Menu Engineering Matrix: Stars, Puzzles, Plowhorses, Dogs
Every menu item is doing one of four jobs: making you money, making you famous, making you busy, or making you tired. The menu engineering matrix is a 40-year-old tool that tells you which is which, using two numbers you already have.
The framework comes from Michael Kasavana and Donald Smith’s 1982 work at Michigan State University, and it remains the standard because it is simple: plot every item by how often it sells and how much profit it contributes per sale. Four quadrants fall out, and each quadrant has a clear playbook.
This article is the deep dive on the matrix itself. For where it fits in the larger process, from costing plates to shipping the redesigned menu, see the full menu design and engineering guide.
The two axes
Popularity is units sold, compared to other items in the same category. Compare entrees to entrees and apps to apps, never a $4 side to a $30 steak. The classic threshold: an item counts as popular if it sells at least 70% of the category average. In a 10-item entree list, category share above 7% is popular; below it is not.
Contribution margin is menu price minus plate cost, in dollars. An item is high-margin if its contribution beats the average contribution of its category. Dollars, not food cost percentage: a 40% food cost item that leaves $15 behind outranks a 25% item that leaves $8. If your plate costs are guesses, fix that first with our food cost percentage method and calculator.
The four quadrants
Stars: popular and profitable
Your signature items. The playbook is protection: keep quality rigidly consistent, never substitute a cheaper ingredient quietly, give them the best positions on the page (first or last in their section), and test small price increases. Stars usually have pricing power; a $1 move on a star often goes unnoticed.
Plowhorses: popular but thin
The burger everyone orders that leaves $6 behind. You cannot cut it, guests would riot, so you re-engineer it. Three levers, in order of preference:
- Trim plate cost. Renegotiate the protein, bake instead of buy the bun, cut the garnish nobody eats. Saving $0.60 on a dish that sells 400 times a month is $2,880 a year.
- Adjust the build. Slightly smaller portion of the expensive component, more of the cheap one. Do this honestly; guests notice a hollowed-out dish.
- Nudge the price. $0.50 to $1.00, paired with a better description so the perceived value moves with the price. How you communicate that matters; see how to announce price increases gracefully.
Puzzles: profitable but ignored
High margin, low sales. The dish is fine; the presentation on the page is failing. Before touching the recipe, try marketing fixes: rename it (specific beats clever), rewrite the description with provenance and preparation detail, move it to the first or last slot in its category, give it the one highlight box on the page, or have servers recommend it by name. If three months of promotion does not move it, it may be a dog wearing a good margin.
Dogs: unpopular and unprofitable
Every dog costs you inventory, prep time, training complexity, and menu real estate. Cut them. The two legitimate exceptions: an item that exists for a strategic reason (the one vegan entree, the kids meal) and an item whose ingredients are fully shared with winners, making its true carrying cost near zero. Otherwise, deleting dogs is the easiest money in the building.
A worked example
Eight entrees, 90 days of sales, category average contribution of $10.90 and average units of 344 (popularity threshold at 70% of average = 241 units):
| Item | Units sold | Price | Plate cost | Margin | Class |
|---|---|---|---|---|---|
| Smash burger | 612 | $15 | $4.90 | $10.10 | Plowhorse |
| Fried chicken sandwich | 548 | $16 | $4.30 | $11.70 | Star |
| Rigatoni vodka | 460 | $19 | $4.10 | $14.90 | Star |
| Fish and chips | 291 | $18 | $7.40 | $10.60 | Plowhorse |
| Hanger steak frites | 244 | $28 | $11.60 | $16.40 | Star |
| Seared salmon | 196 | $26 | $9.80 | $16.20 | Puzzle |
| Mushroom risotto | 214 | $21 | $5.60 | $15.40 | Puzzle |
| Turkey club | 187 | $14 | $5.20 | $8.80 | Dog |
The moves practically write themselves: feature the rigatoni harder (star with a monster margin), take the smash burger from $15 to $15.50 and shave the plate cost, give the risotto the box and a rewrite, and let the turkey club go.
Note what the percentages would have hidden: fish and chips runs a scary 41% food cost but still clears $10.60 a plate. It stays.
Running it in practice
Cadence: quarterly, or after any supplier price shock. The matrix is a snapshot; ingredient inflation redraws it constantly. A star from March can be a plowhorse by October without a single recipe change.
Category by category: run separate matrices for appetizers, entrees, desserts, and drinks. Blending categories makes everything cheap look like a dog and everything expensive look like a puzzle.
Sample size: 60 to 90 days of sales smooths out weather, holidays, and that one big catering order. Two weeks is not enough, and a single blowout weekend can crown the wrong star.
Don’t over-precision it: you will find refinements of the matrix in hospitality textbooks, from David Pavesic’s cost-margin analysis to weighting schemes that fold labor into the margin axis. They are legitimate, and mostly unnecessary at independent scale. A simple two-axis sort on honest numbers, actually repeated every quarter, beats a sophisticated model you run once. The gains come from the decisions (cut, reprice, promote), not from decimal places.
Watch the interactions: some items exist to sell other items. A cheap, popular appetizer that starts 40% of tables is doing margin work that never shows up in its own row. Look at the check level before cutting anything that anchors the start of a meal.
After the matrix: make the layout do the work
Classification is diagnosis, not treatment. The treatment happens in design: your stars and puzzles should occupy the first and last slots of each section, get the fullest descriptions, and receive the page’s single highlight. Pricing presentation matters too, from removing dollar signs to choosing price endings that match your concept; the evidence for each move is in menu pricing psychology.
And the layout only counts if guests see the current version. Once you cut the dogs and reprice the plowhorses, the menu on Google, the QR menu on the table, and the printed to-go menu all need to match. VisibleMenus handles that from one upload: a hosted QR menu, a printable PDF, and a Google menu that stays in sync for $18 one time, then $6/month.
Next up in the series: food cost percentage, the number that feeds the margin axis, and the complete menu design and engineering guide if you want the full process end to end.
