Menu engineering: how to build a more profitable menu

Menu engineering is a way of sorting every dish on a menu by two numbers: how many you sell and how much each one leaves you after its food cost. Put those two numbers on a grid and each dish lands in one of four boxes. The box tells you whether to keep it, reprice it, rework it or take it off.

This page covers the method, a full worked example you can copy, where the method goes wrong, what the research does and doesn't say about menu design, and the UK rules that limit what you can print on a menu.

What menu engineering measures

The method looks at two things for each dish over the same period.

Popularity. The dish's share of all the dishes sold in that menu section. If you sold 720 mains and 120 were roast chicken, roast chicken has a 16.7% share.

Contribution margin. The net selling price minus the plate cost. In the UK, "net" means the menu price without VAT. A £18.00 roast chicken eaten in is standard-rated, so the net price is £15.00. If the plate cost is £4.50, the contribution margin is £10.50.

Menu engineering ranks your dishes by pounds of margin, because a steak at 46.7% food cost can still leave more cash per plate than a risotto at 20%. If you want to understand the percentage itself, read our page on food cost percentage. This page is about pounds per plate and how many plates.

Where the method came from

A 2022 review by Horváth and colleagues in the journal DETUROPE sets out the history: "The first model was created by Miller in 1980", which sorted dishes using food cost. "Kasavana and Smith modified the Miller model in 1982 and used contribution margin instead of production costs." Pavesic followed in 1983 with a model "based on the margin and the cost of producing the food".

The Kasavana and Smith version is the one most calculators use today, and it's the one this page follows. Their book is Kasavana, M. L. and Smith, D. I. (1982) Menu Engineering: A Practical Guide to Menu Analysis, Okemos, MI: Hospitality Publications. We found no online copy, so the formulas below follow the DETUROPE paper's summary of their method. Kasavana still writes about it. In a 2025 article for the American Hotel & Lodging Educational Institute he says the analysis was "initially developed by Michael L. Kasavana and Donald I. Smith at the Michigan State University School of Hospitality Business (circa 1982)".

The four boxes have names that stuck:

• Star: sells well and leaves a good margin. Protect it.
• Plowhorse: sells well, leaves less per plate. Work on cost or price, carefully.
• Puzzle: good margin, few sales. Work on why people don't choose it.
• Dog: low sales, low margin. Usually the first to go.

The two cut-off lines

Every dish is compared against two lines. Getting these lines right matters more than anything else in the method.

The popularity line: the 70% rule

You don't compare each dish against an equal share of sales. You compare it against 70% of an equal share. DETUROPE gives the formula as "(100/number of dishes tested)*70%".

With 8 mains, an equal share is 12.5%. Seventy per cent of that is 8.75%. Any main with 8.75% or more of main-course sales counts as popular.

The 70% factor gives some slack. With a straight equal share, roughly half your dishes would always fail, even on a menu where every dish sells steadily. It's a convention, and we found no published evidence that 70% is the best number. Treat it as the standard setting, and stick with it so your results compare month to month.

The margin line: weighted or simple average

This is where sources disagree, and it changes results.

The classic method uses a weighted average: the total contribution margin from all dishes in the section, divided by the total number sold. In DETUROPE's words: "The total margin divided by the total number of portions sold gave the average margin."

Some teaching material uses a simple average instead: add up each dish's margin and divide by the number of dishes. The menu engineering exercise from Northern Arizona University's hospitality course does this, dividing $30 of total item margins by 10 items, and uses the 7% popularity line for 10 items.

The weighted average is pulled towards your best-selling dishes. If your bestsellers are cheap, the line drops. The simple average treats a dish that sold 30 the same as one that sold 180. Pick one, write it on your worksheet, and use it every time. The worked example below shows a dish that changes box depending on which you pick.

The worksheet

Calculator pages are useful, but you should know what they're doing. The full worksheet is below. Do it in a spreadsheet, one menu section at a time (mains with mains, desserts with desserts).

A. For each dish, record:

1. Menu price (as printed, including VAT)
2. Net price = menu price ÷ 1.20 for standard-rated sales
3. Plate cost (from your costed recipe card)
4. Contribution margin (CM) = net price − plate cost
5. Number sold in the period
6. Total CM = CM × number sold

B. For the section, work out:

1. Total number sold (sum of line 5)
2. Popularity share for each dish = line 5 ÷ line 7
3. Popularity line = (1 ÷ number of dishes) × 0.70
4. Average CM = total of line 6 ÷ line 7 (weighted)

C. Classify each dish:

1. Popularity: high if line 8 ≥ line 9
2. Margin: high if line 4 ≥ line 10
3. Box: Star, Plowhorse, Puzzle or Dog

Use at least four weeks of sales so a quiet weekend doesn't swing the answer. WHY Hospitality's free menu engineering calculator says to use "the sales from the last four to eight weeks", which is a sensible range, though it's their advice rather than a standard.

Worked example: eight mains

The dishes and numbers in this worked example are made up. Four weeks, eight mains, all eaten in at 20% VAT.

The section totals:

• Total sold: 720
• Total CM: £7,112.00
• Weighted average CM: £7,112.00 ÷ 720 = £9.88
• Popularity line: (1 ÷ 8) × 0.70 = 8.75%

So a dish is popular at 8.75% or more and high-margin at £9.88 or more.

Notice the steak. At 8.3% it sits just under the popularity line, so it's a Puzzle even though it sold 60. And look at the burger: it's the bestseller by a long way but leaves the least per plate.

What changes with a simple average

The simple average CM is the eight CMs added up and divided by 8: £82.90 ÷ 8 = £10.36.

Under that line, the mushroom risotto (CM £10.10) drops from Puzzle to Dog. Nothing else moves. Same dish, same sales, a different box, and a different action: you'd promote a Puzzle but you might cut a Dog. This is why you should write down which average you used.

What changes if you forget VAT

If you work from menu prices including VAT, every margin goes up. The risotto's CM becomes £12.60 and the weighted average becomes £12.83. The risotto drops to Dog again.

This happens because VAT is a percentage of price but plate cost isn't. Leaving VAT in favours expensive dishes, so your cheaper ones look worse than they are. Jelly's UK menu engineering guide makes the same point with a £14.00 dish: it says working from the gross price overstates the margin "by £2.33, or roughly 30%".

Getting the inputs right

The grid is only as good as what goes in. Three inputs cause most of the trouble.

Net price and VAT

Under HMRC VAT Notice 709/1, food eaten on your premises is standard-rated, and so is hot takeaway food. Cold takeaway food is usually zero-rated, with exceptions listed in VAT Notice 701/14. The current rates are on GOV.UK's VAT rates page.

So the same sandwich at the same price can leave you a different margin eaten in or taken away. If you sell both ways, the notice explains how to apportion mixed sales and what to do with hot and cold food sold together. For menu engineering, the simplest fix is to run eat-in and takeaway as separate grids, using each one's own net price. Make sure your till records the split, or you'll be guessing.

If you aren't VAT registered, the net price is the menu price.

Delivery apps need their own grid

If you sell through a delivery app, the app keeps a commission on each order, and you pay for the box and bag. So a dish that's a Star in the dining room can be a Dog on the app.

Work out a delivery contribution margin for each dish:

Delivery CM = net price − commission − plate cost − packaging

Net price is the app price without VAT. Commission is whatever rate your contract with the app sets, worked out on the price the contract says. Use your own rate; they differ by app and by deal, and we haven't put a typical figure here because we found no official one. Packaging is the cost of the container, lid, bag and any sauce pots for that dish.

Kasavana makes the same point in his AHLEI article: item costs "should reflect all expenses incurred to produce and serve an item", including "packaging, if any".

Here's a worked example with made-up numbers. A burger sells on the app for £15.00 including VAT, so £12.50 net. Say your contract takes 25% of the £12.50, which is £3.125. Plate cost is £3.60, as in the grid above, and packaging is £0.60. Delivery CM is £12.50 − £3.125 − £3.60 − £0.60 = £5.175, about £5.18. In the dining room, at the same price, the same burger makes £8.90.

Run the app grid on app sales only. Its average CM will be lower than your dining room's, which is fine, because each dish is judged against the other dishes on the same channel.

Service charge isn't part of the price

Leave service charge out of the net price completely. It isn't margin for the dish.

VAT Notice 709/1 section 2.3 says: "If you make a service charge, it's standard-rated". And since 1 October 2024 the Employment (Allocation of Tips) Act 2023 has been in force (the commencement regulations set the date). The statutory code of practice tells employers to "Pass on all tips and service charges to workers without deductions", and says this covers service charges whether they're added automatically or not.

So a 12.5% service charge doesn't make your mains more profitable. It belongs to staff. If you raise menu prices and drop a service charge, or the other way round, rerun the grid, because the net price per dish has changed even if the customer pays the same.

Plate cost

Plate cost needs to be the cost of everything on the plate: the protein, the sides, the sauce, the garnish, the oil in the fryer. Use today's supplier prices. A dish costed two years ago is costed at two-year-old prices.

Jelly's guide claims a plate cost built from raw ingredient prices alone "typically understates true cost by 15–25%". That's their estimate. They don't cite a study, so treat it as opinion. But the direction is right: yield loss from trimming and cooking is real, and if you cost a 250g steak at the price of 250g of raw meat, your margin is wrong.

Three habits help:

• Re-cost recipes when supplier prices change. Your approved suppliers list should show who you buy each ingredient from, so you know which price to use.
• Check what arrives. A delivery check catches short weights and substitutions that raise your real cost.
• Compare theory with stock. If a regular stock count shows you used far more salmon than salmon sales explain, your plate cost or your portioning is wrong.

Number sold

Take this from the till report. Use the same dates for every dish. Leave out staff meals, comps and wastage, because those aren't sales. Record waste separately, as our page on waste management explains. And count specials only if they ran for the whole period, or they'll drag down the popularity line for everyone else.

What to do with each box

The box suggests a direction. Before you act, ask why the dish is where it is.

Stars

Leave the recipe and portion alone. Make sure the team describes it well and that it's always available. Small price rises are possible, but a Star is often what brings people back, so go gently and watch sales after any change.

Plowhorses

These carry your volume. The options are to:

• reduce plate cost without changing what the guest notices (a different supplier, a cheaper cut, a smaller garnish)
• raise the price a little and watch sales
• pair it with a higher-margin side or drink.

Don't cut the portion so much that it stops being the dish people come for. In our example, the burger is a Plowhorse selling 180 a month. A 30p cut in plate cost is worth £54 a month. A 60p rise in net price is worth £108, if sales hold.

Puzzles

The margin is there, but people aren't choosing it. Ask why. Is the name clear? Is the price a jump from everything around it? Do servers know it well enough to recommend it? Taste it again. Use your pre-shift huddle to brief the floor.

If nothing works after a couple of cycles, a Puzzle can take a slot that a better seller could use.

Dogs

Take it off unless there's a reason to keep it. Good reasons include: it's the only vegan main, a regular group orders it every week, or it uses up trim from another dish. If you keep a Dog for a reason, write the reason down so the next review doesn't keep asking.

Removing a Dog also leaves one fewer recipe to prep, store and train on, and one fewer set of ingredients to go out of date. That links to stock rotation: slow sellers are where waste builds up.

Where menu engineering goes wrong

The method is useful. It also has limits that its own literature points out.

It favours expensive dishes. Because it ranks by pounds of margin, a high-priced dish nearly always shows a higher CM than a cheap one. DETUROPE notes the method "favours higher priced foods". That's why Pavesic's 1983 model also looks at the cost of producing each dish. If your grid keeps telling you to sell more steak, check whether that's a real insight or just the maths.

It ignores labour. Margin is net price minus food cost. A dish that takes three minutes to plate and one that takes fifteen show the same margin if their food cost and price match. For a fuller picture of food and labour together, look at prime cost. If a slow dish is clogging the pass on busy nights, that shows up in sales per labour hour, not on this grid.

It ignores how dishes sell together. Your low-margin starter might be why a table orders three courses. A Dog side might be what sells your Star main. The grid looks at each dish alone.

It depends on the average. As the worked example shows, the simple and weighted averages can put the same dish in different boxes.

Small sections give odd results. With 3 desserts, the popularity line is 23.3%. One slow week can move a dish across it. Use a longer period for small sections.

The research on menu design

Most menu engineering guides add a section on menu "psychology": put your best dish in the top right corner, take the pound signs off, use boxes to draw the eye. Some of this has been tested. Much of it hasn't.

The "sweet spot" doesn't hold up well

The claim is that guests look first at a particular spot, often the upper right or the centre, so that's where your Stars should go.

A 2012 eye-tracking study by Sybil Yang in the International Journal of Hospitality Management tested it (there's a free author's copy hosted by Penn State). It recorded 25 people reading a two-page menu in April 2008. The results "suggest traditional menu 'sweet spots' may not exist". People read the menu "like they read a book", starting at the upper left. The study found "no particular frequency-based sweet spot", and it found "sour spots" at the bottom of each page that got less attention.

The paper goes further. It says that menu "design principles as applied to restaurant menus have not been empirically proven to increase purchase intention, actual sales, or even attention".

It's a small study, with 25 people and one menu. It doesn't prove that position never matters. But it's direct evidence, and the sweet-spot claim has very little evidence behind it. In practice, don't hide good dishes at the bottom of a long list, and don't expect moving a dish to the top right to fix a Puzzle.

Price format: some evidence

A 2009 study by Yang, Kimes and Sessarego at Cornell's Center for Hospitality Research tested how menu prices are written (report PDF on Sheryl Kimes's website). Guests given a menu with plain numbers ("14") "spent significantly more than those who received a menu with prices showing a dollar sign or those whose menus had prices written out in words".

That's real evidence, but it's narrow. It was one restaurant (St Andrew's Café at the Culinary Institute of America), at lunch, in the US, with dollars. We found no UK replication. If you want to try numeral-only prices, test it on your own sales over a few weeks rather than assuming the result carries over.

Whatever format you choose, the law still requires a clear price. More on that below.

Descriptions and anchors

Many guides say longer descriptions sell more, or that one very expensive dish makes everything else look cheaper. Treat these as ideas to test. Write descriptions that are accurate and appetising. Then let your sales tell you whether it worked.

UK rules that limit what goes on a menu

Menu engineering tells you what to change. The law sets some limits on how you print it. Three sets of rules matter most.

Price marking

The Price Marking (Food and Drink Services) Order 2003 sets out how prices must be shown. Legislation.gov.uk only holds the Order in its original form, so check for later amendments before you rely on any detail.

• Article 4: prices must include VAT. It also covers "any charge which is payable in addition to the price of any food", which means service charges and cover charges.
• Article 5: if you sell more than thirty items (other than wine), you must show prices for at least thirty of them. For wine with food, you must show prices for at least five, or all of them if you have fewer.
• Article 6: if a dish runs out, its price must be withdrawn "as soon as is reasonably practicable".
• Article 7: prices must be "unambiguous, easily identifiable and clearly legible". For an eating area, they must be shown "at or near the entrance" so people can see them before they go in. And "Any additional charge or minimum charge payable shall be indicated at least as prominently as the price of any food to which it relates."

That last point matters for menu design. You can't put dish prices in bold and the service charge in small print at the bottom. If you drop the pound sign to test the Cornell finding, the price itself must still be clear.

Article 6 also links to the grid. If a Dog keeps running out because you barely stock it, you're supposed to take its price down each time. One more reason to cut it.

Calorie labelling for larger businesses

The Calorie Labelling (Out of Home Sector) (England) Regulations 2021 apply in England only. Under regulation 7, a business is covered if, on the first day of its financial year, it "has 250 or more employees". GOV.UK's implementation guidance says the rules apply "from 6 April 2022" and gives restaurants, pubs, cafes and third-party delivery apps as examples of businesses covered.

Regulation 5 requires the calories to be displayed "at the point at which the consumer chooses what food to buy". The label has to give the calories followed by "kcal" and the portion size. The menu also needs the statement "adults need around 2000 kcal a day". According to the guidance, calories must be "given per portion" and not per 100g, and a menu that only lists children's food doesn't need the 2000 kcal statement.

If you don't comply, you can get an improvement notice. Failing to comply with one is an offence. The Schedule lets an officer impose a "fixed monetary penalty", defined as "a penalty of £2,500".

For menu engineering, the rules mean every recipe change on a covered menu has a second cost: the calorie figure has to be recalculated and reprinted. Swapping an ingredient to cut a Plowhorse's plate cost changes its calorie figure too. Plan reviews so changes go out in batches.

Allergen information

Regulation 5 of the Food Information Regulations 2014 covers food that is "not prepacked", such as restaurant dishes. It lets you give the allergen information "by any means the operator chooses, including, subject to paragraph (3), orally". Paragraph (3) sets conditions if you do it orally. The government's allergen guidance for food businesses has the detail.

So every new or changed recipe needs its allergen record updated before it goes on sale. If you change a Plowhorse's sauce to cut cost, the allergens may change too. Our page on allergen management covers how to keep those records current.

How often to run it

We found no official schedule. Advice from software companies varies: Jelly says to "Re-run the analysis monthly", and WHY Hospitality says "at least every three months". Both are opinions.

A reasonable pattern for most kitchens:

• Every month, update plate costs for dishes whose main ingredients changed price.
• Every quarter, or at each menu change, run the full grid on at least four weeks of sales.
• After any change, check the dish you changed four weeks later. Did sales hold? Did the margin move?

Keep each grid. The value builds over time: a dish that's been a Puzzle for three quarters is a different problem from one that just slipped.

Fold the headline numbers into your regular restaurant KPIs, so the margin mix sits next to sales and labour.

Benchmarks: what's published and what isn't

People often ask what share of a menu "should" be Stars, or what margin a main "should" leave. We found no official UK figure for either. These are the closest published numbers.

On food gross profit, the UKHospitality and Christie & Co Benchmarking Report 2022 surveyed 54 companies with 4,791 managed outlets. Across the whole survey, food gross profit was 67%. The figures are "based on results for the six months to December 2021", a period that still included the end of furlough, so treat them as a 2021 snapshot rather than a current target. The equivalent food cost is about 33% of net sales.

Jelly's guide also describes a rule of thumb of roughly 30% food and drink cost, 30% labour, 30% overheads and 10% profit. It's a trade convention. The same benchmarking report put payroll at 28.3% of sales in that period, so the labour part isn't far off for that sample. The rest of the rule has no published survey behind it.

Use benchmarks to sense-check the whole menu. Menu engineering compares your dishes with each other. A Star on your menu might have a margin that would be a Dog somewhere else, and that's fine.

Running it on a real menu

A few practical notes from the method itself.

Run one section at a time. A starter nearly always leaves less than a main, so on a shared grid every starter would show as a Dog (and you'd end up cutting your whole starter list).

Every dish's sales must come from the same date range and the same till report.

Watch specials and seasonal dishes. If a dish was only available for two of the four weeks, either leave it out or adjust its count, and note which.

Drinks need their own grid. The same method works for cocktails or wine, using drinks costs.

Share the results. The kitchen team controls portioning and waste, which drive plate cost. The front of house team controls what gets recommended. Both need to know which dishes are Puzzles and which are Stars.

If you're building a set of weekly routines around margins, our restaurant management tips cover the other checks that sit alongside this one.

Frequently asked questions

What is the formula for menu engineering?

For each dish, contribution margin is net price minus plate cost. The popularity line is (1 ÷ number of dishes) × 70%. The margin line is the total contribution margin divided by the total number sold. Dishes above both lines are Stars, above popularity only are Plowhorses, above margin only are Puzzles, and below both are Dogs.

Should I use prices with or without VAT?

Without. VAT goes to HMRC, so it isn't margin. Divide standard-rated menu prices by 1.20. Leaving VAT in makes expensive dishes look better than they are.

Why is the popularity line 70% and not 100%?

It's the convention from the Kasavana and Smith method. It stops roughly half your menu from failing every time. We found no evidence that it's the ideal number, but using it consistently lets you compare results over time.

Can I use menu engineering for drinks?

Yes. Run drinks on their own grid with their own costs and sales. Don't mix them with food.

Does menu engineering work for small menus?

It works, but it's noisier. With few dishes, the popularity line is high and small swings move dishes across it. Use a longer sales period.

Do I have to put calories on my menu?

Only if your business is in England and had 250 or more employees on the first day of its financial year, under the 2021 regulations. Smaller businesses aren't required to, though they can choose to.