Menu engineering explained

Menu engineering is the practice of pricing and arranging a menu around two facts about every dish: how much money it contributes after food cost, and how often guests order it. Most menus are built around neither. They grow item by item, inherit prices from an older cost base, and end up quietly steering guests toward the dishes that make the least.

Start with contribution margin, not food cost percentage. Contribution margin is menu price minus plate cost — the actual dollars a dish leaves behind. A $34 steak at 38 percent food cost contributes $21. A $16 pasta at 22 percent contributes $12.50. The pasta has the better percentage; the steak pays more of the rent. Percentage is a control metric, dollars are the decision metric, and menus optimized on percentage alone tend to lose money politely.

Then pull popularity. Export item-level sales from the POS for a full trading cycle — at minimum four weeks, ideally a quarter — and calculate each dish's share of sales within its category. Comparing a starter to an entrée is meaningless; compare like with like.

Plot the two together and you get the four quadrants. Stars are high margin and high popularity: protect them, never discount them, and give them the best real estate on the page. Plowhorses are popular but low margin: re-cost the plate, adjust portion or garnish, or raise the price modestly — guests notice a dollar far less than operators expect. Puzzles are high margin and unpopular: they need a better description, a server mention, or a move up the page before you judge them. Dogs are low margin and unpopular: cut them, and reclaim the prep time and inventory they consume.

Design carries as much weight as the math. Guests read the top right of a page first and the first two items of any list hardest. Boxing, whitespace and a short descriptive line all lift orders. Dollar signs, price columns and trailing zeros invite price comparison; setting prices as plain numbers reduces it. Keep categories to five to seven items — long lists slow decisions and push guests to the safe, usually cheaper, choice.

Re-run the exercise every quarter and after any supplier price move above five percent. Plate costs drift, and a menu costed eighteen months ago is not a menu, it is a historical document. Operators who hold this discipline typically find two to four points of food cost without changing a single supplier.

One caution: menu engineering fails when plate costs are guesses. Before any of this works, every recipe needs a costed spec with real yields. That unglamorous week of costing is the whole foundation.