Menu engineering: find your stars, plowhorses, puzzles and dogs from one month of sales

In our worked example of 12 mains, the dish with the highest food cost earns the most cash per plate, and the cheapest plate to make is one to cut. Three changes add close to $2,000 a month from the same covers.

Menu engineering shows which dishes to keep, fix or cut from one month of sales. Plot every dish in a menu section by popularity (its share of that section's sales) and by cash margin (price before VAT or sales tax, minus plate cost). That puts each dish in one of four groups: stars (popular, high margin), plowhorses (popular, low margin), puzzles (unpopular, high margin) and dogs (unpopular, low margin). Each group has a different fix. Last checked: 5 November 2026.

What is menu engineering?

It is a way to judge each dish by two numbers at once: how often guests choose it, and how much cash it leaves after ingredients. The method most operators use was published by Michael Kasavana and Donald Smith in 1982, and the four names come from it.

Why both numbers? A dish with a low food cost percentage can still earn little cash, and a popular dish can quietly cost you money on every plate. Our GP calculator guide shows why a percentage on its own misleads; menu engineering adds how often each dish sells.

How do you sort dishes into stars, plowhorses, puzzles and dogs?

Work one menu section at a time (mains, then starters, then desserts, then drinks), because a dessert will never sell as often as a main.

1. Pull a month of sales for every dish in the section: how many sold.
2. Cost every dish at today's ingredient prices, including sides and garnish.
3. Cash margin per dish = selling price before VAT or sales tax, minus plate cost. In the UK, divide a VAT-inclusive price by 1.2 first.
4. Popularity threshold = 70% of an equal share. With 12 mains, an equal share is 8.3% of sales, so the threshold is 5.8%. A dish above it is popular.
5. Margin threshold = the average cash margin across everything sold in the section, weighted by how many of each dish sold. A dish above it is high margin.
6. Sort: popular and high margin is a star; popular and low margin, a plowhorse; unpopular and high margin, a puzzle; unpopular and low margin, a dog.

Worked example: 12 mains, one month

Here is a 12-main menu for a neighbourhood restaurant, with 3,200 mains sold in a month. How we costed it: our analysis of US Bureau of Labor Statistics average retail food prices for September 2026 (for example ground beef at $6.79 a pound, sirloin $14.11, boneless chicken breast $4.20, cheddar $6.13 and white potatoes $0.99), applied to recipe weights, plus a fixed allowance for sauces, garnish and oil, plus 10% for trim and waste. Restaurants buy wholesale, so treat the costs as illustrative. The pound columns convert the same plate costs at £1 = $1.35 (European Central Bank reference rates, 4 November 2026) and use UK menu prices that include 20% VAT; the UK cash margin takes the VAT off first.

Thresholds: popularity 5.8% of mains sold; weighted average cash margin $14.09 in the US and £10.42 in the UK. The section takes $60,060 in a month at a 24.9% food cost; the UK version takes £44,417 before VAT at a 75% gross profit.

Two results surprise most owners:

• The steak frites has the highest food cost on the menu (34.8%) and is a star. It leaves £16.24 ($22.17) a plate, more than any other main.
• The mac and cheese has one of the lowest food costs (18.9%) and is a dog. It is cheap to make, but at £12.50 ($14) it earns little cash and few people order it.

What should you do with each group?

Keep a dog only when it does a job the numbers cannot see: the one vegan main, the dish that keeps a group of eight from going elsewhere, or a signature dish regulars ask for by name. Then reprice it so it at least reaches the margin line.

What is each change worth?

Put a number on each action before you reprint, and remember that most extra sales of one dish come from guests who would have ordered another.

• Plowhorse price rise. Add $1 (50p) to the burger. If the 620 a month hold, that is $620 (about £258 after VAT) more cash margin a month. If 5% switch to another plowhorse, you still keep most of it.
• Puzzle promotion. Move the pork chop to the top of the mains and have servers mention it. If it rises from 120 to 200 a month and the extra 80 come from burgers, each switch is worth $20.63 minus $11.66, or $8.97: about $718 a month.
• Dog swap. Replace the mac and cheese with a pasta built on the carbonara's ingredients at a carbonara-level margin. If the 150 who ordered it switch, each is worth about $4.18 more: $627 a month.

Together that is close to $2,000 a month from the same covers, without touching the stars. To set the new prices, use the menu price calculator. For when and how to tell guests about a rise, see our guide to raising menu prices in 2027.

What are the common mistakes?

• Mixing sections. Comparing desserts with mains marks every dessert a dog. Run each section on its own.
• Old costs. In the BLS consumer price index for the year to September 2026, ground beef rose 5.4%, fish and seafood 6.9% and coffee 7.1%, while chicken fell 0.6%. A matrix built on last year's costs can call a plowhorse a star.
• Too short a period. One week is noise. Use at least four weeks, avoid months with a holiday menu, and compare like with like.
• Forgetting sides and modifiers. If the burger's fries or the extra cheese are rung separately, include them in the plate cost and the price, or the margin is wrong.
• VAT in the price (UK). Cash margin must come from the price before VAT, or every dish looks 20% better than it is.
• Changing everything at once. Change a few dishes, then rerun the matrix four weeks later so you know what worked.

For the full costing method in dollars, see our guide to recipe costing and food cost; UK operators can use the food cost percentage guide.

The matrix worksheet to copy

MENU ENGINEERING WORKSHEET     Section: [mains]   Period: [4 weeks to ]

Number of dishes in section (n):             [ ]
Total sold in section:                       [ ]
Popularity threshold = 0.7 x (1 / n):        [ ]%

Dish | Sold (A) | Mix % (A / total) | Price ex VAT or sales tax (B) |
Plate cost (C) | Cash margin (B - C) | Total margin (A x (B - C)) |
Popular? (mix >= threshold) | High margin? (margin >= average) | Group

[ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ]
[ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ]

Average cash margin = sum of total margin / total sold = [ ]

ACTIONS (one per group, with a number)
Stars:      protect [ ], best menu position for [ ]
Plowhorses: price change [ ] by [ ]; cost change [ ]   Value: [ ]/month
Puzzles:    move / rename / server mention [ ]         Value: [ ]/month
Dogs:       remove or replace [ ] with [ ]             Value: [ ]/month

Rerun date (4 weeks after the change): [ ]

Can TablePort handle this for you?

With the Stock add-on (£149 or $199 a month per location, before VAT or sales tax), TablePort runs the matrix from your own sales.

1. Recipes carry plate costs, ingredient by ingredient, and move when a supplier price moves on an invoice.
2. The menu engineering report sorts dishes into stars, plowhorses, puzzles and dogs by fixed rules, so you see which dish quietly loses money.
3. AI-written margin recommendations in the daily report name items to promote or review.

The honest limit: it needs costed recipes. Dishes without a recipe have no plate cost, so start with your best sellers and add the rest as you go. The recommendations are suggestions; the menu is your call. See stock management.