dark-kitchen-test

---
name: dark-kitchen-test
description: Turns your existing kitchen's spare capacity, a standalone unit's real quote, your commission-maths contribution figure and your busiest hour's data into a written answer on whether a delivery-only brand is worth building: the marginal cost of running it from your own kitchen against the full cost of a rented unit, the breakeven order volume either route needs, the registration and planning position for each route, and a dated decision. Use before you sign a lease, buy a second set of pans, or launch a virtual brand nobody has costed.
---

# A dark kitchen decision made on your own numbers, before you sign anything

You give this the spare capacity in your current kitchen hour by hour, a real quote for any standalone unit you are considering, your contribution per order from commission-maths, and your busiest hour's order data from delivery-hours. You get back the marginal cost of running a second brand from your existing kitchen, the full cost of a standalone unit, the order volume either route needs to break even, the registration and planning position for whichever premises you use, and a one page decision dated and ready to act on or shelve. It will not tell you whether a delivery-only brand succeeds. No published figure exists for a single independent UK venue on setup cost, breakeven volume or success rate, and an owner who commits against a borrowed number has committed against a number nobody measured in their own kitchen.

## What it does

1. **Name the route you are actually testing, because "a dark kitchen" is not one decision, it is three.** The first route runs a second menu out of your existing kitchen in hours it is already staffed and already paid for. The second route rents a standalone unit with its own lease, its own kit and its own staff. The third route is not doing it. Write down which route, or which routes, you are costing, because the two real routes have almost nothing in common financially and a single spreadsheet that blends them produces a number that answers neither question.

2. **Cost the existing-capacity route on the marginal cost alone, and nothing else.** The rent, the rates, the lease and the core staff you already carry are a cost whether or not the second brand launches, so none of it belongs in this figure. What belongs is what would not exist without the new brand: the extra ingredients, the extra packaging, the extra minutes of labour during hours that are already staffed but not already full, and the extra platform commission on the new orders. Cost it from real invoices and a real measured shift, the same discipline commission-maths applies to delivery labour generally. A figure with the existing rent folded back in overstates the true cost of the decision you are actually making.

3. **Cost the standalone-unit route in full, from a real quote and nothing invented.** Rent, business rates, kit, a deposit, fit out, utilities, staffing and any platform activation fee, each with a supplier or landlord's own figure and the date it was quoted. This route has no marginal cost, because everything about it is new. Where a figure has not been quoted, it is a gap, not an estimate, because a guessed rent on a spreadsheet is the single most common reason a dark kitchen model looks viable on paper and does not survive its first quarter.

4. **Settle the registration position for whichever premises the food is actually made in.** GOV.UK is direct on the trigger: "When you start a new food business, or take over an existing business, you must register with your local authority", "You are required to register at least 28 days before trading", and "Registration of your food business is free and can't be refused." A second brand cooked entirely in your existing, already registered kitchen does not usually need a fresh registration, but the local authority should still be told the business is trading under an additional name and selling an additional menu from the same site, because the guidance is equally clear that "If you operate in more than one location, you need to register each premise with the local authority in which they are located", which is the rule that catches the standalone unit outright. A new site always needs its own registration, twenty eight days ahead, and it is free.

5. **Get the planning position checked in writing before you sign anything, because the rules genuinely do not give a clean answer for every case.** The Town and Country Planning (Use Classes) Order 1987, as amended by the Town and Country Planning (Use Classes) (Amendment) (England) Regulations 2020, places the sale of food and drink "principally to visiting members of the public where consumption of that food and drink is mostly undertaken on the premises" in Class E, and separately removes from every use class "a hot food takeaway for the sale of hot food where consumption of that food is mostly undertaken off the premises", which sits outside all classes as a use of its own, called sui generis. Both of those definitions assume a member of the public visits the premises to buy the food. A kitchen that sells only to a rider and never to a visiting customer sits awkwardly against both tests, and local planning authorities do not apply the same answer to it: a council's own supplementary guidance, such as Kirklees Council's Hot Food Takeaway Supplementary Planning Document, tests a premises against factors including "the proportion of space designated for food preparation", "the number of tables and chairs" and "hours of opening", none of which was written with a seatless, counterless delivery kitchen in mind. Record the answer your own local planning authority gives in writing, and treat anything short of that as unconfirmed.

6. **Check the rateable value and the small business rate relief position for a standalone unit.** The current guidance states plainly: "You will not pay business rates on a property with a rateable value of £12,000 or less, if that's the only property your business uses", and "For properties with a rateable value of £12,001 to £15,000, the rate of relief will go down gradually from 100% to 0%", figures that apply for the rating year running from 1 April 2026 to 31 March 2027. Record the unit's rateable value as stated by the landlord or the Valuation Office Agency, and whether it is the only property the business uses, because a second property can remove relief that a single premises would have kept.

7. **Work out the breakeven order volume for whichever route you costed, using the contribution per order already established in commission-maths.** Divide the marginal cost from step 2, or the full cost from step 3, by the contribution per order for the new brand's menu, and the result is the number of orders a week the route needs before it is worth the effort at all. Do this before naming a single dish on the new menu, because a route that needs four hundred orders a week to break even is a different conversation from one that needs forty.

8. **Check the new brand's hours against your own kitchen's busiest hour, using the delivery-hours file rather than a guess.** A second menu fired from the same pass, the same fryer and the same small team during the hour your existing brand is already at its limit does not add capacity, it competes for it, and the result is two brands both running late. Record the hours the new brand would actually fire tickets and compare them against the busiest hours already measured, and treat any overlap with a genuinely full kitchen as a reason to either move the new brand to quieter hours or count it against the standalone route instead.

9. **Hand the assembled figures to Codex to build the comparison.** Every judgement is already made: which routes to test, the real quotes, the registration and planning answers and the hours comparison are the owner's. Run:

   ```
   codex exec --sandbox danger-full-access --skip-git-repo-check -C "<working folder>" - < CODEX.md
   ```

   Claude hands over the marginal cost items for the existing-capacity route, the full quoted costs for the standalone route, the contribution per order, the busiest hour data, the registration and planning answers with their dates, and the rateable value if applicable. Codex returns the cost comparison table, the breakeven table, the capacity conflict check and a gaps file. Codex invents no cost and reaches no conclusion on the planning question.

10. **Write the decision, and date every figure behind it.** One page: which route, if any, is being tested, the breakeven volume it needs, the capacity hours it will use, and what happens if it does not reach breakeven by a stated date. Put the date every quote, invoice and written planning answer was obtained in the footer. A dark kitchen model that nobody revisits is a lease that keeps renewing itself on the strength of a spreadsheet built before the first order arrived.

## Then it checks

1. The file names which of the three routes, existing capacity, standalone unit, or do nothing, is being costed, and does not blend the marginal cost of one with the full cost of the other.
2. Every cost in the existing-capacity route is genuinely marginal, meaning it would not be spent if the new brand did not exist, and no share of the current rent, rates or core staff wage appears in it.
3. Every cost in the standalone-unit route carries a named supplier or landlord and a quoted date, and no figure is described as typical, average or approximate.
4. The file states the registration position with its date, and whether the local authority has been told about an additional trading name at an existing site or has issued a fresh registration for a new one.
5. The file states the planning position as either confirmed in writing by the local planning authority, with the date of that confirmation, or as unconfirmed, and no planning conclusion is presented as settled without that written confirmation.
6. The breakeven order volume is calculated from the owner's own contribution per order figure, and the hours the new brand would trade are checked against the existing kitchen's own busiest hour data with any overlap named.

Any check fails: name it, redo that step once. Failed twice: say what is wrong and stop.

## Rules
- Public information only.
- Never invent a fact, a number or a quote.
- Anything sent in someone's name says whose name it is.
- Never state a typical dark kitchen setup cost, a typical breakeven order volume, or a typical success or failure rate for a delivery-only brand. No published figure exists for a single independent UK venue, and a target copied from an industry article is a target built for somebody else's rent and somebody else's kitchen.
- Never fold an existing fixed cost into the marginal cost of the existing-capacity route. That cost exists whether the new brand launches or not, and including it makes a genuinely low cost decision look expensive enough to reject for the wrong reason.
- Never present the planning use class question as settled from the legislation alone. The Use Classes Order was not written with a seatless delivery kitchen in mind, local planning authorities differ in how they apply it, and only a written answer from the actual local planning authority settles it for a specific site.
- Never launch a new brand into hours already shown to be at capacity without recording the conflict. A busy kitchen does not have spare capacity simply because nobody has asked it to prove otherwise.
- This output is a working document prepared for the owner to check against their own quotes, invoices and correspondence, and for their accountant on the cost comparison, their local planning authority on use class, and their local authority's environmental health team on registration. It applies published rules and the owner's own figures to a draft decision. It is not legal advice, planning advice, tax advice or investment advice, and it is not a ruling that either route is viable or lawful.

## Built from
- Food Standards Agency and GOV.UK, "Starting a food business", https://www.gov.uk/guidance/starting-a-food-business, updated 19 August 2026, read 23 September 2026: the registration trigger, the 28 day and free and cannot be refused wording, and the per premises registration rule in step 4.
- The Town and Country Planning (Use Classes) (Amendment) (England) Regulations 2020, SI 2020/757, https://www.legislation.gov.uk/uksi/2020/757/made, read 23 September 2026: the Class E definition and the sui generis exclusion for hot food takeaway with off premises consumption in step 5.
- Kirklees Council, "Hot Food Takeaway Supplementary Planning Document", https://consult.kirklees.gov.uk/portal/pp/spds/hft_spd/hft_spd, no publication date shown on the page, read 23 September 2026: an example of the factors a local planning authority tests a premises against, used in step 5 to show the test was not written for a seatless delivery kitchen.
- GOV.UK, "Business rates relief: Small business rate relief", https://www.gov.uk/business-rates-relief/small-business-rate-relief, rates stated for 1 April 2026 to 31 March 2027, read 23 September 2026: the rateable value thresholds in step 6.

# dark-kitchen-test

## You are given
The route or routes being tested for one UK hospitality business: existing capacity, standalone unit, or both. For the existing-capacity route, the list of marginal cost items only, each with a supplier, an invoice or measured figure, and a date, and confirmation that no item is a share of an existing fixed cost. For the standalone-unit route, the full list of quoted costs including rent, business rates, kit, deposit, fit out, utilities and staffing, each with a named supplier or landlord, a quoted figure and a date. The contribution per order for the new brand's menu, taken from the owner's own commission-maths file. The busiest hour data for the existing kitchen, taken from the owner's own delivery-hours file, and the hours the new brand is proposed to trade. The owner's registration position, stating whether the local authority has been told about an additional trading name at an existing site or has issued a fresh registration for a new site, with the date. The owner's planning position, stating whether the local planning authority has confirmed the use class in writing, the date of that confirmation, and the confirmation reference if one exists. The rateable value of any standalone unit as stated by the landlord or the Valuation Office Agency, and whether it is the only property the business uses. All judgement calls, including which route to test, whether to proceed, and the planning and registration answers, have already been made by the owner and are supplied to you.

## Produce
Write into a `./dark-kitchen-test-output/` folder:

1. `existing-capacity-costs.csv` with these columns in this order: `cost_ref`, `item_verbatim`, `category`, `supplier_or_source`, `amount_gbp`, `basis`, `date_recorded`, `genuinely_marginal`. `cost_ref` is M001 upward. `category` is exactly one of `ingredients`, `packaging`, `labour`, `platform commission`, `other`. `basis` is exactly one of `per order`, `per shift`, `per week`, `one off`. `genuinely_marginal` is `yes` or `no`. Any row where `genuinely_marginal` is `no` is listed in `gaps.md` and excluded from the total.
2. `standalone-unit-costs.csv` with columns: `cost_ref`, `item_verbatim`, `category`, `supplier_or_landlord`, `amount_gbp`, `basis`, `quote_date`. `cost_ref` is S001 upward. `category` is exactly one of `rent`, `business rates`, `kit`, `deposit`, `fit out`, `utilities`, `staffing`, `platform activation`, `other`. `basis` is exactly one of `one off`, `per month`, `per year`. Every row with no `quote_date` is listed in `gaps.md` and excluded from the total.
3. `registration-position.md` recording, on separate lines and nothing else: the route being registered; whether it uses an existing registered premises or a new one; for an existing premises, the date the local authority was told of the additional trading name; for a new premises, the registration date and confirmation it was made at least 28 days before trading; and the source of each statement.
4. `planning-position.md` recording, on separate lines and nothing else: the route requiring a planning answer; whether the local planning authority has confirmed the use class in writing; the date and reference of that confirmation if one exists; and, where no written confirmation exists, the single line `unconfirmed, no written answer from the local planning authority`.
5. `rates-position.md` recording, on separate lines and nothing else, for a standalone unit only: the rateable value as stated; whether it is £12,000 or less, between £12,001 and £15,000, or above £15,000; whether the property is the only one the business uses; and the resulting relief conclusion stated as `full relief`, `tapered relief`, `no relief on value` or `no relief, second property`.
6. `breakeven.csv` with columns: `route`, `total_cost_gbp`, `cost_basis`, `contribution_per_order_gbp`, `breakeven_orders`, `breakeven_period`. One row per route tested. `cost_basis` names which of `existing-capacity-costs.csv` or `standalone-unit-costs.csv` the total was drawn from. `breakeven_orders` is `total_cost_gbp` divided by `contribution_per_order_gbp`, left empty and recorded in `gaps.md` where either figure is missing.
7. `capacity-conflict-check.csv` with columns: `proposed_hour_start`, `proposed_hour_end`, `day_of_week`, `existing_kitchen_status_this_hour`, `conflict`. `existing_kitchen_status_this_hour` is copied from the owner's supplied delivery-hours data. `conflict` is `yes`, `no` or `not stated`. Every row where `conflict` is `yes` is listed in `gaps.md`.
8. `gaps.md` a numbered list of: every existing-capacity cost item not confirmed genuinely marginal; every standalone-unit cost item with no quote date; every route with no contribution per order supplied; every proposed hour with no existing kitchen status supplied or with a confirmed conflict; the registration position where the local authority has not been told or a new registration has not been dated; the planning position where no written confirmation exists; and any figure the owner supplied without a date.

## Rules
- Codex totals, compares, divides and records. It never decides which route to pursue, never estimates a missing quote, never confirms a planning use class, and never decides whether a cost is genuinely marginal. Those were supplied.
- Never estimate a missing cost, rent figure, rateable value or contribution per order. A route with an incomplete cost list reports its total as incomplete rather than as a figure, and every missing item is a line in `gaps.md`.
- Never write a typical setup cost, a typical breakeven order volume or a typical success rate into any file. No such figure exists for a single independent UK venue and none is to be recorded.
- Never treat an unwritten or verbal planning answer as confirmed. Only a dated written communication from the local planning authority supports a confirmed conclusion in `planning-position.md`.
- Never fold a fixed cost the business already carries into the existing-capacity route's total. Any item not marked `genuinely_marginal` as `yes` is excluded from that total and recorded in `gaps.md` instead.
- Never carry a cost, a quote or a planning answer forward from an earlier version of this file. Use only what is supplied in this run.
- Use British English, GBP written as `GBP` or the pound sign, and DD Month YYYY dates. No em dashes in any file you write, and any supplied text containing one is recorded verbatim and flagged in `gaps.md`.
- Every item name, supplier name and route name is copied exactly as supplied, including its punctuation and capitalisation.
- Every file ends with this line: this is a working document prepared for the owner to check against their own quotes, invoices and correspondence, and for their accountant, their local planning authority and their local authority's environmental health team to check before anything is signed. It applies supplied figures to a draft decision and is not legal advice, planning advice, tax advice or investment advice.

## Return
The absolute path of every file written, the row count of each CSV, the total marginal cost for the existing-capacity route or the word incomplete where an item is unconfirmed, the total cost for the standalone-unit route or the word incomplete where a quote date is missing, the breakeven order volume for each route tested, the registration position stated in full, the planning position stated in full including whether it is confirmed or unconfirmed, the rates relief conclusion for any standalone unit, the count of proposed hours in conflict with the existing kitchen's busiest hours, and the `gaps.md` item count.