Restaurant Menu Engineering: The Complete Guide
Most menu prices are set once on instinct and never reviewed again. Menu engineering replaces that with two numbers you already have — how well an item sells and how much it actually contributes — and tells you exactly what to do with each dish.
Updated August 2026 · 9 min read
What is menu engineering?
Menu engineering is the practice of analysing every item on your menu on two axes — popularity and profitability — and then acting on the result: repricing, redesigning, reworking or removing. It was formalised by Michael Kasavana and Donald Smith at Michigan State University in 1982, and the framework has barely changed since, because the underlying maths still holds.
The critical distinction is that menu engineering works on contribution margin, not food cost percentage. A dish with a 35% food cost that sells twice a night contributes less real money than a dish with a 45% food cost that sells forty times. Percentages don't pay rent; margin multiplied by volume does.
The output of menu engineering is never “raise prices”. It is a per-item decision: promote, reprice, rework or retire.
The four categories of menu engineering
Plot every item against average popularity and average contribution margin and you get four quadrants. Each one has a different correct answer.
Stars
High profit · High popularity
Your best dishes: guests order them and they carry real margin. They usually define what people think your restaurant is.
What to do
Protect them. Keep quality and portioning absolutely consistent, give them the strongest position on the menu, and do not discount them. Test small price increases here first — demand is least elastic.
Plow Horses
Low profit · High popularity
Guests love them but they barely contribute. Usually the dish that built the business and has never been repriced while costs moved.
What to do
Attack the cost, not the popularity. Re-engineer portion and garnish, renegotiate the key ingredient, or pair the item with a high-margin side or drink. Raise the price in small steps and watch volume.
Puzzles
High profit · Low popularity
Profitable but nobody finds them. The problem is usually naming, placement or description rather than the food itself.
What to do
Give them visibility: move them into the eye path, add a descriptive name and a short sensory description, brief the floor staff to recommend them. If two rounds of promotion don't move volume, they become Dogs.
Dogs
Low profit · Low popularity
Neither ordered nor profitable. They still cost you: stock lines, prep time, waste and menu space.
What to do
Remove most of them — but check first whether the item is a signature dish, a dietary requirement (vegan, gluten-free) or a small but loyal regular's order. Removing those costs more than it saves.
How to run it: the formula, step by step
You need one clean period of sales data and an accurate cost per item. Everything else is arithmetic.
Choose a clean data period
Take at least four full weeks so every weekday is represented. Avoid periods distorted by a holiday, a closure or a big promotion — those tell you about the event, not the menu.
Pull units sold per item
Export item-level quantities from your POS for the period. Analyse each menu category separately: starters compete with starters, not with mains.
Calculate menu mix %
This is the item's share of units sold within its own category.
Menu mix % = item units sold ÷ total units sold in category × 100Calculate contribution margin per item
Selling price minus the item's food cost. This is the money the dish actually leaves on the table.
Contribution margin = selling price − food costSet the two thresholds
The popularity threshold is the classic 70% rule; the profitability threshold is the weighted average contribution margin of the category.
Popularity threshold = (100 ÷ number of items) × 0.70Place each item and act
Above both thresholds is a Star; above popularity only is a Plow Horse; above margin only is a Puzzle; below both is a Dog. Then apply that quadrant's action, change one variable at a time, and re-run the analysis after a full period.
Worked example
Four items from a single category. The category sold 1,000 units, so the popularity threshold is (100 ÷ 4) × 0.70 = 17.5%, and the weighted average contribution margin is $8.71.
| Item | Units sold | Menu mix % | Contribution margin | Above average? | Category |
|---|---|---|---|---|---|
| Grilled sea bass | 310 | 31.0% | $12.20 | Popularity ✓ · Margin ✓ | Star |
| Cheeseburger | 420 | 42.0% | $5.60 | Popularity ✓ · Margin ✗ | Plow Horse |
| Lamb shank | 120 | 12.0% | $13.80 | Popularity ✗ · Margin ✓ | Puzzle |
| Chicken salad | 150 | 15.0% | $6.10 | Popularity ✗ · Margin ✗ | Dog |
The cheeseburger is the interesting one: the highest-volume item in the category and the second-lowest contributor. A $1.50 increase adds roughly $630 per period at unchanged volume. That single change is worth more than any redesign of the two low-volume items — which is why menu engineering starts with Plow Horses, not with deletions.
Pricing and layout: the part guests actually see
The analysis tells you what to change. Menu design decides whether the change lands.
- Drop the currency symbol
Menus that print 14 instead of $14 consistently show higher average spend in restaurant pricing studies. The symbol is a cue to think about money rather than food.
- Avoid a straight price column
A right-aligned column of prices invites guests to scan by price and order the cheapest thing. Put the price inline, immediately after the description, in the same weight as the body text.
- Use an anchor item
One deliberately expensive item makes everything below it read as reasonable. It does not need to sell; it needs to be there.
- Write descriptions for the items you want to sell
A short sensory description — origin, method, texture — measurably lifts selection. Reserve it for Stars and Puzzles; describing everything removes the signal.
- Respect the eye path
On a single-panel menu, attention lands top-right, then top-left, then centre. Put Puzzles where the eye lands first, not at the bottom of a long list.
Common mistakes
- Analysing too little data
A single week is noise. Weather, a local event or one large party will move a small sample far enough to misclassify items.
- Using stale food costs
The analysis is only as good as the cost side. If ingredient prices moved and your recipe costs did not, every margin in the model is wrong — and Plow Horses look like Stars.
- Deleting every Dog immediately
Check for signature dishes, dietary options and items with a small but loyal following. A menu with no vegetarian option loses whole tables, not one cover.
- Changing several variables at once
If you reprice, rename and move an item in the same round, you cannot tell which change worked. Change one thing, wait a full period, measure.
- Leaving the floor staff out
Servers decide what gets recommended. A Puzzle will not become a Star through menu layout alone if nobody at the table mentions it.
Running menu engineering from POS data
The analysis above is straightforward once — and painful every month. Units sold, recipe costs, category thresholds and the resulting matrix all shift as prices move and seasons turn, and a spreadsheet rebuilt by hand rarely survives past the second cycle.
robotPOS keeps the sales side and the cost side in the same system, so contribution margin per item is current rather than reconstructed. aiR intelligence then places every item on the matrix automatically, tracks how items move between quadrants over time, and flags combo and cross-sell opportunities — across a single venue or an entire chain.
Frequently asked questions
- What is menu engineering in restaurants?
- Menu engineering is the practice of analysing each menu item on two axes — how often it sells and how much contribution margin it produces — and then repricing, redesigning, reworking or removing items based on where they land. It was formalised at Michigan State University in 1982 and remains the standard framework for menu profitability.
- What are the four categories of menu engineering?
- Stars (high profit, high popularity), Plow Horses (low profit, high popularity), Puzzles (high profit, low popularity) and Dogs (low profit, low popularity). Each quadrant has a different correct action: protect Stars, attack the cost of Plow Horses, promote Puzzles, and remove most Dogs.
- What are the three essential components of menu engineering?
- Profitability analysis (contribution margin per item), popularity analysis (each item's share of units sold in its category), and menu design (placement, naming and description that steer selection). Skipping the third means the analysis never reaches the guest.
- How do you calculate menu engineering?
- Contribution margin = selling price − food cost. Menu mix % = item units ÷ category units × 100. The popularity threshold is (100 ÷ number of items in the category) × 0.70, and the profitability threshold is the weighted average contribution margin. Items above both thresholds are Stars.
- How often should you run menu engineering?
- Quarterly for a settled menu, plus any time input costs change significantly. Run it more often and you will be reading noise; less often and you will carry underpriced Plow Horses through an entire cost cycle.
- Is menu engineering the same as food cost percentage?
- No, and confusing the two is the most common error. Food cost percentage measures efficiency on a single dish; contribution margin measures the actual money a dish leaves after ingredients. A low-percentage item that rarely sells contributes less than a high-percentage item that sells constantly.
