Menu engineering is one of the few pieces of restaurant theory that genuinely earns its keep. Sort every dish by how well it sells and how much margin it carries, put them in four boxes, and the boxes tell you what to promote, what to reprice, what to rework, and what to cut. It is simple enough to do on paper and it usually finds money.
There is one assumption underneath it that nobody examines. The whole exercise depends on knowing what each dish costs, and in most restaurants that figure was calculated once, at some point, and has been quietly aging ever since.
The Four-Box Matrix and What It Assumes
The classic version plots popularity against contribution margin and gives you four groups.
Your stars sell well and carry good margin, and they are the dishes you protect and push. The workhorses also sell constantly but barely pay their way, which is where a small price or spec change moves real money. Puzzles are the good dishes nobody finds, and that is usually a menu design or description problem before it is anything else. Whatever is low on both counts sits in the bottom corner as a candidate for removal.
The actions that follow are sensible and well tested. A star gets moved to a better position on the page. A workhorse gets repriced or reworked until it pays properly, while a puzzle usually needs a better description more than anything else. The bottom corner comes off the menu.
The assumption sitting under all of it is that a dish belongs in one particular box. And that placement depends entirely on the contribution margin, which is the selling price minus the dish cost. The selling price you know exactly. The dish cost is where the trouble is.
What Happens to the Boxes When Prices Move
Ingredient prices do not hold still, and they do not move together.
A dish costed at four euros eighteen months ago is not a four-euro dish now. Some ingredients went up sharply while others only drifted, and one or two actually came down. The individual moves are small enough that nobody notices them on a single invoice, which is exactly how price changes get past you in the first place. Compounded across a year and across every line on a recipe, the total is not small.
The important part is that the moves are uneven. If everything rose by four percent, every dish would shift by roughly the same amount and the matrix would hold its shape. What actually happens is that a beef dish takes an eleven percent hit while a pasta dish takes two, and the ranking between them changes.
That is what makes a stale matrix worse than no matrix. A dish that has slid from a healthy margin into a thin one is still sitting in the star box, promoted, and holding the best position on the page. You are actively pushing the thing that stopped paying months ago. Meanwhile something you were about to cut may have become perfectly viable because its main ingredient came down.
Kitchens tend to discover this the slow way, through a food cost percentage that keeps creeping despite nothing obviously changing.
Recosting Before You Re-engineer
The order matters. Recost first, then run the matrix, because old costs produce confident answers to the wrong question.
Recosting means going back to what you actually paid recently, not what the recipe card says you paid. For a menu of thirty dishes that is a real piece of work by hand, which is why it so rarely happens between menu changes. The practical shortcut is to recost by ingredient instead of by dish. Identify the twenty or so ingredients that carry most of your spend and find their current prices. Let those flow through to every recipe that uses them. Most of the movement in your menu will come from that handful.
While you are in there, check two things that distort dish cost independently of price. The first is whether the recipe reflects what the kitchen actually plates, because a spec that says a hundred and eighty grams against a line serving a hundred and ninety-five is understating your cost before any price movement is considered. That is a portioning question and not a costing one, though it lands in the same number. The second is yield, especially on anything trimmed, reduced, or butchered in house. Costing a portion on raw weight when a third of it goes in the bin makes the dish look better than it is.
Then build the matrix. On current numbers it will not look like the one you drew last time.
Which Dishes Move Most and Why
Some dishes are stable and some are volatile, and knowing which is which is more useful than any single snapshot of the matrix.
Dishes built on one dominant expensive ingredient move the most. A steak is a beef price with a garnish attached, so it tracks the beef market almost one for one. When the ingredient moves eight percent the dish moves nearly eight percent.
Dishes with many cheap components are stable. A pasta with six ingredients, none of them dominant, absorbs individual price moves because they partly cancel out. Those dishes tend to stay in their box for years.
Seafood is its own category and worth watching separately, because it moves on availability instead of along a smooth cost curve, and it can shift a lot in a fortnight.
The practical output is a shortlist: half a dozen dishes whose margin moves with the market, with everything else left alone between proper menu reviews. Those six are the ones worth a monthly look, and they are usually the ones with the biggest headline prices too, which makes repricing them a delicate commercial decision and not a mechanical one.
Making It a Quarterly Habit
Menu engineering usually happens when someone gets worried, which means it happens rarely and under pressure.
A better rhythm is quarterly for the full exercise and monthly for the volatile shortlist. The full pass is the one where you actually change the menu: reposition, rework, reprice, cut. The monthly check on the volatile dishes is much lighter and exists to catch a dish sliding underwater between full reviews, so you find out within weeks instead of at the next menu print.
What makes this sustainable has little to do with discipline. It comes down to whether the costs are current without anyone having to go and fetch them. If recosting is a two-day job, the quarterly review will happen twice a year at best. When recipe costs update themselves from the invoices already arriving, the review becomes a conversation about what to do instead of an exercise in data gathering, and that is the difference between a habit and an intention.
Frequently Asked Questions
What is menu engineering? Analysing every dish on your menu by how well it sells and how much margin it contributes, then sorting them into four groups so you know which to promote, reprice, rework, or remove.
How often should you re-engineer a menu? A full pass quarterly, with a lighter monthly check on the handful of dishes whose cost tracks ingredient markets closely. Anything less frequent and the analysis is running on costs that have moved since.
Why does menu engineering fail? Most often because the dish costs feeding it are out of date, so dishes sit in the wrong box and the resulting actions push the wrong items. The second most common reason is recipes that do not match what the kitchen actually plates.
Do I need software to do menu engineering? The analysis itself works on paper. What is hard by hand is keeping dish costs current between reviews, which is the part that decides whether the analysis is right. That is where live recipe costing earns its place.
Working From Numbers That Are Current
Menu engineering is good theory undermined by stale inputs. Recost against what you are paying now and check that the recipes match what leaves the pass, then build the matrix and act on it. Keep a short list of market-sensitive dishes under a monthly eye and leave the stable ones for the quarterly pass.
Stockifi reads supplier invoices line by line and updates every recipe as ingredient prices move, so the margin on each dish reflects this week and not last year. See how recipe building works.