Menu Engineering and Food Costing: Pricing for Profit, Not for a Percentage

Most first-time operators price every dish the same way: take the food cost, divide by 0.30, round to something that looks like a price. It's tidy, it's defensible to your accountant, and it costs you money at both ends of the menu.

It leaves dollars on the cheapest plates — a pasta with $4 of food prices out at $14 when the restaurant down the street gets $18. And it prices the expensive ones out of the market: a steak with $14.50 of food lands at $48, nobody orders it, and the item that would have thrown off the most dollars per plate never sells.

The percentage isn't wrong. It's the wrong number to price with. Three mechanics fix it: cost a dish so the number is true, price it in dollars instead of ratios, and run the menu-engineering matrix as a calculation instead of a wall poster.

If you haven't nailed down food cost as an aggregate yet — benchmark bands by concept, prime cost, the theoretical-versus-actual gap — start with restaurant financial basics. That's the portfolio view. This is the per-item version underneath it: theoretical food cost isn't a separate calculation, it's the sum of every recipe card in this guide, one dish at a time.

The Recipe Cost Card

A recipe cost card is one page per menu item: every ingredient, what it costs, how much the recipe uses, divided by portions produced. The output is one number — the food cost of one plate. Build a card for every item you sell and you have theoretical food cost; there's no other way to get it. Software can hold the cards, but somebody has to sit down with a scale and an invoice and build the first one.

AP cost, EP cost, and yield percentage

AP cost (as-purchased) is what the invoice says. EP cost (edible-portion) is what the part you actually plate costs, after everything you threw away. Two formulas: yield % = EP weight ÷ AP weight, and EP cost = AP cost ÷ yield %.

Buy whole ribeye subprimals at $12.49/lb and break them down yourself. 100 pounds in, 78 pounds of portionable steak out — a 78% yield.

$12.49 ÷ 0.78 = $16.01/lb EP

The invoice says $12.49. The plate costs $16.01. Cost that steak off the invoice and you've understated your most expensive item by about three dollars a plate. Any ingredient with real trim — whole proteins, anything with peel, core, root end, bone, or rind — costs more than the invoice says. Cost off AP and you're wrong on every card, always in the direction that flatters your margins.

Trim loss isn't the same for every ingredient

Trim loss covers fat, bone, skin, peel, seed, core, rind, and moisture lost in cooking.

Those bands are a sanity check, not your numbers. Yield depends on the cut, the supplier, the season, and — the one nobody plans for — who's holding the knife. Two cooks breaking down the same case of chicken can come out five points apart.

So yield-test your own product: weigh the case, break it down, weigh the usable output, divide. Three times on your top proteins, averaged, and re-test whenever the supplier, the cut, or the prep cook changes. If your menu leans on whole proteins, prep equipment that standardizes yield — a slicer, a portioner, a scale on the prep bench — narrows that variance more reliably than telling people to be careful.

Building the card: a full worked example

Rigatoni Bolognese, batched at twelve portions.

Twelve portions: $45.89 ÷ 12 = $3.82 per portion. Then add a Q factor — what goes to the table but isn't in the recipe: bread and butter, finishing parm, the oil you dressed the plate with. Run it as a percentage of food cost or a per-cover figure. Here it's $0.35.

Plate cost: $3.82 + $0.35 = $4.17. Target 30% food cost and $4.17 ÷ 0.30 = $13.90. You've hit your number exactly. Hold that price — the next section is about what's wrong with it.

Portion Control Is What Keeps the Card Honest

The card is theoretical until the line matches it. A card that says four ounces and a cook who puts down five isn't a costing problem, it's a costing lie — and it surfaces eight weeks later as a food cost number nobody can explain.

Portion scoops (dishers) are numbered by how many level scoops fill one quart: a #8 is 32 ÷ 8 = about 4 oz, a #12 about 2.67 oz, a #16 2 oz. Once the card says "#12 scoop" instead of "about 2.5 oz," the portion stops being a judgment call. Scoops are color-coded by size, but the mapping and exact volumes vary between manufacturers — buy one vendor's set and standardize on it. Same for ladles, sized in ounces. Portion scales are the other half: a calibrated scale at the prep bench and one on the line is what makes a 6 oz patty actually 6 oz. Portion scoops, ladles, and portion scales are the cheapest line item in this guide and the fastest payback — low hundreds of dollars for a small kitchen.

The card should state the tool, not just the weight. "6 oz" is a hope. "6 oz, #5 scoop, checked on the scale for the first three" is a procedure.

Where portion drift actually costs money

The per-plate number is invisible — nobody fights over eleven cents. So run it out at 1,900 entrées a month, with a half-ounce of drift on the proteins and two ounces on a hand-scooped fry basket:

About $4,800 a year from portions nobody would notice — and note which line is biggest. The cheapest ingredient with the loosest portion leaks more than the expensive protein everybody guards, because nobody weighs a fry basket and the volume is enormous. That $4,800 is one of the concrete things producing the theoretical-versus-actual gap restaurant financial basics covers at the portfolio level.

Food Cost Percentage Is a Ratio. Ratios Don't Pay Rent.

Contribution margin is menu price minus item food cost, in dollars. It's what's left from that plate to pay labor, rent, utilities, insurance, debt service, and eventually you. Food cost percentage tells you how efficiently a dish converts ingredients into revenue; contribution margin tells you how much money it made. Pricing to a flat percentage answers the first question and ignores the second.

The same trap, in numbers

Two dishes: the Bolognese at $4.17 plate cost, and a ribeye — 12 oz off that 78%-yield subprimal at $16.01/lb EP, plus potato, vegetable, and compound butter at $2.15, plus the same $0.35 Q factor — at $14.51.

On the pasta the percentage got better — 30% down to 24.5% — and you made $3.10 more per plate, because cheap ingredients divided by 0.30 produce a small number no matter what the guest would have paid.

On the steak the percentage got worse — 30% up to 38.2% — and the dish became sellable. At $48.37 it prints and sits there. At $38 it moves, and each one contributes $23.49, nearly double the pasta. Its food cost percentage is the worst on the menu and it's the most profitable plate in the building. Both moves were correct and they went in opposite directions, which only makes sense if you stop treating the percentage as the target.

Percentage measures efficiency. Contribution margin measures what's actually left to run the business. You pay rent in dollars, not in ratios.

One caveat: per-plate margin is half the equation. The pasta at $12.83 × 420 covers contributes $5,389 a month; the ribeye at $23.49 × 96 covers contributes $2,255. Per plate the steak wins by a mile; per month the pasta pays more of the rent. You need both — which is what the matrix combines.

Use both, in the right order

Food cost percentage is still a sanity check and the number a lender or landlord asks about; in aggregate it tells you whether the menu sits in a workable band, and restaurant financial basics covers those bands. The order is what changes. Per item:

1. Card the dish for a true plate cost with yields applied.
2. Set a target contribution margin in dollars: monthly non-food operating costs divided by the covers you realistically expect, plus the per-cover profit you want. Not every item has to clear it; the menu as a whole does.
3. Check the resulting price against what the market bears.
4. Then look at food cost percentage — as a check, not a target. Wildly out of band means look again; a few points off is noise.

One omission worth naming: labor cost per dish. Some operators charge each plate for the skilled minutes it consumes. More accurate, meaningfully more complex. Get the food side right first.

The Menu Engineering Matrix: Stars, Plowhorses, Puzzles, Dogs

The framework comes from Michael Kasavana and Donald Smith at Michigan State in 1982, adapting the BCG growth-share matrix to menu items. Still the right tool, as long as you calculate it. Two axes, both measured within a menu category — appetizers against appetizers, entrées against entrées, never the whole menu at once: popularity, the item's share of that category's unit sales off a POS sales mix report, and profitability, its contribution margin in dollars off the recipe card.

Stars sell well and make money. Plowhorses sell well and don't — volume drivers, quietly underpriced or overcosted. Puzzles make money when they sell, which isn't often. Dogs do neither. Pulling the sales mix is your POS documentation's job; what matters is the window — weeks, not a weekend.

The actual thresholds, not the vibe

Skipping this turns the matrix into a gut call with a diagram attached.

Popularity threshold. If every item sold equally, each would take 100 ÷ number of items percent of the category. The convention is that an item counts as popular at 70% of that equal share — (100 ÷ number of items) × 0.70. Eight entrées: (100 ÷ 8) × 0.70 = 8.75%.

Profitability threshold. The category's weighted average contribution margin — total category contribution margin divided by total units. Weighted, not a simple average, because a $23 margin on 96 plates shouldn't count the same as a $9.60 margin on 505.

Weighted average contribution margin: $24,602 ÷ 1,900 = $12.95. So the lines are 8.75% and $12.95, and every item lands in a quadrant off those two numbers. That's the whole calculation — a spreadsheet you can build in an hour once your cards exist.

What to actually do with each quadrant

Stars — leave them alone. Don't discount, don't shrink, don't quietly swap to a cheaper supplier for another forty cents. Price, quality, and demand are already in balance, and the most common way operators break a Star is by optimizing it. Feature it, photograph it, put it where people look.

Plowhorses — fix cost or fix price, carefully. Either find cost in the recipe — re-yield the protein, tighten the portion, swap a component, renegotiate at the volume you're actually buying — or nudge the price. A one-dollar increase on the burger is 505 × $1 = $505 a month, about $6,000 a year, from one line on a menu. But volume is what made it a Plowhorse: small increments, timed to a reprint, then watch the mix for four weeks.

Puzzles — fix exposure, not the recipe. The ribeye makes more per plate than anything else you sell and 96 people a month order it. Move it higher in the category, give it a real description, get servers to name it out loud, or offer a smaller portion to make trial cheap. If a quarter of that doesn't move the mix, it's a Dog — but exposure gets the first shot.

Dogs — cut them, with two exceptions. The eggplant parm is 4.6% of covers at a $10.20 margin and occupies a menu line, a prep station, cooler space, and ingredients nothing else uses. Keep one only if it's a strategic loss leader — a $3 side priced to get people in the door, which always scores as a Dog while doing a job the matrix can't see — or if it prevents a veto. If it's your only vegetarian entrée, it isn't earning $898 a month; it's earning whatever the four-top spends that would otherwise have gone elsewhere. Say the exception out loud: "we keep this because it's the vegetarian option" is a decision, "we keep this because we always have" isn't.

Where the matrix lies to you

It doesn't see labor. Contribution margin is price minus food cost. A Star that takes twelve minutes of a skilled cook may be a worse deal than a Plowhorse that takes ninety seconds. Similar margins, different prep — the fast one wins on a Friday.

It scores relative to the category, not your business. Every category produces Stars and Dogs by construction, because the thresholds come from the category's own numbers. A Star in a weak appetizer section is just the least-bad appetizer. Look at absolute dollars too.

Threshold items are knife-edge. The Bolognese is a Plowhorse by twelve cents — $12.83 against $12.95. Shift one item's mix and it becomes a Star without the dish changing at all. Anything within a few percent of a line is "watch," not "act."

Small samples produce premature answers. A new item hasn't had time to build a mix, and a slow week, a holiday, or one server pushing one item all distort it. A matrix built on a weekend isn't wrong so much as early.

Menu Design and Placement: Getting the Engineering Onto the Page

The matrix tells you what should happen. Design is how you make it happen without renaming half the menu.

The golden triangle. The common heuristic says the eye goes center, then upper right, then upper left on a single-page menu — the spots for Stars and the Puzzles you're rescuing. It's a widely used convention, not a law: eye-tracking research on menus is genuinely mixed, and several studies find people read more sequentially than the triangle suggests.

Price anchoring. A genuinely expensive item at the top of a category makes everything below it read as reasonable. The $38 ribeye doesn't need volume to earn its spot — above the $17 pasta, it reframes $17 as the sensible choice. That's the honest defense of a Puzzle you'd otherwise cut. Two cautions. The anchor has to be real, a dish you'd be proud to serve, because a fake one reads as a trick and guests spot it more easily than menu consultants assume. And you'll find a specific percentage lift attached to this claim across a lot of menu-engineering content — it traces through secondary sources to a citation we could not verify against the original research, so we're not repeating the number. The mechanism is well established; the figure isn't something to build a budget on.

Numerals instead of dollar signs. A 2009 Cornell Center for Hospitality Research study by Yang, Kimes, and Sessarego, run at the Culinary Institute of America's teaching restaurant, tested numeral-only ("18"), dollar-sign ("$18.00"), and spelled-out ("eighteen dollars"). Guests given the numeral-only menu spent significantly more, and the other two didn't differ meaningfully — the win wasn't hiding the price in words, it was dropping the currency cue. Source: Cornell Chronicle Don't oversell it: one study, one table-service restaurant, 2009. A counter menu, a digital board, and a delivery app listing are different reading environments, and nothing in that study says the effect transfers.

Three things that matter more:

• Kill the price column. Right-aligned prices with leader dots invite shopping down the column. Set the price in the same type at the end of the description and the guest reads the dish, not the list.
• Describe flavor, and highlight one item per category, not six. A short, concrete description on a Star or Puzzle beats a box around it, and if everything is boxed, nothing is.
• Shorten the menu. A long menu gives Dogs places to hide, slows the line, spreads prep across more SKUs, and increases spoilage.

Jurisdiction check — menu labeling and allergens. Federal calorie-labeling rules apply to chains with 20 or more locations, so as a first-time independent they almost certainly aren't yours. But some states and cities layer their own allergen-disclosure requirements on top, independent of chain size, and a few require specific language on the menu itself. Ask your health department what your menu has to disclose rather than assuming the federal chain rule is the whole picture.

Repricing When a Key Ingredient Moves

Costing isn't a project you finish. The card you built in March is fiction by August if you never touched it.

A rule of thumb many operators use: a sustained supplier price change of roughly 10% or more on an ingredient is worth an immediate recost of every dish that uses it, rather than waiting for a scheduled review. Practitioner guidance, not a standard — pick a threshold that fits your volatility and your reprint cycle. The word doing the work is sustained. One high invoice is noise; three in a row is a price.

When it triggers: new AP cost, re-apply the yield, rebuild the extended costs, get the new plate cost and contribution margin. Then choose from four responses.

1. Absorb it. Turns entirely on the ingredient's share of your food spend. Canned tomatoes at 2% of purchases going up 15% is 0.3% of food cost — absorb it, reprint nothing. Beef at 30% of purchases going up 15% is 4.5% of food cost, and absorbing that means eating several points of margin indefinitely. Do the share-of-spend math first.

2. Reprice — using contribution margin, not a percentage bump. Ground beef goes from $4.79 to $6.15/lb AP. At the same 72% yield, EP moves from $6.65 to $8.54/lb. The batch goes from $45.89 to $51.56, so plate cost goes from $4.17 to $4.65 — a 48-cent increase on a dish selling at $17.00.

The percentage method multiplies the cost increase by roughly 1 ÷ your food cost percentage — here, about four times. Not automatically wrong; if labor and rent moved too, a bigger increase may be right. But it should be a decision you made, not an artifact of the formula you happened to use.

3. Reformulate. Change the dish instead of the price: shift the ratio (more soffritto, less beef in a ragù), swap a component that isn't moving, or resize a portion deliberately and once rather than letting it drift. Preserves the price point, and it has a ceiling — do it too often and you've changed the dish people come for.

4. Add a transparent, time-bound surcharge. In early 2025 egg prices spiked hard — the USDA's outlook at the time projected egg prices rising more than 20% on the year against roughly 2% for food overall. Several chains responded with a disclosed per-egg surcharge rather than reprinting menus; Waffle House's was the most publicly discussed, posted on signage and removed later in the year as prices came back down. Source: Forbes A surcharge is honest, fast, needs no reprint, and is reversible in a way a menu price increase never really is. It also puts the increase in the guest's face on every ticket and invites a conversation with a server who didn't set the price, which a quiet 50-cent adjustment doesn't. Use it for a genuinely temporary, externally-caused, widely-reported spike the guest has already read about — not for anything else.

Jurisdiction check — surcharges. Whether you may add a surcharge at all, how it must be disclosed (menu note, table tent, receipt line, pre-purchase notice), and whether it can be a flat fee versus a percentage are set by state and sometimes city law. Several states have passed price-transparency or "junk fee" laws that squarely affect restaurant surcharges, and credit-card surcharge rules are a separate thicket. Check your state attorney general's or consumer-protection agency's current guidance before you add the line, not after — and whether the surcharge is itself taxable is another state-by-state question for your bookkeeper or state revenue department.

A recost cadence, not a one-time event

"Check occasionally" isn't a process. A workable cadence many operators use, adapted to your own volatility:

None of this is a regulatory or accounting requirement. It's an operating habit, and the argument for it is that operators who keep it find problems while they're small. If the full-menu recost isn't on a calendar it won't happen, and the version that happens in a panic after a bad P&L is always worse than the one on a Tuesday in February.

How This Differs by Track

The mechanics are identical everywhere. What changes is the data you can get and the margin you need.

Restaurant. The most complete version: a real POS sales mix, categories big enough for the thresholds to mean something, and a check average that gives room to move prices in small increments. The discipline is patience — weeks of data, and no repricing after one strong Saturday.

Food truck. A short, station-limited menu means fewer items per category. Less room for a Dog to hide, but at five items the popularity threshold is (100 ÷ 5) × 0.70 = 14% and one item's swing moves the whole picture. Lean on absolute contribution margin per item, and on station capacity as a hard constraint the matrix can't see — the food truck equipment guide covers how menu breadth ties to it.

Ghost kitchen. Platform commission comes off the top of the same ticket your contribution margin is trying to protect. A $17 pasta with $12.83 of margin isn't that once the platform takes its share, plus packaging, plus the promotions you ran to stay visible. A delivery-only menu needs a wider per-item contribution margin than a dine-in menu to land in the same place — price for it from the start rather than discovering it in month four. For commission tiers and the breakeven math on direct versus marketplace ordering, see the commission math on delivery platforms; for the equipment side, the ghost kitchen equipment guide.

An engineered menu also changes what the kitchen needs. If the burger and fries carry your volume, griddle and fryer capacity is your throughput ceiling; cut four Dogs and you may have freed a station and a chunk of cooler space you were paying to run. Settle the menu, then buy the kitchen — the complete restaurant equipment checklist covers what a full kitchen costs new versus used.

Frequently Asked Questions

How do you calculate food cost for a recipe?

Take each ingredient's as-purchased cost from the invoice, divide by its yield percentage to get edible-portion cost, multiply by the quantity used, and total the extended costs. Divide the batch total by portions produced, then add a Q factor for bread, garnish, and condiments that go out with the plate but aren't in the recipe. The step most people skip is the yield adjustment, which understates cost on every ingredient with real trim.

What is a recipe cost card?

A one-page record of what one menu item costs to produce: every ingredient with its as-purchased cost, yield percentage, edible-portion cost, quantity used, and extended cost, divided by the portions the batch makes. Built for every item, the cards are what theoretical food cost is made of — the sum of your cards weighted by what sold, not a separate calculation. The card should state the portioning tool, not just the weight.

What's the difference between food cost percentage and contribution margin?

Food cost percentage is item food cost divided by menu price — a ratio measuring efficiency. Contribution margin is menu price minus item food cost — dollars, measuring what's left to pay labor, rent, and profit. Pricing to a flat percentage leaves money on cheap dishes and prices expensive ones out of the market: a steak at 38% food cost can contribute $23 a plate while a pasta at 24% contributes $13, and you pay rent with the $23.

What are stars, plowhorses, puzzles, and dogs in menu engineering?

The four quadrants of the menu engineering matrix developed by Kasavana and Smith in 1982, plotting each item in a category on popularity and contribution margin. Stars are high on both — feature them and leave them alone. Plowhorses sell well at a low margin — find cost in the recipe or nudge price carefully, since volume makes a small increase compound. Puzzles have high margin and low sales — fix exposure, not the recipe. Dogs are low on both and usually get cut, unless one is a loss leader or your only item for a dietary restriction.

How do I price a new menu item?

Card it first for a true plate cost with yields applied. Then set a target contribution margin in dollars: monthly non-food operating costs divided by the covers you realistically expect, plus the per-cover profit you want. Add that to the plate cost for a starting price, check it against what comparable restaurants near you charge, and look at food cost percentage last. If the percentage is a few points off but the dollars work, the dollars win.

When should I raise menu prices because of ingredient costs?

A common operator rule of thumb is that a sustained price move of roughly 10% or more on a meaningful ingredient warrants an immediate recost of every dish using it — emphasis on sustained, since one high invoice is noise. Then choose deliberately: absorb it if the ingredient is a small share of food spend, reprice using contribution margin so you pass through the actual increase rather than multiplying it by a percentage formula, reformulate, or add a disclosed temporary surcharge. Whether a surcharge is permitted and how it must be disclosed varies by state and city.