Menu engineering: how to use your sales data to decide what stays on the menu

Last year, 60% of full-service restaurant operators took items off their menus because food costs went up. So did 43% of limited-service operators. That’s from the National Restaurant Association’s 2026 State of the Industry report, and it matches what we hear from owners: costs rose, something had to go, and most people decided what to cut by feel.

There’s a better way, and it’s been around since 1982. It’s called menu engineering, and it uses two numbers you already have in your point-of-sale system: how often each item sells, and how much money it leaves you after food cost. This post walks through how it works, with a worked example you can copy, what the research actually says about menu design, and where the classic method falls short.

What menu engineering is

Michael Kasavana and Donald Smith developed menu engineering at Michigan State University and published it in 1982. The idea is simple: judge every item on two things at once.

• Popularity: how many of this item you sell compared with the other items in the same part of the menu.
• Contribution margin: the menu price minus the food cost of one plate. It’s the dollars each sale leaves you to cover labor, rent, and everything else.

Food cost percentage alone can mislead you. A $30 steak at 40% food cost leaves $18. A $12 pasta at 25% leaves $9. The steak looks worse on a percentage basis and puts twice as many dollars in the bank. Menu engineering looks at the dollars.

The four boxes

Put the two measures together and every item lands in one of four boxes.

How to run it, step by step

You need a spreadsheet and a report from your POS. Do one section of the menu at a time.

1. Pick a group and a period. Sandwiches, entrées, or pastas, over the last month or quarter. Only compare items that compete with each other. Don’t mix sandwiches with drinks.
2. Pull units sold for every item in the group.
3. Work out each item’s menu mix: its units divided by the group’s total units.
4. Set the popularity line. It isn’t the average. Kasavana and Smith use 70% of an equal share: (1 ÷ number of items) × 70%. With five items, that’s 14%. An item at or above the line is popular.
5. Work out each item’s contribution margin: menu price minus plate cost.
6. Set the margin line as the weighted average: add up (margin × units sold) for every item, then divide by total units. Some online guides use a plain average of the margins instead. That isn’t the original method, and it gives popular items too little weight.
7. Sort each item into its box.

Here’s what the table tells you to do:

• Italian combo (Star): it sells the most and earns the most per sale. Don’t touch the recipe or the price in a way guests would notice.
• Turkey pesto (Plowhorse): a $1 increase takes its margin to $10.90, above the line. If most buyers stay, it becomes a second Star.
• Caprese (Plowhorse): it only just clears the popularity line and has the thinnest margin. Look at the mozzarella and the portion before the price.
• Meatball (Puzzle): the best margin on the list, and hardly anyone orders it. Move it to the top of the list, give it a line that makes it sound worth ordering, and have the counter recommend it for a month.
• Prosciutto & fig (Dog): low sales and a below-average margin. Rework it or replace it, unless it’s the item that brings certain regulars in.

What the research says about menu design

Menu design is full of confident advice. Some of it holds up. Here’s what the studies actually found, and how much weight to put on each.

• First and last get noticed. In a 2011 study that included a real restaurant, items listed first or last in their category were up to twice as popular as the same items listed in the middle. This is the strongest, easiest change on this list: put your Stars and Puzzles at the top or bottom of their section.
• The “sweet spot” is probably a myth. The idea that eyes land first on the upper right of a menu goes back to a 1987 industry report. A 2012 eye-tracking study found diners read menus in order, like a book, with no sweet spot. A 2003 study that swapped two pages of a menu for four months found no significant change in what sold.
• Dollar signs may cost you. A 2009 Cornell field test found that menus showing prices as plain numbers (“20”) brought in more per check than “$20.00” or “twenty dollars.” It was one restaurant at lunch, and the authors asked for it to be repeated, so treat it as a cheap experiment worth trying, not a rule.
• A more expensive option makes the next one look reasonable. A 2024 study found that adding a pricier “decoy” bundle sharply increased how many people chose the upgraded bundle next to it. It was a survey experiment, not real sales, but it matches what most operators see with good, better, best pricing.
• Descriptions help, with a caveat. A 2001 cafeteria study found descriptive names raised sales of those dishes by 27%. The paper hasn’t been retracted, but its lead author was later found to have committed research misconduct on other work, so don’t lean on the exact number.

Where the classic method falls short

Menu engineering is a starting point, not an autopilot. Watch for these:

• It ignores labor. Contribution margin only subtracts food cost. A dish that takes three times as long to make can look like a Star and still cost you. Researchers added a labor rating to the model back in 1995. At minimum, flag the items that slow your kitchen down.
• Takeout changes the margin. Packaging comes out of every to-go order, and delivery apps take a commission on top. DoorDash’s delivery commission is 15%, 25%, or 30% depending on the plan. If a big share of an item goes out the door, run its numbers with those costs included.
• Cutting one item moves demand to others. A study of two years of data from 48 steakhouses found that removing or repricing one item shifted sales to similar items. Watch the whole group for a few weeks after a change, not just the item you touched.
• Some Dogs earn their spot. A dish with a loyal following, a vegetarian option that lets a group say yes, or the item a regular orders every week can be worth keeping even if it scores badly.
• Seasons move costs. Ingredient prices and availability shift through the year, so a Star in March can be a Plowhorse by August.

How often to run it

Kasavana recommends checking the numbers every quarter and revising the menu about twice a year. Also run it whenever you change the menu seasonally or a key ingredient jumps in price.

The short version

• Compare items only within the same section of the menu.
• Judge them on units sold and dollars per sale, not food cost percentage.
• Use 70% of an equal share for the popularity line and the weighted average for the margin line.
• Fix Plowhorses, sell Puzzles harder, protect Stars, question Dogs.
• Put the items you want to sell first or last in their section.
• Account for labor, packaging, and delivery fees before you trust the result.
• Check every quarter, and revise twice a year.