Menu engineering sorts every dish on your menu into one of four categories, stars, plowhorses, puzzles, or dogs, based on popularity and profit, so you know exactly which items to promote, reprice, or cut. Restaurants that run this analysis at least twice a year typically find 15-25% of their menu is quietly underperforming without anyone noticing during service.
Most menus grow item by item over years, based on what a chef wants to cook, what a supplier is pushing, or what worked at another restaurant. Few menus get rebuilt based on what’s actually selling and what’s actually profitable. Menu engineering fixes that by giving every item a place on a simple two-axis matrix, using data you already have in your point-of-sale system.
Key takeaways
- The matrix: every item sorts into stars, plowhorses, puzzles, or dogs based on popularity and profit
- Framework origin: Michael Kasavana and Donald Smith, Michigan State University, 1982
- What it needs: per-item sales counts and food cost, both already in most POS systems
- Realistic finding: 15-25% of most menus are underperforming items nobody has flagged
The four-box matrix explained
The matrix plots two things against each other: how often an item sells (popularity) and how much profit each sale generates (contribution margin, not the item’s price).
| Category | Popularity | Profitability | What it means |
|---|---|---|---|
| Stars | High | High | Your best items. Protect these, don’t mess with the recipe or price carelessly |
| Plowhorses | High | Low | Popular but thin margin. Guests love them, but they’re not earning their keep |
| Puzzles | Low | High | Profitable when ordered, but too few guests order them |
| Dogs | Low | Low | Neither popular nor profitable. Candidates for removal or a full redesign |
Popularity threshold: an item counts as “high popularity” if it sells at or above its expected fair share. With 20 items on a menu, each item’s fair share is 5% of total sales (100% / 20 items). Many operators use 70% of fair share as the practical cutoff, so on a 20-item menu, anything selling above roughly 3.5% of total covers counts as popular.
Profitability threshold: an item counts as “high profitability” if its contribution margin sits at or above your menu’s average contribution margin across all items.
Worked example
Here’s a simplified six-item menu analysis to show the matrix in action:
| Item | Price | Food cost | Contribution margin | Units sold (month) | Category |
|---|---|---|---|---|---|
| Ribeye steak | $34 | $13 | $21 | 180 | Star |
| Grilled chicken | $19 | $6 | $13 | 320 | Plowhorse |
| Seafood pasta | $26 | $8 | $18 | 90 | Puzzle |
| Veggie burger | $16 | $5 | $11 | 60 | Dog |
| House salad | $12 | $3 | $9 | 210 | Plowhorse |
| Chef’s tasting plate | $29 | $9 | $20 | 45 | Puzzle |
Average contribution margin across this menu is roughly $15.30. Grilled chicken and house salad sell in high volume but sit below that average margin, making them plowhorses: guests love them, but they’re not carrying their weight financially. Seafood pasta and the chef’s tasting plate earn well above average margin but sell in low volume, making them puzzles worth promoting harder before cutting.
What to do with each category
Stars: protect and feature
Don’t change the recipe, don’t quietly raise the price too aggressively, and make sure these items stay visible on the menu. These are working exactly as intended.
Plowhorses: raise margin without losing volume
Plowhorses are popular for a reason, usually price or familiarity. Look for ways to improve margin without guests noticing: a slightly smaller portion paired with a better plate presentation, a cheaper cut that tastes the same after the right preparation, or a modest price increase justified by a menu description upgrade.
Puzzles: promote before you cut
A puzzle item might be underselling simply because it’s poorly placed on the menu, poorly described, or unfamiliar to guests. Try repositioning it in a more visible spot, adding a compelling description, or having servers mention it, before assuming it needs to go.
Dogs: redesign or remove, but check first
Some dogs exist for a reason: dietary accommodation, a loss-leader that drives other orders, or a seasonal item early in its run. Review each dog individually. If there’s no strategic reason for it to stay, removing a true dog simplifies your menu, reduces inventory complexity, and cuts prep time with minimal revenue impact.
How to build your own matrix
Step 1: Pull per-item sales data
Export unit sales for every menu item over a 60-90 day window from your POS system.
Step 2: Calculate contribution margin per item
Subtract each item’s food cost from its menu price. If you don’t have exact food cost per item, a reasonable estimate from your recipe costing is enough to start.
Step 3: Calculate your averages
Find the average popularity (fair share divided by number of items, adjusted by the 70% threshold) and average contribution margin across the full menu.
Step 4: Plot and categorize
Sort every item into stars, plowhorses, puzzles, or dogs based on where it falls against both averages.
Step 5: Assign an action to each category
Don’t stop at categorizing. Every item needs a specific next step: protect, reprice, promote, or reconsider.
Common mistakes to avoid
Using revenue instead of contribution margin
A high-priced item can look profitable by revenue while actually carrying a thin margin because its food cost is also high. Always use contribution margin (price minus cost), not price alone.
Running the analysis once and never again
Ingredient costs, guest preferences, and your menu itself all change. Rerun the matrix at least twice a year, or whenever you make a significant menu change.
Cutting every dog immediately
Some low-scoring items have a purpose beyond direct profit. Review each one before removing it.
Ignoring how items interact
A puzzle item that pairs naturally with a plowhorse (a wine pairing suggestion alongside the grilled chicken, for example) might be worth keeping even at low individual volume, because it lifts the check on orders that include it.
You need per-item sales data before any of this works
Menu engineering is only as good as the sales data behind it. Without accurate per-item counts and revenue, the matrix is guesswork. This is also the same underlying data that reveals your real covers and revenue patterns by service period, which matters for the food cost and profit margin work menu engineering feeds into.
Resos includes covers and revenue reports that surface the sales patterns you need to run this analysis with confidence, alongside your POS export.
Frequently Asked Questions
What is menu engineering?
What is the menu engineering four-box matrix?
Who invented the menu engineering matrix?
How do I calculate contribution margin for menu engineering?
Should I remove every dog item from my menu?
The bottom line
Menu engineering replaces gut feel with a straightforward two-axis analysis: popularity and profit. Most menus haven’t been reviewed this way in years, and most operators find real opportunities in the plowhorse and puzzle categories once they run the numbers. Build the matrix from 60-90 days of real sales data, assign a specific action to every item, and rerun it twice a year.
A visual diagram of the four-box matrix is planned as a follow-up to this guide.
Related guides: Restaurant KPIs | Food cost percentage | Restaurant profit margin