Menu Engineering Matrix: Stars, Plowhorses, Puzzles, and Dogs

A menu engineering matrix compares how many units each dish sells with the contribution it earns per sale. The result is a practical first pass: stars score high on both, plowhorses sell often but earn less per plate, puzzles earn more but sell slowly, and dogs lag on both. Use it to choose what to inspect or test—not as an automatic menu purge.

Start with one menu category and one sales period, using units sold from your POS item-sales report. For the costing steps behind menu price and plate cost, see the complete guide to restaurant food costing and menu pricing.

What does a menu engineering matrix measure?

The menu engineering matrix compares two measures for items in the same category: popularity, based on units sold, and contribution (often called contribution margin), the dollars left from each sale after food cost. Comparing dishes within a category—such as mains with mains—keeps a high-volume drink from skewing the result for dinner entrées. Use the same sales period and cost basis for every item, too.

For each dish, calculate its contribution as menu price minus plate food cost. Then divide the item’s unit sales by all units sold in that category to get its menu mix. The guide to food cost percentage versus gross margin explains why the percentage alone can hide how many dollars a sale contributes.

This is the classic food-menu definition, and it is a simplified measure, not full item profit. It excludes labor, rent, and other operating expenses, and it leaves out other per-order selling costs. If packaging, delivery commissions, or payment fees vary by item or channel, calculate them separately. For discounted sales, use the revenue the restaurant actually keeps rather than treating the discount as a separate cost.

How are stars, plowhorses, puzzles, and dogs classified?

Set two cutoffs for the category and period: a popularity threshold and an average contribution. Michael Kasavana’s AHLEI explanation of menu engineering uses 70% of an equal share for popularity and compares each item’s contribution with the sales-weighted average.

Item contribution = menu price − plate food cost
Menu mix % = item units sold ÷ category units sold × 100
Popularity threshold = 70% ÷ number of items in the category
Total contribution dollars = sum of (item contribution × units sold)
Average contribution = total contribution dollars ÷ category units sold

An item counts as high popularity when its menu mix is at or above the popularity threshold, and as high contribution when its contribution is at or above the category average.

Here is a four-item example for one period. The four dishes sold 200 units in total, so the popularity threshold is 70% ÷ 4 = 17.5%, or 35 units. Their total contribution is ($10 × 100) + ($11 × 60) + ($14 × 25) + ($7 × 15) = $2,115, making the weighted average $2,115 ÷ 200 = $10.575, or about $10.58 per sale.

The chicken rice bowl is both popular and above the $10.58 contribution line, so it is a star. The burger is popular but below that line, making it a plowhorse. The salmon earns more than average but falls below 35 sales, so it is a puzzle. The salad is below both cutoffs, so it is a dog.

The burger misses the line by only 58 cents, but closing that gap takes more than 58 cents. Because it makes up half of all sales, every gain in its contribution also lifts the average. It would need about $11.15 per sale to reach the line—and at that point the chicken rice bowl, still at $11, would drop below the new average and become a plowhorse. Items near a cutoff can change class without anything happening to their own recipe, so check them against a second period before acting.

What should a restaurant do with each quadrant?

Treat each label as a starting point for investigation, not as a decision. The next move depends on recipe accuracy, guest expectations, service capacity, and the role the dish plays in the wider menu.

• Stars: protect the recipe and availability, then give the dish clear menu visibility. Check customer response before raising its price; high sales alone do not show how guests will react to a change.
• Plowhorses: keep the demand while checking portion size, ingredient cost, waste, and price. A small recipe or price adjustment can matter on a dish that sells often, but test it against guest response.
• Puzzles: make the dish easier to notice or understand. Review its name, description, photo, position within its menu category, and staff recommendation before changing the recipe or cutting the price. The digital menu optimization guide covers presentation choices that can support this test.
• Dogs: check the data and the dish’s role before removing it. It may be a new item, a seasonal option, a dietary choice, or a price anchor that helps guests choose another dish. If it has no clear purpose, simplify, rework, or retire it.

What does the menu engineering matrix leave out?

The matrix gives a relative snapshot, not a complete operating picture. Changing the items, category boundaries, or sales period can move the cutoffs and reclassify dishes even when a recipe has not changed.

The matrix also treats each dish largely in isolation. Hospitality research has highlighted the limits of traditional menu analysis, including its narrow cost inputs and its omission of labor effort. Later work examined how menu-item substitutes affect pricing and placement decisions. A dish with strong contribution per plate may still slow the kitchen or draw sales away from another profitable choice. See the research on limitations in traditional menu analysis and menu-item substitutes.

For delivery sales, food cost alone also misses order-specific costs. Subtract channel costs such as packaging and commissions, and use the revenue kept after any discount, before treating the result as item profit; the delivery-order profit guide shows that calculation.

When should a restaurant repeat menu engineering?

Repeat the analysis when the business has a fresh, comparable sales period and current recipe costs. Recheck after a meaningful price, portion, supplier, promotion, or menu change, and compare similar days or seasons so a short-term event does not decide a permanent menu change.

Save the inputs beside the results—item price, plate cost, units sold, menu category, any channel-specific costs, and the dates of the POS export and recipe costs—so the next comparison uses the same basis. Then change one thing at a time—such as a description, position, recipe cost, or price—and review the next comparable period. Once the choices are clear, a digital menu can keep published item details current without reprinting the QR destination.

A useful menu matrix ends with a measured test, not a label.