The Menu Engineering Matrix Is Only Half the Job
Answer first: the classic menu engineering matrix sorts every item by popularity and contribution margin into four categories — stars, plowhorses, puzzles and dogs. It is a useful diagnostic and an incomplete instruction. It tells you which items make money and which don't. It does not tell you whether your kitchen can execute them, what they cost you in labor and complexity, or where a guest's eye actually lands on the page. A matrix built on margin alone will hand you a menu that is profitable on a spreadsheet and unexecutable on a Friday.
Menu engineering, done completely, has three layers: the math, the operation, and the page. Most operators do the first, skip the second, and outsource the third to a designer.
How does the menu engineering matrix work?
Two axes, four boxes.
High popularity | Low popularity | |
High contribution margin | Stars — protect and feature | Puzzles — reposition or reprice |
Low contribution margin | Plowhorses — re-engineer the cost | Dogs — remove or rebuild |
Contribution margin = menu price − plate cost. Note that it is a dollar figure, not a percentage. This distinction is the single most consequential detail in menu engineering, and it is where most operators go wrong.
A dish selling at $18 with a 28% food cost contributes $12.96. A dish selling at $46 with a 38% food cost contributes $28.52. The second item has the worse percentage and puts more than twice the money in the bank per cover. You deposit dollars, not percentages.
Percentage thinking pushes operators toward cheap, high-margin, low-check items and away from the premium plates that actually pay the rent. It is the reason so many menus are simultaneously "well costed" and unprofitable.
Why isn't the matrix enough?
Because it treats every item as equally executable, and no kitchen works that way.
The matrix has no variable for:
Station load. Two stars on the same station will collide at 8:15 on a Saturday. The matrix cannot see the collision; the expo can.
Prep hours. An item can carry an excellent contribution margin and consume four hours of skilled labor before service. That labor is real cost, and it never enters the plate cost.
Ingredient overlap. An item that shares no ingredients with anything else on the menu carries the full weight of its own purchasing, storage, rotation and spoilage. Its true cost is higher than its recipe.
Skill dependency. A dish that only one cook can execute is not a menu item. It is a scheduling constraint.
Waste exposure. Fresh, unshared, short-shelf-life product priced against forecast covers — the matrix sees the margin, not the risk.
This is why the second layer exists.
What is the operational layer of menu engineering?
Score every candidate item against the kitchen you actually have — before you decide what stays.
Dimension | Question | Why it matters |
Station | Which station fires this, and what else does that station carry? | Balances the line under pressure |
Prep | How many labor minutes before service? By whom? | Uncosted labor is where margin disappears |
Overlap | How many other items use these ingredients? | Drives purchasing efficiency and waste |
Skill | How many cooks on the current roster can execute it correctly? | Determines whether it survives a call-out |
Hold | Does quality survive a ticket backup? | Consistency on the worst night, not the best |
An item that scores well on margin and badly across this table is not a star. It is a liability with good numbers, and it will produce inconsistency long before it produces profit.
The reverse is also true, and more interesting: an item with a merely adequate margin that shares ingredients, balances the line, holds well and any cook can execute is worth more to the operation than its contribution margin suggests. It stabilizes every other item on the menu. Cheese courses, well-built salads, braises and one-pan dishes usually live here.
How should the menu be laid out?
The third layer is the page, and it is not decoration — it is where the mix you engineered either happens or doesn't.
Principles that hold across formats:
Fewer items, better executed. Every line you add is purchasing complexity, prep load, training time and waste exposure. Short menus are not a creative limitation. They are an operational strategy.
Position your stars where the eye lands. The top of a section and the first item of a category carry disproportionate weight. Give those positions to items you want ordered, not to items tradition placed there.
Break the price column. A right-aligned column of prices invites guests to read down the cheapest path. Set prices immediately after the description, same size, no currency symbol, no leading dots.
Use description length as emphasis. A longer, more specific description raises perceived value and order rate. Spend that length on stars and repositioned puzzles, not evenly across the page.
Limit choice per category. Beyond roughly seven items in a section, decision fatigue pushes guests toward the familiar — which is usually your plowhorse.
None of this works if the underlying mix is wrong. Design amplifies engineering. It does not substitute for it.
How often should a menu be re-engineered?
Read the mix monthly. Re-engineer quarterly. Redesign the physical menu once or twice a year, or when the product changes.
Monthly reading matters because menu mix drifts silently. Guests migrate toward items that are easy to order, familiar, or simply positioned well — and food cost rises without a single recipe changing. By the time an annual review catches it, you have paid for three quarters of drift.
What to look at, in order:
Mix percentage by item — is anything above 15% of covers? Is anything below 2%?
Contribution margin in dollars, ranked — not percentage
Total contribution by item — margin × covers. This is the real ranking, and it reorders the list every time
Station distribution of the top ten — is the line balanced or is one station carrying the restaurant?
The fourth question is the one almost nobody asks, and it is the difference between a menu that is profitable and a menu that is profitable and survives a full house.
The structural point
Menu engineering is often sold as a pricing exercise. It isn't. A menu is a financial document, a production plan and a piece of communication at the same time, and it has to be right in all three registers simultaneously.
The most profitable move available to many restaurants this quarter costs nothing and takes a week: cut the menu, rebalance the line, and put the stars where guests will actually see them.
Frequently asked questions
What is contribution margin in menu engineering?
Menu price minus plate cost, expressed in dollars. It measures what each cover deposits, which is what pays fixed costs. Food cost percentage measures efficiency, which is a different and lesser question.
Should I remove all my dogs?
Not automatically. Some low-margin, low-popularity items exist for strategic reasons — a vegetarian option, a children's dish, an ingredient anchor that supports other items. Remove the dogs that serve no structural purpose; rebuild the ones that do.
How many items should a menu have?
Fewer than you currently have, in almost every case. The operational answer is: as many as your worst-staffed station can execute correctly on your busiest night. That number is usually well under thirty across all categories in full service.
Does menu engineering work for hotel outlets?
Yes, with an added constraint — hotel outlets share labor and inventory across breakfast, banquet and à la carte. The engineering has to run per outlet with costs separated, or the averages hide everything worth seeing.
Will raising prices fix a low contribution margin?
Partly, and only if the mix is engineered first. A uniform price increase across an unengineered menu raises the price of items guests already over-order at low margin, which can accelerate the drift rather than correct it.
How long does a menu engineering project take?
Reading the mix and producing the matrix takes days. Re-engineering the operation around it — station balance, prep, recipe standards, training, new costing — typically takes three to four weeks before the new menu goes live.
Paul Forciniti is a restaurant and hotel operations consultant with 25+ years in professional kitchens — line cook in Buenos Aires, culinary training in France, executive chef of a five-star international hotel, owner-operator, and author of Your Food Is Not the Problem and From Concept to Cover.
Want your menu read as a financial document? See how menu development works, or request an advisory review.
Comments