Restaurant menu engineering: optimising your menu profitability (2026)

The short version. Menu engineering is a method for analysing your menu along two axes: a dish's popularity and its contribution margin. You sort each dish into one of 4 categories (star, plowhorse, puzzle, dog), then act. Documented result: 3 to 7% additional revenue in 6 months, without chasing a single new customer.

Context / Definition

Menu engineering was born in the 1980s in American hospitality management schools. In Europe it remains underused — most operators set their menu on instinct, on feel, or because "the chef likes making this dish". Result: menus draining margin without anyone really knowing. To go further on the levers of complete restaurant profitability, this is a solid starting point.

What is the menu engineering matrix?

The matrix is simple: two axes, four boxes. On the x-axis, popularity (does it sell?). On the y-axis, contribution margin (does it pay?).

You get four categories. Each calls for a different decision.

The 4 dish categories

The matrix itself doesn't take 10 minutes to fill in if you know your numbers. The problem is that most restaurants don't know their precise per-dish food cost in real time. They estimate. And that's where everything derails. (Names from the original Kasavana-Smith framework.)

How to build your matrix concretely

Steps in order. No shortcuts.

1. Pull your sales data over a representative period (minimum 4 weeks, ideally 3 months). Number of covers sold per dish, per category.
2. Calculate contribution margin per dish: net selling price minus actual food cost (not theoretical — actual, with current supplier prices).
3. Set your popularity threshold: average sales over the period. A dish selling below average = low popularity.
4. Set your margin threshold: average contribution margin across the menu. Below = low margin.
5. Place every dish in the matrix using these two thresholds.
6. Decide for each category (see table above).

Update your recipe cards and reformulation the moment you touch a recipe. If you don't, you're working with bad numbers for the following weeks.

Case study — Lunch Wagon, 2024

In 2024, the Lunch Wagon's signature burger was what's called a star in popularity: the most ordered item on the truck, the one customers came for. Problem: its real food cost was at 42%. On every burger sold, 42 cents of every euro went to ingredients. Contribution margin was disastrous.

The natural temptation is to leave it alone. It sells, customers love it, don't touch. That's the classic mistake. The dish you love selling can be the very one eating your margin.

I did the opposite. Full ingredient redesign: changed beef supplier (equivalent quality, different sourcing), reduced sauce gram weight by 30g, removed a decorative garnish that added nothing on the palate. Result: food cost down to 29%. Popularity intact — customers noticed nothing. Over 6 months, that single decision delivered +4% overall profitability on the truck.

No magic. Just control. To understand how to track KPIs to drive menu decisions, it's the same logic: don't decide without data.

Reference table — actions by category

The puzzle deserves special attention. A high-margin dish that doesn't sell is often a readability problem, not a recipe problem. Bad placement on the menu, a name that doesn't land, cold copy. Before dropping it, try repositioning it visually. That lever is free.

Common mistakes

• Rebuilding the menu without sales data: if you don't know what actually sells, you're working blind. Four weeks of records is enough to have a solid base.
• Touching prices only, not recipes: raising the price of a plowhorse without reformulating risks turning it into a dog if customers don't follow.
• Treating every dish with the same logic: a starter, a main and a dessert don't have the same contribution margin thresholds. Run the matrix separately by dish category.
• Not updating recipe cards after reformulation: your food cost numbers go stale immediately. To cut overall food cost, live data is non-negotiable.
• Doing the exercise once and never returning: supplier prices move. A 2026 star can become a plowhorse in 2027 if you don't watch sourcing.

Conclusion

Three takeaways.

First: menu engineering isn't a menu overhaul. It's a data read followed by targeted decisions on specific dishes. You don't need to change everything — often 3 or 4 decisions are enough.

Second: the best-selling dish isn't necessarily your ally. An uncontrolled food cost on a star is your worst enemy. Reformulating without the customer noticing is a skill worth developing.

Third: it only works with live numbers. Real per-dish food cost, sales week by week. Not estimates, not last year's averages. Daily control is the prerequisite.

Last updated 2026. Written by Cyril Quesnel, founder of Onrush, chef-entrepreneur (La Verrerie 2015-2018, Lunch Wagon 2023-2026).