How the matrix works
Menu engineering, introduced by Michael Kasavana and Donald Smith in 1982, judges every item on two axes and ignores everything else:
- Popularity: the item's share of units sold in its category (its menu mix). The threshold is 70% of the average share: with 8 items, average is 12.5%, so anything selling 8.75% or more of units is "high". The 70% cushion stops a menu from having half its items flagged unpopular by definition.
- Profitability: the item's contribution margin (price − plate cost) compared with the sales-weighted average margin across the category. At or above average is "high".
⭐ Star: popular + high margin
- Protect it: consistency, prominent placement
- Test small price increases
- Never take it off
🐴 Plowhorse: popular + low margin
- Raise price in small steps
- Cut plate cost (portion, garnish, sub-recipe)
- Pair with a high-margin add-on
🧩 Puzzle: high margin + unpopular
- Rename, re-describe, reposition on the menu
- Server recommendation, feature it
- Consider a price cut if margin allows
🐕 Dog: unpopular + low margin
- Remove, unless it serves a purpose
- Or reprice/re-engineer into a Puzzle
- Frees prep time and inventory
Why margin, not food cost percentage
This is the part that surprises people who've been managing to a food cost target. A ribeye at 40% food cost that contributes $22 a plate makes more money than a pasta at 22% that contributes $9. Percentage is a control ratio for the kitchen; contribution margin is what pays rent. Menu engineering ranks by margin on purpose, and the usual result is that one or two of your "high food cost" items turn out to be Stars you'd been thinking about cutting.
Running it well
- One category at a time. Entrées against entrées, apps against apps. Mixing a $6 side with a $32 steak makes the average meaningless.
- Use a real period. A month is the minimum; a quarter smooths specials and weather.
- Re-cost first. Plate costs from last year's prices will misclassify. USDA expects food inflation around 3% again in 2026, and individual items swing far more.
- Act on two items per quarter. Reprice one Plowhorse, reposition one Puzzle, and measure. Wholesale menu changes are hard to read.
- Re-run after every change. Prices and mix move; the classification is a snapshot.
Beyond the four boxes
The matrix is deliberately simple, and it has known blind spots: it ignores labor (a Star that takes 20 minutes of line time may be less profitable than the numbers say), it ignores what an item pulls along (a low-margin burger that sells a high-margin beer), and it treats "average" as the bar when your average may itself be too low. Use it to decide where to look, then check the prime cost and the price ladder before acting.
Need plate costs first? Cost a recipe →
Frequently asked questions
What is menu engineering?
Sorting items by popularity (menu mix) and profitability (contribution margin) into Stars, Plowhorses, Puzzles and Dogs, from Kasavana and Smith, 1982.
How is "popular" decided?
Selling at least 70% of the average share: (1 ÷ items) × 0.7.
How is "profitable" decided?
Contribution margin at or above the sales-weighted category average.
What do I do with a Plowhorse?
Nudge price up, trim plate cost, pair with a high-margin add-on. Keep it.
Should I remove Dogs?
Usually, unless they serve a purpose (kids, dietary, a regular's order).
How often should I run it?
Quarterly, and after any price or recipe change.
Sources
- Menu engineering (Kasavana & Smith, Menu Engineering: A Practical Guide to Menu Analysis, 1982).
- Ivancsóné Horváth et al. Examining a menu on the basis of the Kasavana–Smith model. Deturope 14(1), 2022 (70% popularity threshold, quadrant definitions).
- Egan, Introduction to Food Production and Service (Penn State): contribution margin.
- USDA ERS: Food Price Outlook.