Menu Engineering for Restaurants and Cafés: Stars, Plowhorses, Puzzles and Dogs, with a Worked Example

What menu engineering is, and where it comes from

Michael Kasavana and Donald Smith set out the method in their 1982 book “Menu Engineering: A Practical Guide to Menu Analysis” (Wikipedia summarises the history). It is also called menu analysis or the menu engineering matrix. Every dish gets two scores, and the scores put it in one box of the grid.

The point is to stop judging the menu by gut feeling or by food cost percentage alone. A dish with a low food cost percentage can still earn little per plate, and a dish with a high one can be your most valuable item. The matrix makes that visible in an afternoon.

What you need before you start

• Units sold per dish for one period, from your POS item-sales report. Four weeks is a good default: long enough to smooth out one busy weekend, short enough that recipes and prices haven't changed in the middle.
• The plate cost of each dish: every ingredient at today's purchase price, including what goes out with it. Bread, pickles, rice, sauces and the small salad that comes with a grill belong in the cost. Our guide to food cost and menu pricing shows how to cost a recipe.
• Selling prices before VAT. The tax is not yours, so leave it out of the margin. If your menu prices include VAT, divide by 1 plus the rate first.
• Your categories, as guests see them on the menu: mains with mains, drinks with drinks.

If a large share of your orders comes through a delivery app, run delivery as its own analysis. The platform's commission comes out of the same plate, so a dish that is a star in the dining room can be a plowhorse on delivery.

How to do menu engineering: the formulas

The calculation below follows the method as taught in Penn State's Introduction to Food Production and Service textbook. Do it for one category at a time.

1. Contribution margin (CM) per plate = selling price before VAT − food cost per plate.
2. Total CM per dish = CM per plate × units sold.
3. Average CM for the category = total CM of all dishes ÷ total units sold. Do not take a simple average of the per-plate margins: dividing by units sold gives each dish its real weight.
4. Menu mix % = units sold of the dish ÷ total units sold in the category × 100.
5. Popularity threshold = (100% ÷ number of dishes in the category) × 70%. With 6 dishes: 16.7% × 0.7 = 11.7%.
6. Classify: CM at or above the average is high; menu mix at or above the threshold is popular. Then read the group off the matrix.

A worked example in Saudi riyals

A casual restaurant has six mains. Four weeks of POS data and the recipe costs give this:

The average CM is 34,120 ÷ 1,240 = SAR 27.52. Anything earning that or more per plate is high-margin. The popularity line is 11.7% of 1,240, about 145 plates.

What the example shows

• Food cost % points the wrong way. The lamb kabsa has the worst food cost percentage on the list (53%) and is a star, because each plate leaves SAR 34. The pasta has the best percentage (26%) and is a dog. A menu run on food cost % alone would push the wrong dishes.
• The plowhorses carry the room. Shawarma and burger are 63% of mains and over half the total margin. Nothing in this analysis says remove them; it says each plate could earn a little more.
• Borderline is borderline. The mixed grill clears the popularity line by five plates. Treat it as a star to watch, not a result to act on.

In Jordanian dinars, Kuwaiti or Bahraini dinars or the Omani rial, work in three decimals (JOD 4.250) so rounding doesn't move a dish across a line.

What to do with each quadrant

The actions below build on the ones the textbook method recommends, with notes on how they play out in a Gulf or Jordanian restaurant.

Stars: protect them

• Keep the recipe, the portion and the plating exactly as they are. A star that quietly shrinks stops being a star.
• Give it a good photo and a clear place in its category. Staff should know it by name.
• Test a small price rise only if it has been a star for several periods in a row.

Plowhorses: earn more per plate without losing the plate

• Check the cost first: portion size, the side that comes with it, a cheaper supplier for the same quality.
• Raise the price in small steps, then measure. Use the break-even check below before you decide.
• Offer a paid upgrade (extra meat, a larger size, a drink) rather than changing the base dish guests come for.
• The textbook also suggests moving a plowhorse to a less prominent spot. Do that gently: guests who come for the shawarma and can't find it will not order the salmon instead.

Puzzles: make them easier to choose

• Move it higher in its category, give it a photo and a description that says what it is: “grilled salmon, lemon-garlic butter, saffron rice”, in both languages.
• Ask staff to suggest it, and let them taste it so the recommendation is honest.
• If it still doesn't move after a few weeks, a slightly lower price or a smaller portion at a lower price can make it an easier choice. Keep the margin above average, or it becomes a dog.

Dogs: replace, unless there is a reason to stay

• Remove it, or replace it with a new dish that uses ingredients you already stock.
• Keep it if it does another job: the only vegetarian main can decide where a group of eight eats, and a dish that uses up trimmings with little labour costs you almost nothing to offer. If you keep it, make it cheaper to produce.
• Taking a dish off also simplifies prep and stock, which the matrix doesn't measure.

Testing a price change: the break-even check

Before raising a plowhorse's price, work out how many plates you can afford to lose. The rule: plates needed at the new price = current total CM ÷ new CM per plate.

In the example, the shawarma plate earns 9,240 over four weeks at SAR 22 a plate. At SAR 34 instead of 32, each plate earns SAR 24, so 9,240 ÷ 24 = 385 plates earn the same. The price rise pays as long as sales fall by fewer than 35 plates, about 8%. If sales hold at 420, the dish earns SAR 840 more per four weeks.

Then measure. Change one thing at a time, run it for four normal weeks, and compare with the period before. If you change the price, the photo and the position at once, you won't know which one worked.

Menu layout and placement: what the evidence shows

Most menu engineering advice online says to put your best dishes in a “sweet spot” or “golden triangle” where the eye lands first. The research is weaker than that advice suggests.

• No sweet spot found. In an eye-tracking study of 25 diners reading a two-page mock menu, Sybil Yang found people read it in order, like a book, left to right and down the page, and found no evidence of the classic sweet spot. The sweet-spot idea traces back to a 1978 scan-path study by a graphic designer, William Doerfler (Nation's Restaurant News, 7 February 2012).
• Mains first. The same study found diners spent their first pass on the main-course section and built the rest of the meal around their choice, with a “sour spot” at the bottom corners where general information and salads sat.
• Price format had a small effect. At a Culinary Institute of America restaurant, lunch guests given prices as plain numerals spent significantly more than those whose menus showed a dollar sign or prices written in words. The authors note that party size and time at the table mattered more, and that the result may only hold for lunch at that restaurant (Yang, Kimes and Sessarego, Cornell, 2009).
• Items are not independent. Noone and Cachia showed, with two years of data from 48 steakhouse outlets, that pricing and placement decisions that account for substitute dishes gave better revenue results than treating every dish on its own (International Journal of Hospitality Management, 2020).

Those studies are small, mostly American and mostly about printed menus. Nobody has shown that a layout trick moves sales by a fixed amount in a Riyadh café or an Amman restaurant. What holds up is plainer: guests read in order, find the category they came for, and choose from what they notice and understand. So put the dishes you want chosen near the top of their category, describe them clearly, and check the effect in your own sales. Our restaurant menu design guide covers layout and wording in detail.

Why a digital menu makes menu engineering cheaper

Menu engineering is a loop: analyse, change one thing, measure, repeat. With a printed menu every change is a reprint, so most restaurants run the loop once a year, if at all. With a QR menu the change takes a minute and costs nothing, so you can test monthly.

In AlaMenu's dashboard you can:

• Move a dish up or down within its category, and move whole categories up or down, to test placement for a puzzle.
• Add or swap a photo for each dish and an image for each category.
• Rewrite the name and description in Arabic and English, which guests also search.
• Change a price and have it live the next time someone scans, without reprinting the QR code on the tables.
• Hide a dish by marking it unavailable while you rework it, then switch it back on.

What it doesn't do: AlaMenu doesn't import your POS sales or build the matrix for you, so the analysis stays in your spreadsheet. There is no separate badge or “popular” label either; if you want to flag a dish, say so in its description. In Saudi Arabia, a new or changed dish also needs its calories and allergens on the menu; see Saudi menu requirements.

The Free plan covers one location with up to 10 items per category; Pro adds unlimited items. See the QR menu features and pricing, or open the demo menu to see how photos and categories look on a phone.

Where menu engineering goes wrong

• Comparing across categories. A juice and a mixed grill can't be judged against the same average. Run drinks, starters, mains and desserts separately.
• Chasing the average. Because the lines are averages, some dishes always fall below them. Replace every dog and a new set of dishes falls below the new average. The textbook advice is to temper the numbers with common sense.
• Ignoring labour. The matrix counts food cost only. A dish that needs ten minutes of a cook's time at the peak costs more than its CM shows.
• Out-of-date costs. Lamb, chicken and cooking oil prices move. If the plate costs are six months old, the margins are guesses.
• Acting on one month. Change dishes that land in the same group two periods running, not one.

A monthly menu engineering routine

1. First week of the month: export last month's item sales from the POS.
2. Update plate costs for any dish whose main ingredient changed price.
3. Fill the spreadsheet: price before VAT, food cost, units sold. Let the formulas do the rest.
4. Pick at most three changes: one plowhorse price or cost, one puzzle placement or photo, one dog to replace or re-cost.
5. Make the changes on the menu the same day and write down the date.
6. Next month, compare the same dishes before and after. Keep what worked, undo what didn't.

Questions

Sources

1. Wikipedia: Menu engineering (Kasavana and Smith, 1982; sweet-spot evidence)
2. Egan, Introduction to Food Production and Service, Penn State: 2.6 Menu Analysis (Engineering)
3. Nation's Restaurant News: Study counters prevailing restaurant menu theories (7 February 2012)
4. Yang, Kimes and Sessarego: $ or Dollars: Effects of Menu-price Formats on Restaurant Checks (Cornell, 2009)
5. Noone and Cachia: Menu engineering re-engineered (International Journal of Hospitality Management, 2020)

Keep reading

• How to Price a Restaurant Menu: Food Cost, Margin and VAT, with a Worked Example

  How to price a restaurant menu: cost each dish with yield and waste, pick a food cost target, check margin and VAT, with a worked shawarma example.

  15 min read

• Restaurant Menu Design: A Practical Guide for Printed and QR Menus

  Restaurant menu design that works on paper and on a phone: category order, how many dishes, names, photos, type sizes, Arabic layout, prices and a checklist.

  16 min read

• How to Write a Bilingual Arabic-English Menu: Dish Names, Descriptions and Layout

  How to write a bilingual Arabic English menu: when to keep or translate dish names, 29 Gulf and Levant dishes in English, and the right-to-left traps.

  12 min read