Menu engineering is the practice of looking at every dish on two axes at once — how profitable it is and how often it sells — and then doing something specific about each group. It is one of the few restaurant exercises with a genuinely high return for an afternoon of work.
The two numbers per dish
Contribution margin. Menu price minus food cost. In dollars, not as a percentage — this is the part people get wrong.
A dish with a thirty per cent food cost sounds better than one at forty. But a $12 dish at 30% contributes $8.40, and a $34 dish at 40% contributes $20.40. You bank dollars, not percentages.
Popularity. Units sold over a period, as a share of its category.
Plot every dish on those two and four groups appear.
The four groups and what to do
High margin, high popularity. Your winners. Do not touch the recipe, do not touch the price without care, and make them impossible to miss on the menu. Protect these.
High margin, low popularity. The opportunity. These make money when they sell and they do not sell. The fix is usually presentation or description rather than price — move it, describe it better, have servers recommend it. If it still does not sell after that, it does not suit your customers.
Low margin, high popularity. Dangerous, because volume disguises them. Every sale is a lot of work for little contribution. The fix is cost, not price: re-examine portioning, the protein spec, the garnish. A small cost reduction on a high-volume dish is worth more than a large one on a rare dish.
Low margin, low popularity. Remove them. This is the group people defend hardest and the removal is almost always right — every dish carries inventory, prep, training and menu space.
The exercise
One category at a time. Starters, then mains. Comparing a starter to a main tells you nothing.
Use real food cost, costed from current invoices, not from a spreadsheet written eighteen months ago. Prices have moved and your costings almost certainly have not.
Use a real period. A full month at least, and be aware of seasonality.
Include everything on the plate. The garnish, the sauce, the bread, the oil. Restaurants routinely cost the protein and forget the rest, which understates food cost by a few points across the board.
Where the menu itself does work
Position matters. People look at certain areas of a menu more than others. Your high-margin dishes should be where eyes land, not buried in the middle of a long list.
Length works against you. A long menu means more inventory, more waste, more prep, slower service, and a customer who defaults to whatever is familiar because choosing is work. Shortening a menu improves prime cost and usually improves what people order.
Description sells. A dish with a sentence that makes it sound like something specific outsells the same dish listed as three nouns.
Avoid a column of prices. A right-aligned column invites scanning down for the cheapest. Prices set inline after the description get read as part of the dish.
What not to do
Do not raise prices across the board. Menu engineering is per dish, which is the entire point. A blanket increase raises your winners into resistance and leaves your losers unfixed.
Do not cut portions on your winners. The dishes people come for are the ones to leave alone. Take cost out of the high-volume, low-margin group instead.
Do not keep a dish because one regular likes it. Cost the dish honestly, including the inventory it alone requires. You are usually better off making that regular something special off-menu.
Doing it again
Menu engineering is not a one-off. Costs move, seasons move, and mix changes as soon as you change the menu.
Quarterly is a reasonable rhythm, and it takes an afternoon once the first one is done. The numbers come out of your POS if it reports sales by item — which is the reason to check that report exists before you buy one. What a restaurant POS system actually does covers what to look for.
And the whole exercise feeds prime cost, which is where the result shows up. Prime cost: why food and labour only make sense together has that side.