
The licensing and compliance guide explains why a signal such as labeling maturity is a compliance signal. And every one of them is run at peak by someone in the role the Bureau of Labor Statistics describes in its profile of food service managers: staff, supplies, sanitation, customer issues, budgets, on nights and weekends.
Signal 12, brands per address, is a question about that person's attention, which the hiring and training guide treats as the constraint that fails first.
Where the study feeds decisions
Signal 15 feeds the commission strategy in the pricing and profit guide. Signals 10, 22, and 25 feed the storefront work in the marketing and growth guide. Signals 1 to 4 feed the facility decision. None of them feeds a decision by themselves; they feed the gates.
Common questions
Why not just name the leading operators?
We name public operators as examples, but a ranking of leaders would be stale within a season and we cannot verify private performance. The signals stay true.
Can I study operators in other countries?
Yes, by the same method where you can order or observe, and the platform and regulatory differences are part of what you learn. Signals 3 and 21 are where those differences show.
Which model should a new operator copy?
None of them wholesale. Copy the signal-25 discipline from whichever operator's bag arrives best, and build the rest from your own radius and tier.
How often should I run the study?
Once before launch, and again before any growth move that resembles a model above.
