Prime Cost from 68.4% to 61.1%: how we decided which dishes to remove from the menu to gain profitability at a 14-table trattoria using the Restaurant Model Canvas and the Standard Recipe Generator

Which menu items should you cut to gain profitability?

The items that leave the menu are those with a low CONTRIBUTION MARGIN in dollars per unit sold that also carry little weight in the mix, not the ones that sell least:

at this 48-seat trattoria, 17 of the 62 references were cut, all of them below 4.10 USD of unit contribution. The textbook mistake is reading the unit ranking and trimming the tail, because a slow item leaving 11 USD clean supports a weak night far better than a bestseller leaving 2.80 USD. This house's four top sellers ranged from 2.80 to 4.05 USD, moved 31% of the units and delivered barely 14% of total contribution, so the menu was working against the owner's cash while sales climbed 9% year over year. Theoretical food cost on that menu, built dish by dish with standard recipes and real waste, came to 31.8% of sales, comfortably inside the 28-35% optimal range published by the National Restaurant Association (2025), while inventory returned a real cost of 38.9%.

Seven points of gap the P&L never showed

Seven points of gap. On annual revenue in the 500 thousand to 1 million USD band, those seven points are roughly 51 thousand USD leaving the business every year without ever appearing as a loss on any line of the income statement, because a P&L that closes 40 days late reports a hemorrhage that already stopped. The benchmark matters: median limited-service food cost was 32.4% of sales in 2024 according to the National Restaurant Association, and casual dining runs around 30-34%, so the theoretical number was healthy and the problem lived in execution, not in the recipe book. Root cause showed up over three days of blind weighing on the house ragù: the portion served swung between 142 and 218 grams depending on who worked the line, a 53% spread on one single recipe. Forty-one of the 62 references had no closed standard recipe, so every cook portioned from memory, and memory on a Friday shift at half past nine always portions upward.

The blind weighing of the ragù: 142 to 218 grams

A 62-item menu for 48 seats is not generosity, it is physical impossibility: nobody sustains 62 technical sheets with eleven employees across kitchen and floor. That spread does not get fixed with a scolding at the pass or a gram chart taped to the wall; it gets fixed by removing references until the number of sheets fits inside the head of the team that has to execute them twice a day. We loaded all 62 references into the menu-engineering matrix of the Masterestaurant method, the tool Diego F. Parra uses to cross unit contribution in dollars against mix share and against prep labor hours, the variable almost everyone forgets. Each dish landed in a quadrant and left with a single instruction: cut, reformulate the recipe, raise price or relocate on the menu. The 17 that were cut shared three traits: contribution under 4.10 USD, less than 1.2% of the mix each, and at least one exclusive ingredient used in no other dish.

How the cut was made with the Masterestaurant contribution matrix?

That last criterion returns the most money, because every orphan ingredient drags inventory, waste and a slot of walk-in space nobody accounts for.

The menu dropped from 62 to 45 references in a single printed version. Ninety days after the cut, real food cost fell from 38.9% to 33.1%, and the gap against theoretical —which rose to 32.2% once eight recipes were reformulated— settled at 0.9 points, a range normal waste explains without any loss of control. Average ticket moved from 18.40 to 20.15 USD with no across-the-board price increase: fourteen prices were adjusted, 17 dishes disappeared and the mix drifted on its own toward high-contribution references, now visible because they were no longer buried among sixty-two lines of cramped type. On the same cover volume, monthly contribution grew by roughly 4,200 USD. The owner recovered his draw, which had been falling for fourteen months, and for the first time in nine years he knew what each plate left him before cooking it.

The paradox of the short menu and the regular who orders the usual

Every owner fears the same thing when cutting: the loyal guest finds his usual dish gone and never comes back. What would have happened had we kept the 17 out of caution? Real cost would have stayed at 38.9%, technical sheets would have stayed open because 62 simply do not fit, and the owner would have financed a minority's whim out of his own draw. We measured the fallout: of those 17 dishes, twelve moved fewer than three units a week, and of the complaints logged in the dining room during the first month, eleven of fourteen were settled by offering a neighboring reference from the same quadrant. The tension resolves on the gram side: a shorter menu executes better, and a guest who always receives the same portion of the same ragù comes back more often than one who loved a dish nobody could cook the same way twice.

Transferable lessons by revenue band

Below 500 thousand USD a year, your first step this week is to weigh your five top sellers across three services and write down the real gram range, telling nobody. Between 500 thousand and 1 million —this case's band— calculate dollar contribution for the ten references moving the most units and compare it with the menu average: if your bestsellers sit below it, you already have the diagnosis. Above 1 million, demand closed standard recipes at 100% before touching a single price, because without a sheet there is no theoretical cost and without a theoretical cost there is no gap to measure. Above 5 million, the celebrity-chef archetype running a large-format concept lives on signature plates nobody dares touch: audit what share of total contribution those untouchables actually carry. Above 10 million, a multi-unit group starts by unifying the master recipe book across locations before making any cut.

Limits of this case

I would not expect these numbers in a delivery-only operation dominated by aggregators: here the dining room carried 78% of sales, and when an aggregator commission takes 18 to 30 points off the ticket, the main lever stops being the menu and becomes the channel mix, because a dish leaving 4.10 USD of contribution in the dining room can turn negative once it goes out through an app. Nor would I expect it in fine dining, where the National Restaurant Association places normal food cost between 34 and 40%: there a 38.9% cost is not a leak, it is the business model, and menu cuts get decided on concept coherence before contribution. And in a business built on one protagonist ingredient heavily exposed to the beef cycle —the US herd sits at roughly 86 million head, its lowest since the 1950s according to the USDA (2025)— margin moves through purchasing and contracts, not by deleting references.

Root-cause diagnosis: every symptom, its cause, and the number that exposed it

SYMPTOM: seven points of gap between theoretical and actual food cost. ROOT CAUSE: 41 of the 62 items had no closed standard recipe, so every cook portioned from memory. What exposed it was a three-day blind weighing of the house ragù: the served portion ranged from 142 to 218 grams depending on who worked the line, a 53% spread on the same recipe. SYMPTOM: rising sales with a falling owner's draw. ROOT CAUSE: the sales mix had drifted toward the lowest-contribution items because a 62-dish menu displayed them first. The four best sellers left between 2.80 and 4.05 USD of contribution margin each; together they made up 31% of units and barely 14% of total contribution. SYMPTOM: a 29.5% Labor Cost the owner believed could not move without firing someone. ROOT CAUSE: he did not have too many people, he had too much MENU.

Root-cause diagnosis: every symptom, its cause, and the number that exposed it — in practice

Sixty-two items forced 4.5 hours of daily mise en place and kept 23 ingredients turning less than once a week. Cutting to 45 dishes dropped mise en place to 2.8 hours and freed 28 weekly hours without touching headcount. SYMPTOM: protein waste nobody could explain. ROOT CAUSE: six beef items using different cuts, each purchased in small volumes. Beef costs were already squeezed by the contraction of the US cattle herd to roughly 86 million head, its lowest since the 1950s according to the USDA (2025), and buying six cuts instead of two multiplied both the price per kilo and the trim that ended up in the bin. SYMPTOM: a P&L that always arrived late and always with a surprise. ROOT CAUSE: a 40-day accounting close and zero weekly Prime Cost control. The owner made September decisions with July numbers. I got this wrong for years by recommending elegant monthly dashboards: an independent restaurant needs one ugly number on time, not a pretty one that is stale.