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

A family-run Antep-style restaurant in old Gaziantep had a 47-item menu and the owner kept asking why margins felt thin. We ran the 1982 Kasavana-Smith menu engineering matrix on 90 days of POS data. The verdict: 6 stars, 12 plowhorses, 8 puzzles, 21 dogs. Trimming the dogs alone added +18% to monthly profit.

The Four Quadrants Explained

The matrix plots each menu item on two axes: popularity (units sold) and profitability (contribution margin). Every dish lands in one of four cells:

• Stars: high popularity, high margin. Your hero items — Ali Nazik, lahmacun, lamb tandır in our case.
• Plowhorses: high popularity, low margin. They sell, but the margin per plate is thin. Think rice-and-beans plates, ayran, tea.
• Puzzles: low popularity, high margin. Great profit if customers ordered them — they don't.
• Dogs: low popularity, low margin. Why are they still on the menu?

The threshold for "high" is typically 70% of average sales (popularity) and the menu's average contribution margin (profitability).

How to Build the Matrix

You need 4 weeks of sales data and each item's food cost. Contribution margin = sell price − food cost. Example: a 180 TL Ali Nazik with 52 TL food cost has 128 TL margin. Compare each item's unit sales to the average and its margin to the menu average.

Spreadsheet works in 10 minutes. thMenu's Pro analytics panel auto-generates the matrix from 90-day sales data — one click and you see the four quadrants with each dish positioned visually.

Action per Quadrant

Each quadrant gets a different play. Stars: protect, feature prominently, test small price increases (5–8%). Plowhorses: shrink food cost via portion control or cheaper sides, push upsells. Puzzles: reposition them — feature higher on the menu, train servers to recommend, add photos, sometimes rename. Dogs: remove or reformulate. If 20% of your menu is dogs, prep waste is killing the kitchen.

In Gaziantep we cut 16 of 21 dogs outright, reformulated 3 (smaller portions, repriced), and kept 2 for cultural/sentimental reasons. Food waste dropped 23% the next month, prep time per service shrank by 1.5 hours, and the dining room ran smoother because servers stopped recommending obscure items nobody really wanted.

FAQ

How often should I rerun the matrix? Every 90 days, plus an extra pass at season changes. Tourist-heavy spots may need monthly checks during peak.

How do I classify brand-new dishes? They need at least 30 days of sales data to qualify. Mark them as "unranked" until then, then run the standard rules.

Should drinks and desserts share the matrix with mains? No — run them separately. Drink margins dwarf food margins, and combining them skews the average and produces bad calls.

Found this helpful? Share it.