Five Nights to a Tighter Restaurant P&L: Food Cost, Labour, Rent, the Platform's Cut and the Weekly Read

Ask a struggling operator for last month’s food cost and you usually get a pause, then a guess. Ask for rent as a share of revenue and you usually get silence. That is not carelessness. It is the natural state of a business that reads its accounts once a month, three weeks after the month has ended, in a pack built for the accountant rather than the owner — by which point every line that drifted has run for six weeks or more, and the cause has gone cold.

A turnaround starts differently. It does not open the whole P&L at once. It reads one line at a time, in a deliberate order, against a published ceiling, and asks the same question of each: what did this line actually consume, against what actually came in? That order — food cost, labour, rent, the platform’s cut, and then the weekly discipline that keeps all four closed — is the method behind every diagnostic we run across the UAE and wider GCC, India and Singapore. It is also the spine of Five Nights to a Tighter P&L, the free five-night email course you can join from the toolkit: one leak per night, then the series ends.

This page is the whole method in one place, read in the order a turnaround reads it: the line that moves most for the least effort first, the line you can no longer change last, and each night carrying its own arithmetic so you can run it on your own numbers tonight. The bands throughout are the published ceilings behind the Restaurant Operating Index — food at or under 32% of revenue, labour at or under 30%, prime cost (the two together) between 55% and 62%, rent between 6% and 12%, and delivery commission between 3% and 6% of total revenue. They are diagnostic working rules, not promises, and every concept has its own texture around them. Where an example uses figures, they are illustrative round numbers chosen so the arithmetic is easy to check.

One illustrative restaurant runs through all five nights. Say it takes AED 300,000 a month, net of VAT, and every line sits inside its band:

Line	Share of revenue	AED
Revenue	100%	300,000
Food cost	30%	90,000
Labour (fully loaded)	28%	84,000
Prime cost	58%	174,000
Rent	10%	30,000
Delivery commission	5%	15,000
Four lines together	73%	219,000
Left for everything else	27%	81,000

Check it: 90,000 + 84,000 = 174,000, which is 58% of 300,000; add 30,000 and 15,000 and the four lines take 219,000, or 73%; what remains is 81,000, or 27%. That remainder carries utilities, marketing, repairs, licences, insurance, finance, depreciation and the owner’s return. Healthy does not mean a fat margin on any single line. It means four lines each held inside their range so the remainder is real — and the five nights are how each of them is held.

Night one — Food cost

Food cost is the line a turnaround touches first, because it moves the most for the least effort: no lease to renegotiate, no roster to rebuild, just discipline applied to what the kitchen buys, prepares and plates.

The read. Food cost is not the supplier total and it is not a feeling. It is consumption against sales: opening stock, plus purchases, minus closing stock, divided by food revenue for the same period. That is what actually left your shelves as a share of what actually came in — including everything no recipe accounts for: the spoiled delivery, the over-trimmed fillet, the staff meals, the plate that walked. If you take one habit from this night, take the denominator seriously. A percentage computed on a guessed revenue figure or a skipped stock count is fiction, and every decision built on it inherits the fiction.

The ceiling — and what it does not say. GGB reads food cost against a published ceiling of 32% of revenue, and reads it beside labour, because the two together form prime cost with its own ceiling of 62%. The band is a smoke alarm, not a target to sit at. Cross it and something upstream is leaking. Sit comfortably under it with a wide gap between what the costed recipes say the food should have cost and what the stock movement says you actually spent, and you are still paying for food you never sold. On the illustrative month, the costed recipes at the mix you sold might say 28% — AED 84,000 — while the count says 30%, or 90,000. That AED 6,000 gap is the most informative number in the kitchen, and it has exactly four addresses.

Yield: the invoice is not the cost. The kilo you buy is not the kilo you serve. Between the two sit trim, peel, bone, skin and shrink, and if the recipe card was costed on the raw invoice weight, every plate is quietly more expensive than the card claims. The fix is a yield test: weigh the item as delivered, prep it as you actually prep it, weigh what is usable, and cost the recipe on the yielded weight. Run it on your five most expensive ingredients first — that is where the gap is priced highest.

Portioning: drift you cannot see on one plate. No cook decides to over-portion. Portions drift — a heavier hand on the ladle, a garnish that grows, a “make it nice” for a regular — until the kitchen is serving a dish that no longer matches its costing. Per plate it is invisible; across a service it is real money; across a month it can be the whole variance. Portion tools on the line — scoops, scales, marked ladles, a photographed spec per dish — are unglamorous and they work. The spec is the contract between the menu price and the plate.

Purchasing and waste: discipline beats haggling. The dramatic version of purchasing is renegotiating suppliers. The profitable version is duller: order against par levels instead of habit, receive against the invoice with a scale in reach, check invoice prices against agreed prices, and re-cost the menu when they move. Supplier price creep is a slow leak by design — a dirham here, a substitution there — and it only shows if someone compares this quarter’s invoices with the prices the recipes were costed at. The fourth address is the bin: spoilage, over-production and the special nobody costed. A waste log kept for two weeks — what went out, why, roughly what it cost — usually names the cause faster than any report.

What the method can move is a matter of record in the one engagement we publish by name. Parco Group’s Jebel Ali operation began its reset with the line twelve points past the ceiling — 44% — and finished a 120-day programme of purchasing, portioning, menu pricing and waste control at 29%, with average daily sales rising from AED 6,000 to AED 14,000 over nine months. The consented figures are on record; they are documented, not a promise of your outcome. The point of citing them is narrow: a line in the forties is not a life sentence. It is an unread line. The full theoretical-versus-actual method is in food cost control: closing the gap; the pricing side — which dishes deserve their place on the card — belongs to menu engineering, and the Menu Engineering Matrix plots the card itself, margin against popularity, dish by dish.

Night two — Labour

The second night belongs to labour, and it opens with an uncomfortable reframe: most labour problems are not pay problems. Pay is what the line shows; the schedule is what drives it. A roster built from habit will hold its shape for years after the demand that justified it has moved, and the P&L pays the difference every week.

The read. Labour cost is the fully loaded number — wages plus everything that rides with them — divided by revenue for the same period. Loaded is the operative word. In the GCC the gap between a salary and what that person actually costs the business is wide: visas, medical, insurance, accommodation, transport and the end-of-service entitlement accruing quietly across every contract. Count the payslip alone and the labour line you are managing to is fiction. The components are walked line by line in restaurant staffing cost per head; use that loaded figure, or the percentage flatters you.

The ceiling. GGB reads labour against a published ceiling of 30% of revenue, inside the 25–30% band, and always beside food cost, because the pairing is the point. A scratch kitchen legitimately runs food lean and labour a little rich; a prep-light concept runs the reverse. Either can look healthy alone while the sum sinks the model, which is why prime cost — the two together, banded 55–62% — is the survival gauge and the components are where the work happens. On the illustrative month, labour at 28% is AED 84,000; let it drift two points to the ceiling and the line is 90,000 — AED 6,000 a month that nobody decided to spend.

The roster is a habit. Here is how most rosters were actually built: someone constructed one in the opening month, it survived contact with reality, and it has been photocopied — with small mutations — ever since. New menu, new delivery mix, a mall footfall pattern that shifted two years ago: the revenue curve moved and the roster did not. Nobody decided this. That is precisely the problem — the schedule is the largest controllable cost in the business that nobody is currently deciding.

Draw the curve. The diagnostic costs nothing. Pull one ordinary week of covers — or orders, or revenue — by hour, from the POS. Draw it as a curve. Lay the roster over it. Two findings appear in almost every operation that runs this honestly:

• Hours stacked where the rush used to be — a heavy mid-shift from an era when lunch was the story, while the real peak has migrated to evenings or to delivery windows.
• The peak running thin — the half-hour either side of the true rush understaffed, which is where slow tickets, dropped tables and lost repeat visits are manufactured, while quiet afternoons carry people with nothing to sell.

The mismatch between the two lines is the leak, drawn in ink.

Productivity, not headcount. The lens that keeps this honest is revenue per labour hour: the period’s revenue divided by the labour hours scheduled to earn it. On the illustrative month, AED 300,000 across 2,500 scheduled hours is AED 120 per hour; the same revenue across 2,200 better-placed hours is about AED 136. Headcount says how many people you employ; revenue per labour hour says what each scheduled hour is doing for the business. It is the difference between cutting and shaping. A blind cut lowers hours and revenue together and calls it discipline. Shaping moves hours from the flat parts of the curve to the steep parts — staggered start times, prep pushed into quiet windows, sections opened and closed with demand — and the same guests get served by a calmer line.

What to move first. Start where the curve says, not where the argument is easiest: the two or three widest gaps between scheduled hours and demand. Stagger arrivals in fifteen- or thirty-minute steps instead of shift-block starts. Put a name against the schedule each week — a roster nobody owns reverts to habit within a month. And read the two numbers weekly, labour percentage and revenue per labour hour, so the shape holds. The Labour Productivity instrument runs this read — your labour line against the 30% ceiling and the productivity lens beside it — free, on-device, in a few minutes; how many staff a restaurant needs builds the roster forward from forecast covers.

Night three — Rent

The third night’s read is the strangest cost on the P&L: the one you agreed to before the first customer ever walked in. Food cost responds to management within days. Labour responds within a roster cycle. Rent does not respond at all — it arrives on the same day, at the same size, however service went. Which is exactly why it has to be read differently.

The read: one division. Occupancy is read as a ratio: annual rent divided by realistic annual revenue. Annual on both sides — restaurants are seasonal, and a ratio computed on your best month is a comfortable lie. Include what genuinely rides with the lease — service charges, chilled-water and similar recurring occupancy charges — because the business pays the whole line, not the headline. Then write the percentage down. In diagnostic work this is the number owners most often cannot produce: operators who know their food cost to the decimal and have never once divided rent by revenue. It takes a minute, and it reframes the renewal, the refit and the second site.

The band, and the arithmetic that runs backwards from it. The published band is 6–12% of revenue. Because rent is fixed and revenue is not, the useful way to read the band is in reverse — from the lease to the revenue it demands:

Monthly rent	Revenue needed at the 12% ceiling	Revenue needed at the 6% floor
AED 20,000	AED 166,667	AED 333,333
AED 30,000	AED 250,000	AED 500,000
AED 45,000	AED 375,000	AED 750,000

The illustrative restaurant pays AED 30,000 and takes 300,000, so it sits at 10% — inside the band. If a road closure or a mall re-tenanting took revenue to 200,000, the same lease would read 15%, three points past the ceiling, and nobody would have renegotiated anything. That is the defining behaviour of the line: the ratio moves when revenue moves. A lease that read comfortably at projection turns heavy the moment the denominator changes, which is why the occupancy read belongs in the weekly rhythm beside food and labour, on realistic current revenue rather than the revenue the business plan promised.

Why the ratio beats the market rate. Per-square-foot is how leases are marketed; it is not how they are survived. The market rate compares your lease with other leases. The ratio compares your lease with your business — and the business is what pays. Two restaurants can pay identical rent on identical terms while one is comfortable and the other is drowning, because the denominators differ. This is also why “the district commands these rents” is never an answer to an occupancy problem: the street does not pay your rent; your revenue does.

The three honest ways out. When the ratio runs heavy there are exactly three honest exits, and none of them is “work harder on food cost”. Cutting a variable line to fund a fixed one does not fix the occupancy problem; it decides who funds the gap, and it is usually the owner.

• Grow revenue into the lease — with a plan, not a hope. Legitimate when the gap is modest and the plan is specific: named channels, menu work, hours, covers. “Sales will pick up” is not a plan; it is how a heavy ratio buys itself another expensive year.
• Renegotiate or right-size. Renewal windows are leverage; evidence is more. Arrive with the ratio, the revenue record and real alternatives. Sometimes the answer is not a lower rent but less space — subletting a floor, shedding a mezzanine, a format that earns from a smaller footprint.
• Exit. The hardest and sometimes the cheapest. A location the revenue cannot carry consumes cash indefinitely; an exit prices the loss once. The arithmetic is brutal but it is arithmetic — run it before the reserve, not after.

Lease terms and exit clauses are legal matters — take proper advice on your own contract. The operator’s job is to arrive at that conversation with the numbers already read. Everything above is cheaper as prevention: before any new lease or renewal, run the ratio on realistic projected revenue and on the pessimistic case, because the rent stays fixed in that scenario too. The traps that compound a heavy ratio are walked in restaurant lease and fit-out; the Break-Even Calculator shows the covers per day a given rent demands before you commit; and for the site you already run, the Rent vs Revenue Check reads your ratio against the band and shows the revenue your lease actually demands, in about a minute.

Night four — The platform’s cut

The fourth night follows the money out through the apps. Delivery is the cost line that grew fastest in most restaurants over the last five years, and the one most P&Ls still treat as a footnote — netted quietly out of revenue where nobody reads it. For a delivery-heavy operation the platform’s cut is the same order of money as food or labour, and it deserves the same discipline.

Off the top: how the cut works. When an order comes through an aggregator, the platform takes its share of the order value first — before food cost, before packaging, before anyone in your kitchen is paid. The per-order rate is commonly 15–30% of order value depending on contract and delivery model: whether the platform’s riders or yours carry the order, what marketing placement you take, category, exclusivity. When Khaleej Times surveyed Dubai operators in 2020, the quoted range was 25–30%, and higher with everything stacked on top. Treat any range as context. The number that decides your margin is the one in your own agreement — effective rate per order, add-ons included — and the aggregator commission tracker keeps the published take rates in one cited place. The mechanics matter more than the headline: the cut is a share of revenue while your costs live below it, so a few points of commission move the bottom line by far more than a few points.

Two percentages wearing the same name. The per-order rate is not the Index line. The Index bands delivery commission at 3–6% of total revenue, and the bridge between the two is how much of your business runs through the apps:

Delivery share of revenue	at 15% commission	at 25%	at 30%
10%	1.5%	2.5%	3.0%
20%	3.0%	5.0%	6.0%
30%	4.5%	7.5%	9.0%
40%	6.0%	10.0%	12.0%

The illustrative restaurant does a fifth of its business through the apps at 25% and sits at 5% of revenue — inside the band. Let delivery grow to 40% of revenue at the same rate and commission alone takes 10% of everything it sells, as much as it pays in rent. The dish never changed price; the channel changed everything around it. That is why margin has to be read by channel: a healthy dine-in room can subsidise a losing delivery screen for months without anyone noticing.

The stack the dine-in price never carried. A dish priced for the table carries food cost and its share of the room. The same dish, same price, through an app, has to absorb three costs the table never billed it: the commission, off the top of the order; packaging, a real cost on every single order that scales with volume rather than revenue; and promotion drag, the discount or delivery offer that won the order, cut from the same ticket. Walk one order. A dish listed at AED 100, sold through a 20% offer, is an AED 80 order. Commission at 25% takes 20. Packaging takes 4. The plate still costs the kitchen 30 — the food is paid for on the dish, not on the discount. Contribution: 80 − 20 − 4 − 30 = AED 26, against AED 70 for the same plate at the table before its share of the room. Nothing on the menu changed. Stack that across a month of busy-looking delivery volume and a dish that earns honestly at the table can hand money back through the app.

Contribution per order: the only honest read. Blended P&Ls hide delivery problems by construction — strong dine-in margin papers over weak delivery margin until the combined number sags and nobody can say why. The honest read is per order, by channel:

Order value − platform’s cut − packaging − promotion − food cost = contribution per order

Run it on real orders — your three best-selling delivery items first, actual tickets, not list prices. You are sorting the delivery menu into items that still leave a sensible contribution after the full stack and items being quietly subsidised by the rest of the card. Marketplace promotions belong in the same arithmetic: they win orders, and the discipline is to treat each one as a priced decision rather than a reflex — run with the post-stack margin in front of you, aimed at the orders you actually want more of, never as a permanent state of the listing.

Decide the mix deliberately. None of this argues for leaving the platforms. It argues for choosing. Price the delivery channel for its real cost stack — a delivery menu is legitimate engineering, not trickery. Move what repeat demand you can onto channels you own, where an order does not pay the full cut. Then decide, with the channel numbers in front of you, how much volume you want through a third party, instead of letting the apps decide by default. The wider economics — packaging spec, promotion design, direct channels — are walked in delivery aggregator economics; the Delivery Margin Recovery instrument runs the contribution read item by item and shows where the recovery is.

Night five — The weekly P&L

The final night is not about finding a new leak. It is about the discipline that keeps the first four closed — because every line above will drift again the moment nobody is reading it. Food cost, labour, rent, the platform’s cut: none of them stays fixed by being fixed once. They stay fixed by being read.

Month-end is an autopsy. The standard operating rhythm — wait for the accountant’s pack, read it three weeks into the next month, wince, carry on — reviews the patient after the outcome is decided. A food-cost drift that begins in the first week of March has run six weeks or more by the time the March statement is discussed. The cause is cold: the supplier price moved, the portion crept, the promotion ran long, and nobody can now say which. Month-end is for reconciliation. Control happens weekly, or it does not happen. On the illustrative month, a two-point drift on the food line is AED 6,000 a month — AED 1,500 a week. A weekly reader meets it in week two; a month-end reader meets it four to six weeks later, with the habit embedded and the cost multiplied.

The five lines. The weekly read is deliberately small — five lines on one page:

1. Sales for the week, against the same week last month and, once you have it, last year.
2. Food cost as a share of sales — purchases adjusted by the weekly count, never purchases alone.
3. Labour as a share of sales — the fully loaded figure, not just wages, with overtime flagged separately.
4. Occupancy — the week’s share of rent against the week’s sales, so a heavy lease is never invisible between renewals.
5. The platform’s cut — commission, packaging and promotions on delivery, read against delivery sales, not blended away.

Anything more belongs in the monthly pack. The weekly page is small so that it actually gets read.

Read against the ceilings. Numbers without reference points are weather. Each line is read against the published bands — food at or under 32%, labour at or under 30%, prime cost at or under 62% — with occupancy and delivery read against your own ratio and contract, the way nights three and four set out. What the bands give the weekly read is a verdict: within the band or above it, by how much, and in which direction it is moving. Three weeks of the same line drifting the same way is not noise. It is a cause with an address, and the profitability audit method is how it gets chased down.

One action, not five. Here is the part most weekly routines get wrong: they end in a list. Five observations, five intentions, and by Thursday the operation has absorbed none of them. The discipline is to end the twenty minutes with one action — chosen because it addresses the widest gap against the bands — with a name on it and a check the following week. One action a week is fifty-two real corrections a year, which is more than most turnarounds need. Owner attention is the scarcest ingredient in the building; the weekly read is how it gets spent where the money is.

The docket habit. Make it physical. Print the page — the same one-page docket, on the pass of your week: same day, same table, same twenty minutes, before service rather than after, when there is still a decision left in you. Sales at the top, the four cost lines under it, the one action written at the foot with a name and a date. Operators who keep the printed file gain something the screen never gives them: a spine of weeks, flipped through in thirty seconds, where a drifting line is visible as a trend before it is a crisis. Software makes the collection faster and a multi-outlet picture easier — that same discipline scaled across branches is the HO Control System — but it cannot supply the habit. Build the twenty-minute routine first; automate it once it exists.

Five lines, one method

Read the five nights back to back and the pattern is the method. Every night asks the same question — what did this line actually consume, against what actually came in — of a different line, in the order a turnaround works: the line that moves most for the least effort first, the line you can no longer change last, and then the rhythm that keeps all of them honest.

Two restaurants with the same revenue end up on opposite sides of profitability not because one is busier but because one holds four lines inside their bands and the other lets them drift a few points each, at once, quietly. Run the illustrative month again with the food line four points past its ceiling, labour two points past its own, and commission three points heavier because delivery grew: those three drifts take AED 39,000 more from the same AED 300,000 — 18,000, 12,000 and 9,000 — and leave 42,000 where 81,000 stood. The remainder has halved, and none of the lines moved dramatically enough to alarm anyone on its own. That is the whole story of “revenue is fine but the profit is gone”: a few points on several lines, compounding, discovered late. The full room and the empty account coexist far more often than the queue outside suggests — and the wider cost structure those bands sit inside is set out in restaurant profit margins in the UAE.

A discipline this simple has one honest failure mode: reading your own numbers with your own assumptions, every week, alone. That is what the founder’s second opinion exists for — one page of your P&L, reviewed personally, free, with limited monthly capacity, returning a straight read of which line deserves your next month. And if the whole structure needs more than a read — the four lines back inside their bands before any growth is attempted, because growth on a broken structure buys more of the same problem — the turnaround door is where that conversation starts.

Run the first night tonight

Start where the series starts. The Restaurant Profit Leak Audit takes five numbers — revenue, food cost, labour, rent and delivery commission — and ranks your three likeliest leaks against the bands above with an estimated monthly impact, in about two minutes, free and computed on your device. Print the docket it returns, put twenty minutes in the diary for the same day next week, and the five nights become the one habit that holds. It is smaller than the problem, which is exactly why it works.