Menu Engineering Explained: Stars, Plowhorses, Puzzles and Dogs

The problem with looking at food cost alone

Food cost percentage is the number most operators reach for when judging a dish, and on its own it is close to useless for menu decisions. It measures cost as a share of price and says nothing whatsoever about how often the dish sells.

The consequence is a predictable mistake. A pasta with a $2.10 plate cost sold at $14.00 runs a 15% food cost, which looks excellent. A steak with a $9.00 plate cost sold at $30.00 runs 30%, which looks mediocre. Judged on percentage the pasta wins comfortably. Judged on what actually reaches your bank account, the steak returns $21.00 per plate against the pasta's $11.90.

Menu engineering fixes this by ignoring percentages entirely and working with two things that genuinely matter: how much money a dish makes per sale, and how many times it sells.

The two measurements

Contribution margin

Menu price minus plate cost. The dollar amount each sale contributes toward rent, wages and profit. A dish is classified as high or low margin relative to the average across your own menu, not against any universal benchmark — which is why the method works equally well in a cafe and a fine dining room.

Popularity

The dish's share of total covers, compared against an adjusted equal share. Divide 100 by the number of menu items, then take 70% of that figure. On a sixteen-item menu the threshold is 4.4% of covers.

The 70% adjustment matters. Real menus are never evenly distributed — a few dishes always carry a disproportionate share — and holding every item to a perfectly equal share would classify most of the menu as a failure.

The four categories

Stars: popular and profitable

The dishes doing the most for you. The main risk with a Star is well-intentioned damage: because it is popular, it becomes the obvious candidate for a promotional discount, which means discounting your most profitable dish. Protect the recipe specification instead, since portion drift here costs more than anywhere else on the menu.

Plowhorses: popular, low margin

Almost always where the largest recoverable sum sits. Demand is already proven, which means price sensitivity is usually lower than operators fear. A modest increase typically costs very little volume, and because the volume is high the margin gain is substantial.

The alternative is reducing plate cost through substitution, portion adjustment or better yield. Which route to take depends on whether the dish's appeal rests on its size, its ingredients, or simply its familiarity.

Puzzles: profitable, unpopular

These are worth real effort before deletion. Ask why nobody orders it. Is it buried at the bottom of a section, described poorly, unfamiliar, or priced above what guests expect from that part of the menu? Position and description shift ordering behaviour more than most operators believe, and converting a Puzzle into a Star is the highest-value outcome this exercise can produce.

Dogs: unpopular and unprofitable

Usually remove them. The gain exceeds the dish's own numbers, because a shorter menu means less inventory, less prep, less waste on slow-moving ingredients, and an easier read for the guest. The exception is a dish serving a real need — a vegetarian, vegan or allergen-safe option that allows a whole table to book.

A worked example

A sixteen-item menu sold 2,400 dishes last month with an average contribution margin of $11.20. The popularity threshold is 4.4%.

The burger sells 420 units — 17.5% of covers, comfortably the menu's engine — at $16.00 against a $6.20 plate cost. Its $9.80 margin sits below the menu average, making it a Plowhorse. Raising the price to $17.50 lifts margin to $11.30 and, across 420 units, produces roughly $630 of additional monthly contribution from a single change to a dish guests already choose.

The sea bass earns $18.40 per plate, nearly matching the ribeye, but reaches only 2.9% of covers. That is a Puzzle. Moving it to the top of its section and rewriting the description costs nothing and could convert the menu's second-best margin into a volume seller.

The chicken salad at $8.90 margin and 2.6% of covers is a Dog. Unless it is the menu's only light option, it is occupying space and inventory that something else could use.

Where the exercise goes wrong

• Inaccurate plate costs. The single biggest failure. If yield percentages are not applied, margins on protein dishes are overstated and the classification is wrong from the start.
• Mixing menu sections. Starters, mains and desserts must be run separately. A dessert will rarely clear a threshold set by main courses.
• Too short a period. A single week misclassifies anything seasonal. Use a month at minimum.
• Treating the grid as a verdict. The categories indicate what kind of action a dish needs, not how urgent it is.
• Acting on every dish at once. Changing a whole menu simultaneously makes it impossible to learn what worked.

The number that decides priority

The grid tells you what kind of intervention each dish needs. Total contribution — margin multiplied by units sold — tells you which to do first.

In the example, the ribeye is a Star contributing $19.20 across 180 covers, or $3,456. The burger is a Plowhorse contributing $9.80 across 420 covers, or $4,116. The Plowhorse matters more to the business despite the worse category, which is exactly why it is the first thing to fix.

Rank every dish by total contribution alongside its category and work down the list. Two or three changes made deliberately and measured properly will do more than a wholesale menu rewrite, and you will actually know which change produced the result.

Frequently asked questions

What are the four menu engineering categories?

Stars are popular and high margin, Plowhorses are popular and low margin, Puzzles are unpopular and high margin, and Dogs are unpopular and low margin.

How do I calculate contribution margin?

Subtract the plate cost from the menu price. It is the dollar amount each sale contributes toward fixed costs and profit, and it is measured in dollars rather than percent for exactly that reason.

Why 70% of the equal share for popularity?

Because real menus are never evenly distributed. A handful of dishes always take a disproportionate share, so an unadjusted equal-share threshold would classify most of the menu as unpopular.

What is the fastest win from menu engineering?

Repricing Plowhorses. Demand is proven, so a modest increase usually costs very little volume while improving margin across high unit numbers.

Should starters and mains be analysed together?

No. Run each menu section separately, since a starter or dessert will rarely reach a popularity threshold calculated across main courses.

How often should I redo menu engineering?

Quarterly, and after any menu change or significant supplier price movement. Categories move as costs and preferences shift.

Run the numbers

Use the free Menu Engineering Calculator to apply everything above to your own figures.