Your menu has four kinds of dishes

Ask an owner which dish makes them the most money and most will name the one with the lowest food cost percentage. It sounds right. It is usually the wrong answer, and a menu built on it quietly pushes guests towards the dishes that earn the least.

The percentage tells you how much of the price goes on ingredients. It does not tell you how many euros each plate leaves behind, or how many plates you sell. Those two numbers together are what pays the rent.

There is an old and very simple method for putting them together. It is called menu engineering, and it sorts every dish on a menu into one of four groups in about an hour.

Two numbers per dish

The method was described by Michael Kasavana and Donald Smith in their book Menu Engineering, and it still works because it asks only two questions of each dish.

The first is what the dish leaves behind. Take the menu price without VAT and subtract the plate cost. What is left is the contribution: the euros that plate puts towards wages, rent and profit.

Contribution = price without VAT − plate cost

The second is how often it sells. Count how many of each dish went out over a month, from the till, and compare dishes only within the same section: mains with mains, starters with starters.

Each answer gets a line to compare against. For contribution, the line is the average contribution per plate across the section, weighted by how many of each were sold. For popularity, the line is 70 per cent of an equal share: with six mains, an equal share is one in six, and 70 per cent of that is 11.7 per cent of everything sold.

Six mains at Acacia

Acacia is the example restaurant I use for worked sums. The salmon is the same plate whose cost we worked out in the food cost post; VAT is taken out at 10 per cent. Here is one month of mains:

Altogether 1,320 mains went out and left €16,473 behind, which is an average of €12.48 a plate. The popularity line is 11.7 per cent, or 154 plates in the month.

The four groups, and what to do with each

Star: sells well and earns well. The salmon. Leave the recipe alone, give it the best spot on the menu and make sure the kitchen can always deliver it. Stars are the dishes you lose money by tinkering with.

Plowhorse: sells well, earns below the line. The burger and the schnitzel. Guests love them, so taking them off is a mistake. The usual moves are a small price rise, a cheaper garnish or a smaller side that nobody misses. If the burger went from €16 to €17 and still sold 420, it would leave €0.90 more per plate, or €378 a month. Whether it still sells 420 is the part you test, not assume.

Puzzle: earns well, sells below the line. The ribeye and the risotto. The money is there; the guests are not. Look at where the dish sits on the menu, what it is called, how it is described and whether the floor ever recommends it. If thirty guests a month who now order the schnitzel ordered the risotto instead, that is €72 more a month from a dish already on the menu.

Dog: sells little and earns little. The lamb shank. It takes hours of oven time and a portion of fridge space for sixty plates a month. Rework it, reprice it, or replace it. If it stays because a regular loves it, that is a fair reason, as long as it is a decision and not a habit.

Where food cost percentage gets it wrong

Now look at the same six dishes by food cost percentage alone. The ribeye is the worst on the menu at 44 per cent, and the risotto is the best at 17. A menu run on percentages would hide the steak and push the risotto.

Yet every ribeye leaves €17.41 behind and every risotto €12.85. Guests do not pay you in percentages. The steak that looks expensive on the costing sheet earns more euros per plate than anything else on the menu. The percentage is a useful alarm for a single dish that drifts. It is a poor way to compare one dish with another.

Once a month, one hour

• Take last month's sales per dish from the till, one section at a time.
• Use a costed plate for every dish. Without it, the contribution is a guess.
• Take VAT out of every price before anything else.
• Work out the two lines, sort the dishes into the four groups and pick one action per group.
• Change one thing at a time and look again next month, so you know which change did the work.

The matrix has known limits. It counts ingredients but not the minutes a dish takes on the line, and a dish that is cheap in food can be expensive in labour. Book 1 of the Menu Trilogy, The Choice Algorithm, takes the method further, counts the cook's minutes as well, and covers the layout and wording that move a puzzle up.