Menu engineering

What menu engineering is

Menu engineering is the practice of scoring every dish against two measures, then acting on where it lands. The two measures are:

1. Contribution margin. The cash gross profit a dish leaves behind after its plate cost. Not the percentage, the pounds and pence.
2. Popularity. The share of total dishes sold that this one dish accounts for, measured over a defined trading period.

Score both, plot them, and every dish on the menu falls into one of four groups. Each group has a different, well established set of actions attached to it. That is the whole discipline. It is not complicated, it is just rarely done, because it requires sales mix data that most kitchens never pull from the till and plate costs that most kitchens have not refreshed in a year.

The framework was set out by Michael Kasavana and Donald Smith at the Michigan State University School of Hospitality Business in 1982, and the underlying logic has not needed much updating. What has changed is the volatility. When ingredient prices moved two or three percent a year, an annual costing review was defensible. It is not defensible now.

Why it beats intuition

Every head chef has a view on which dishes make money. In my experience that view is right about half the time, and the half it gets wrong is expensive.

The reason is that the two things a kitchen notices most, how often a dish is called and how hard it is to make, are only loosely related to what it earns. A dish that goes out sixty times a service feels like a winner on the pass. If it carries a £4.10 margin against a menu average of £9.60, it is not a winner, it is a volume problem.

The opposite is also common. A dish nobody rates, with an awkward garnish and a name that means nothing to the customer, often carries the best margin on the board. It is not underperforming because it is bad. It is underperforming because it is buried in the third section, described in four words, and never mentioned by the floor team.

The menu engineering matrix

The matrix is a four box grid. The vertical axis is contribution margin, split at the average margin across the menu. The horizontal axis is popularity, split at a threshold of 70 percent of an equal share of sales. If you have twenty dishes, an equal share is 5 percent each, so the popularity threshold is 3.5 percent.

Stars

High popularity, high margin. These are the dishes doing the commercial work. The instinct is to leave them alone, and broadly that is right, but there are three active moves worth making. Lock the specification so the portion does not creep upwards over a busy summer. Secure the supply line for the key ingredient, because a star that goes off for three weeks costs you more than any other line on the menu. And test a modest price rise. Stars are the dishes most likely to carry a price rise, because the customer is choosing them on desire rather than price. Treat that as a hypothesis to test on one dish at a time, holding everything else constant, and read the units four weeks later rather than assuming the rise was free.

Plowhorses

High popularity, low margin. The classic example is the burger, the fish and chips, the roast. The mistake operators make here is to delist them, which cuts footfall along with the margin problem. The correct approach is to work on the cost side while leaving the demand alone:

• Re-examine the specification line by line. A 20g reduction in a protein portion is rarely noticed and often worth 3 to 5 percent of plate cost.
• Renegotiate the core ingredient, or move to a different cut or grade that survives your cooking method.
• Attach margin around it. A plowhorse with a £3.50 side and a £5 beer alongside it is a different proposition on the basket level.
• Reduce the labour cost inside it, since a plowhorse with a fourteen minute pick up is costing you more than its ingredients suggest.

Puzzles

Low popularity, high margin. The single biggest available win on most menus, because you are not trying to change the economics of the dish, only the number of people choosing it. Interventions, roughly in order of effect:

• Move it. Position on the page is worth more than almost anything else you can do.
• Rewrite the description. Two concrete details beat five adjectives. Provenance, method, or a single specific ingredient.
• Reprice it relative to its neighbours so it reads as the sensible choice rather than the expensive one.
• Brief the floor team with one sentence they can actually say. Most puzzles are puzzles because nobody has ever recommended them.

If a puzzle stays a puzzle after four weeks of all four interventions, it is not a puzzle. It is a dog with good arithmetic, and it should come off.

Dogs

Low popularity, low margin. Cut them. The resistance is usually emotional or historic, occasionally legitimate, for instance a dish that exists to cover a dietary requirement and quietly keeps a table of six from going elsewhere. That is a real reason to keep a dog. "The chef likes it" is not.

Removing dogs does more than tidy the P&L. It shortens the menu, which shortens the decision, which reliably improves both average spend and ticket times. How many dishes fall into that bottom quadrant is a question your own sales mix answers. Score the menu first, then decide, rather than working to a rule of thumb about how many should go.

The data you need first

Menu engineering is only as good as its inputs. Before any analysis, gather:

Eight weeks of sales mix is the practical minimum. Twelve is better. Anything shorter and you are engineering around a bank holiday or a heatwave.

The formulas

All of these run on the net selling price, which is your menu price with VAT stripped out. If you are VAT registered and pricing a £26.00 main, your net selling price is £21.67.

Applying these to set a price, rather than to score one you already have, is covered in the guide to pricing a menu.

The last one is worth dwelling on. Your weighted gross profit is what lands in the accounts, and it is driven by the mix, not by the average of your dish level percentages. A menu where every dish sits at a respectable 70 percent GP can still deliver 63 percent weighted, if the dishes people actually order are the ones at the bottom of the range.

A worked menu profitability analysis

Here is a section of a menu, costed and scored. Ten main courses, 1,000 covers over the period, average contribution margin £13.42, popularity threshold 7.0 percent. A second example, laid out as an audit report on a fictional restaurant, is the sample menu profitability analysis.

Illustrative figures for demonstration. Replace with your own EPOS and costing data.

What the table is telling you

Four things jump out, and none of them are visible from the menu itself.

The sirloin is a star despite a 58 percent GP. It has the second highest cash margin on the section and the highest sales mix. If this operator had been managing to a GP percentage target, the sirloin would have been flagged as a problem and either repriced upwards or removed. Both moves would have cost real money.

The risotto is the best percentage on the menu and close to the worst performer. A 79.5 percent GP is meaningless at 41 covers. This is the archetypal puzzle: the margin is excellent, the demand is not there, and the fix is on the page rather than in the pan.

The sea trout is losing on both axes. A 49.3 percent GP at 5.1 percent of mix is a dog. Samphire and sea trout are both volatile lines, so it is also carrying risk. It comes off or it gets rebuilt around a cheaper fish with the same treatment.

The two plowhorses account for 37 percent of covers. That is the real story of this section. More than a third of everything that leaves the kitchen carries a below average margin. Taking 40p out of the fish and chips plate cost is worth more annually than any change to the top of the section.

Menu pricing formula

The basic mechanic is simple. To price for a target gross profit:

That gives you a floor, not an answer. Three adjustments turn it into a price.

1. Price against the section, not against the cost

Customers do not know your plate cost. They read your prices relative to each other. A section with a £34 top price makes a £26 dish read as measured and sensible. Remove the £34 dish and the same £26 now reads as the expensive one, and its sales fall, even though nothing about it has changed.

2. Use the spread deliberately

The gap between your cheapest and most expensive dish in a section should sit at roughly 2.5 times. Wider than that and the low end feels like the safe choice for everyone. Narrower and you have given up the chance to trade people upwards.

3. Round with intent

In UK casual dining, whole pounds outperform .95 endings on perceived quality without measurably reducing volume. In pubs and high street sites, .50 endings still read as fair value. Fine dining drops the currency symbol entirely and uses whole numbers. Pick the convention that matches your room and apply it consistently, because a menu with mixed conventions reads as careless.

Handling a price rise

Do not put every price up by the same amount on the same day. Regulars notice uniform rises immediately. Instead, raise stars and puzzles, hold plowhorses, and let the removal of dogs absorb the optics. Pair the change with at least two genuinely new dishes so the menu reads as a new menu rather than as the old menu costing more.

Menu psychology and layout

Once the arithmetic is right, the page does the rest of the work. The effects below are well documented and, more usefully, they are cheap to implement.

• 01

  Position beats everything

  On a single page menu, the top right area and the first two items of any section carry disproportionate attention. On a folded menu it is the top right panel. Put your puzzles there, not your cheapest dish.

• 02

  Section length caps at seven

  Beyond about seven options in a section, decision quality drops and customers retreat to the familiar, which is usually your plowhorse. Guests spend on the order of 109 seconds reading a menu, so a long section spends that budget on deciding rather than on being sold to. Gallup, widely cited Shorter sections sell your margin, not their habits.

• 03

  Remove the price column

  Prices aligned in a right hand column invite a vertical scan on cost alone. Set the price immediately after the description, same size, same weight, no leader dots. A Cornell study found guests given prices without a currency symbol spent 8.15 percent more than those shown a symbol. Yang, Kimes & Sessarego, Cornell, 2009

• 04

  Anchor high

  One deliberately premium dish at the top of a section raises the perceived reasonableness of everything beneath it. It does not need to sell in volume to earn its place.

• 05

  Describe with specifics

  Named provenance, method and a single distinguishing ingredient outperform strings of adjectives. "Herdwick lamb, three day brined, anna potato" does more than "succulent locally sourced lamb". The method, with before and after examples, is in the guide to writing menu descriptions. A field experiment found descriptive dish names lifted sales of those items by 27 percent, though it did not raise what guests were willing to pay. Wansink, Painter & van Ittersum, Cornell HRAQ, 2001

• 06

  Use boxes sparingly

  A box or a rule draws the eye reliably, which means it stops working the moment you use more than one per page.

• 07

  Watch the decoys

  A dish that exists mainly to make its neighbour look good is legitimate, but it needs to be genuinely orderable. Customers do occasionally choose the decoy, and it should not be a loss maker when they do.

Menu design best practice

Design here means the physical and typographic decisions, not the branding. A well engineered menu can still be undermined by the way it is set.

• One page where possible. Every fold or page turn loses attention and pushes customers back toward the first thing they read.
• Type at 10pt minimum, and 11pt if your room is dimly lit. An unreadable menu is ordered from by memory, and memory favours plowhorses.
• Generous leading between dishes. Crowded blocks get skimmed, and skimming benefits the familiar.
• Do not italicise whole descriptions. It reduces reading speed measurably and dilutes emphasis where you actually want it.
• Allergen information handled properly. Under UK rules you must be able to provide allergen information for all fourteen allergens on unpackaged food, and prepacked for direct sale items require full ingredient labelling under Natasha's Law. A clear line pointing to a staffed conversation is usually better on the page than a grid of symbols that clutters the layout.
• Print costs shape behaviour. If reprinting is expensive, the menu will not change when it should. Design for a format you can economically reprint every quarter.

Menu innovation without margin loss

Menus need to move. Regulars need a reason to come back, seasons change what is worth buying, and a menu that has not changed in two years reads as tired regardless of how good the food is. The risk is that every new dish arrives uncosted and quietly drags the weighted GP down.

The discipline that prevents this is simple: nothing goes on the menu without a costed specification and a target classification. Before a dish is written on the board, you should be able to say what its plate cost is, what its contribution margin will be, and whether you are launching it as a star, a puzzle you intend to promote, or a deliberate plowhorse to drive footfall.

For groups and multi site operators, this is what a structured NPD sprint is for: developing a batch of dishes against a defined margin and operational brief, rather than one at a time and after the fact.

Building your menu engineering template

You can run all of this in a spreadsheet. The minimum viable template has one row per dish and these columns:

1. Dish name and section
2. Menu price, gross
3. Net price, being the menu price divided by 1.20
4. Plate cost, from the recipe spec including yield loss
5. Gross profit percentage
6. Contribution margin in pounds
7. Units sold in the period
8. Sales mix percentage
9. Total cash margin, being units multiplied by contribution margin
10. Classification, driven by a formula against your two thresholds

Add two summary cells: average contribution margin across the menu, and the popularity threshold. The classification column is then a nested condition against those two cells. Sort by total cash margin descending and the top five rows will tell you where your business actually comes from.

Two habits make the template useful rather than decorative. Update plate costs from invoices monthly, not from memory. And keep the historic versions, because the interesting information is in the movement between quarters rather than in any single snapshot.

Nine mistakes to avoid

1. Costing on the gross menu price. Including VAT in the calculation overstates every GP figure by a wide margin.
2. Ignoring yield. A whole fish at £14 per kilo is not £14 per usable kilo. Costing on purchase weight understates plate cost systematically.
3. Managing to percentage alone. It pushes you toward cheap dishes and away from the ones carrying your cash margin.
4. Using too short a sales period. Six weeks that include a bank holiday will misclassify half your menu.
5. Forgetting labour. Two dishes with identical plate costs are not equally profitable if one takes twelve minutes on the pass.
6. Deleting plowhorses. You remove the margin problem and the footfall together.
7. Changing everything at once. If eleven variables move simultaneously you learn nothing about which one worked.
8. Skipping the floor team. A repositioned puzzle with an unbriefed front of house is a repositioned puzzle that stays a puzzle.
9. Treating it as a one off. Ingredient costs move continuously. An engineered menu is only engineered as at the date it was engineered.

Working with a menu engineering consultant

Everything above is something an operator can do without help, and plenty do. Bringing someone in is worth it in three situations.

You do not have the time. A full analysis on a thirty dish menu, done properly with verified plate costs, is around forty hours of work. Most head chefs do not have forty hours.

You need the answer to be uncontested. Menu decisions are political. A dish stays on because it was the owner's idea, or because the sous chef developed it. An external analysis takes the personalities out and leaves the arithmetic.

You are operating across sites. Multi site menu engineering is a genuinely different problem, involving specification control, supply consistency and the fact that the same dish can be a star in one location and a dog in another.

I spent five years as head chef developing food across seven pub sites, after two spells in a 3 AA Rosette kitchen, which matters here for one specific reason: the recommendations have to survive a real service. Reducing a portion by 20g is a spreadsheet exercise until you have to plate it, and a redesign that ignores what a section can actually execute at eight o'clock on a Saturday is not a redesign, it is a document.

Menu engineering FAQs

Want it one section at a time? Menu profitability by section covers breakfast, starters, lunch, mains and desserts.