Menu engineering: find your best and worst dishes in 5 steps
Menu engineering shows which dishes to feature, reprice or retire. Learn the Stars, Plowhorses, Puzzles and Dogs matrix with a worked example.

It is the end of a long Saturday. The kitchen is clean, the till is closed and you are scrolling through the week's sales report. The burger sold more than any other main. The lamb shank, the dish your chef is proudest of, hardly sold at all. You have a feeling about which dishes pay the bills, but if someone asked you to prove it, you could not.
That feeling is worth checking, and menu engineering is a simple way to do it. It turns a sales report into a short list of dishes to feature, reprice, explain better or retire, and you only need a spreadsheet to start.
What menu engineering is
Menu engineering judges every dish on two questions: how often do guests order it, and how much does each plate earn? The method was introduced in 1982 by Michael Kasavana and Donald Smith in their book Menu Engineering: A Practical Guide to Menu Analysis, and their matrix is still the version most restaurants meet.
"Earns" means contribution margin: the selling price minus the cost of the ingredients, as an open food service textbook defines it. It is an amount of money, not a percentage. Compare each dish with the rest of its category on both measures and it lands in one of four boxes:
| Popularity | Earns more than average | Earns less than average |
|---|---|---|
| Sells well | Star | Plowhorse |
| Sells less | Puzzle | Dog |
How to run it, step by step
1. Pick one category and a fair period
Work through starters, mains, desserts and drinks separately, because dishes in the same category compete with each other. Use sales from a few weeks, so that a quiet Tuesday and a busy Saturday both count. Your till report gives you the number of each dish sold.
2. Work out each dish's share of sales
Divide the number sold of a dish by the total sold in the category. This is its menu mix share.
3. Draw the popularity line
If every dish sold equally, each would have a share of 100 percent divided by the number of dishes. The usual convention, which the textbook above follows, is to call a dish popular when it sells at least 70 percent of that equal share. With eight mains, an equal share is 12.5 percent, so the line sits at 8.75 percent. The 70 percent is a convention rather than a law, so keep it the same from one review to the next.
4. Work out each dish's margin
Take the selling price before any sales tax and subtract the cost of the ingredients on the plate. Our guide to menu pricing shows how to cost a dish, including the trimming and cooking losses that make a plate cost more than the ingredients on the scale.
5. Draw the margin line
Add up the margin earned by every plate sold and divide by the number of plates sold. This weighted average is the line: a dish at or above it earns more than average.
A worked example
These numbers are made up, for eight mains and one month. Amounts are in whatever currency you price in, and the average margin comes to 12.70.
| Dish | Price | Ingredient cost | Margin | Sold | Share of sales | Result |
|---|---|---|---|---|---|---|
| Roast chicken | 19.00 | 5.70 | 13.30 | 180 | 18.0% | Star |
| Ribeye steak | 32.00 | 14.40 | 17.60 | 150 | 15.0% | Star |
| Beef burger | 16.00 | 5.60 | 10.40 | 240 | 24.0% | Plowhorse |
| Margherita pizza | 13.00 | 2.60 | 10.40 | 170 | 17.0% | Plowhorse |
| Mushroom risotto | 15.00 | 3.60 | 11.40 | 120 | 12.0% | Plowhorse |
| Pan-fried sea bass | 24.00 | 8.40 | 15.60 | 70 | 7.0% | Puzzle |
| Lamb shank | 26.00 | 9.10 | 16.90 | 40 | 4.0% | Puzzle |
| Caesar salad | 12.00 | 3.00 | 9.00 | 30 | 3.0% | Dog |
Notice what the percentages hide. The ribeye costs 45 percent of its price in ingredients, the highest on the menu, yet it earns more per plate than anything else. The pizza costs only 20 percent, but earns less per plate than the chicken. And the lamb shank from the opening scene earns well and sells rarely: a Puzzle.
What to do with each box
The textbook's advice for each box is sensible and cautious:
- Stars. Leave the recipe and portion alone, make them easy to find, and have staff mention them. A slight price increase is worth testing.
- Plowhorses. Guests love them and they keep the kitchen busy, so change them gently. Raise the price a little, or trim the accompaniments or portion without making the plate feel smaller.
- Puzzles. The problem is visibility, not the dish. Move it higher in its section, describe it better and ask staff to recommend it. Take care not to make it popular but unprofitable.
- Dogs. Usually retire them. Keep one only if it has a job, such as using up leftovers at low labour cost, or being the only vegan main.

How often to repeat it
Run it after every menu change, once the new dishes have a few weeks of sales behind them, and again when supplier prices move enough to change your margins. A review each season is a good rhythm for most kitchens. Change one or two things at a time, so that you can tell what worked.
Putting it to work on a phone menu
On a phone, guests read in scroll order, in one column. That gives you simple levers:
- Order. Put the dishes you most want noticed near the top of their section. Our guide to restaurant menu design for phones explains why.
- Descriptions. Give a Puzzle one or two clear lines about what is in it and how it is cooked, the answer you would give a guest who asked.
- Prices. Keep every price right next to its dish, in the same format. Hiding a price does not raise a Plowhorse's margin, and guests who have to hunt for it notice.
- Highlights. Mark a few dishes you are proud of, not every third one.
A word on "golden triangles", the idea that guests' eyes land on a sweet spot of a printed menu. An eye-tracking study from 2012 found no evidence of such a spot: readers moved through a two-page menu mostly left to right and top to bottom, like a book. The idea is contested, and on a phone a single column has no corners at all.
Mistakes to avoid
- Too little data. One week, or a week with a private party, can mislead you.
- Mixing categories. A dessert is not a Dog because it sells less than a burger.
- Reading it as a verdict. The classic matrix treats dishes as independent of each other, but researchers point out that guests often choose between similar dishes, so removing one moves sales to others.
- Counting only ingredients. A dish that needs hours of prep looks as good as one that takes minutes. Researchers who study restaurant costing argue that managers often lack a full picture of their costs, which leads to inaccurate and sometimes unprofitable prices.
- Retiring a dish that matters. Check why a Dog is on the menu before you delete it.
How Tastiana shows the menu you decide on
Tastiana does not analyse sales or margins, so the sums above stay with your till report. It reads the menus already on your website, PDFs included, and turns them into a guide guests read on a phone without zooming, with search, prices and descriptions exactly as printed, your menu's own dietary and allergen labels, Help me choose suggestions that only use real dishes from your menu, answers to guests' questions grounded in it, and a button to your existing booking service. It checks your website daily and larger changes wait for your approval, and it never changes a dish or a price: you edit your menu as you always do. Try the demo to see it on your own phone.
Questions restaurants ask
What are Stars, Plowhorses, Puzzles and Dogs?
The four boxes of the menu engineering matrix. Stars sell well and earn more than average, Plowhorses sell well but earn less, Puzzles earn well but sell less, and Dogs do neither.
Do I need special software for menu engineering?
No. A sales report showing how many of each dish you sold, plus each dish's price and ingredient cost, is enough, and a spreadsheet can do the sums.
Should I remove every Dog from the menu?
Not automatically. Check why the dish is there first: it may use up leftovers cheaply or be your only vegan main. Retire the ones that have no job.
What is contribution margin?
What a dish earns after its ingredients: the selling price minus the ingredient cost, as an amount of money rather than a percentage.