Restaurant Menu Engineering: A Practical Guide

Restaurant menu engineering is a way to judge every dish on two numbers: how much money it leaves after its own cost, and how often guests order it. You sort each dish by those two numbers, then decide what to keep, reprice, move or drop. The data is already in your POS. A spreadsheet does the rest.

We build restaurant websites through our restaurant website design service, so weigh that. Every fact below about menu engineering was checked against Michael Kasavana’s own articles, Cornell’s hospitality research and the National Restaurant Association in October 2026.

Short answer: for each menu category, work out each item’s contribution margin (price minus the item’s direct cost) and its share of items sold. An item is popular if its share is at least 70% of an equal share. Its margin is high if it beats the weighted average margin for the category. That gives four groups. Stars are retained, plowhorses repriced, puzzles repositioned and dogs removed. Then check the results again next month.

The Menu Engineering Matrix at a Glance

As of October 2026, from Michael Kasavana’s own description of the method:

Sources: Kasavana, The Power of Menu Engineering, Part Two, Kasavana, Part One.

What Is Menu Engineering (And Why It Pays More Than a Price Increase)

Menu engineering was developed by Michael Kasavana and Donald I. Smith at Michigan State University’s School of Hospitality Business, around 1982 (source). Their book, Menu Engineering, came out from Hospitality Publishers in Lansing, Michigan that year (source).

Kasavana describes it as a mix of psychology, managerial accounting, marketing and graphic design (source). The accounting part tells you what each dish earns. The other parts help you change what guests order.

Does it pay more than a price increase? Not always, and nobody can promise you a number. The honest claim is narrower. A blanket price rise treats every dish the same. Menu engineering lets you act on one dish at a time. You might raise the price of a busy dish that earns little, give a better spot to a dish that earns well but sells slowly, and drop a dish that does neither. That is more targeted than moving every price at once.

The Business Case for Menu Engineering

Ingredient costs move, and they don’t all move together. The National Restaurant Association reported that wholesale food prices edged down 0.1% from July to August 2026 and sat 1.9% below a year earlier. They were still more than 33% above February 2020 levels (source).

Inside that average, single items swung hard. Fats and oils were up 21.2% on the year. Eggs were down 59.4%. The Association noted that how much relief a restaurant gets depends on its own menu mix (source). That is the case for menu engineering in one line: your costs depend on what you sell, so you need to know what each dish earns.

Kasavana lists the benefits as better profit margins, better cost management and less food waste (source). He also says recipe costs should be updated whenever the menu is re-engineered (source). With costs moving like this, an analysis built on last year’s costs can point you the wrong way.

The Two Variables That Drive Everything

Most operators watch food cost percentage. Kasavana defines it as an item’s ingredient cost, including garnishes and sides, divided by its selling price (source). It is useful for purchasing and portion control. It is not the number menu engineering ranks dishes by.

Menu engineering uses two other numbers:

• Contribution margin (CM): the menu price minus the item’s direct cost. It is the money each sale leaves to cover labor, rent and profit (source).
• Menu mix (MM): each item’s share of the total items sold in its category. It is the measure of popularity (source).

Kasavana’s own example shows why food cost percentage can mislead. A chicken dish that costs $1 and sells for $3 runs a 33% food cost. A steak that costs $2.50 and sells for $5 runs 50%. The chicken looks better on food cost. The steak leaves $2.50 per plate against the chicken’s $2 (source).

He adds a warning. Items with a low food cost percentage tend to be cheaper and lower in contribution margin, so pushing them can pull your average check down. A low food cost and a high gross profit can go together, but one doesn’t guarantee the other (source).

How to Calculate Contribution Margin and Item Popularity

The math is subtraction, division and a little multiplication. You need two inputs per item: how many sold, and what one portion costs you to make.

Kasavana says to compare items only against items they compete with. That means appetizers against appetizers and entrées against entrées. Keep lunch and dinner apart too. He notes the data can be averaged over a four-week cycle (source). A guest choosing a starter is not choosing between a starter and a dessert, so mixing categories would distort both numbers.

Contribution Margin: The Formula Every Operator Needs

Contribution margin = menu price minus the item’s direct cost

The hard part is the cost, not the formula. Kasavana says item costs should ideally include (source):

• food, including waste and loss,
• the extra labor the item needs,
• condiments,
• packaging.

Many operators start with food cost only, because that is what their recipe cards hold. That is a fair first pass. Just know that a dish with heavy prep or takeout packaging earns less than its food cost alone suggests.

Once you have CM per item, multiply it by the number sold. That gives the item’s total contribution for the period. Add those up across the category and you have the category’s total CM. You need that number for the margin test below.

Item Mix Percentage: How to Find Your Popularity Threshold

Menu mix % = items of this dish sold ÷ total items sold in the category × 100

The popularity line in Kasavana’s method is the 70% rule. If every item sold equally, each would get a share of 1 divided by the number of items. An item counts as popular if its menu mix is at least 70% of that equal share. With four items, the equal share is 25%, so the line is 17.5% (source).

The margin line works differently. It is the weighted average contribution margin: the category’s total CM divided by the total number of items sold. Kasavana’s example is $3,444.80 of total CM over 1,000 items, an average of $3.44. Items at or above that average count as high margin (source).

Note that the margin line is weighted by sales. A simple average of the item margins would treat a dish sold ten times the same as a dish sold four hundred times.

A Worked Example (Our Own Arithmetic)

The numbers below are illustrative inputs we made up to show the method. They are not data from a real restaurant. Four entrées, sold over one four-week period:

Total sold is 880. Total CM is $10,775 ($3,800 + $4,650 + $1,485 + $840). So the weighted average CM is $10,775 ÷ 880 = $12.24. With four items, the popularity line is 17.5%.

Look at the burger. It is the best seller, yet it earns the least per plate. Its food cost is about 41%, against 25% for the pasta. A price rise from $16 to $17 lifts its CM to $10.50. At the old price it brought in $3,800 a period. At the new price it needs to sell about 362 to bring in the same amount. So in this made-up case, the burger could lose up to 38 orders a period before the change cost money. Your own numbers will differ. Run the same check before you change a price.

The Menu Engineering Matrix: Stars, Plow Horses, Puzzles, and Dogs

With both lines drawn, every item lands in one of four boxes. Kasavana calls stars “winners”, plowhorses “marginal”, puzzles “potential” and dogs “losers”. His short list of moves is: stars retained, plowhorses repriced, puzzles repositioned and dogs removed (source).

Cornell’s Sheryl Kimes describes the same idea with names borrowed from a well-known business matrix: stars, cash cows, question marks and dogs. In her version, stars and cash cows are candidates for price increases, and question marks may suit a price cut (source). Two experts, slightly different advice. Treat the moves as starting points, not rules.

Stars: Your Menu’s Best Performers

Stars sell well and earn well. Kasavana’s advice is to retain them (source). Don’t change the recipe, the portion or the name without a reason.

Retaining doesn’t mean ignoring them. Kimes lists stars as candidates for a price increase (source). If guests already choose a dish often, a small rise may hold up. Test it on one star at a time and watch its count in the next period.

Keep stars easy to find on every version of your menu: print, website and ordering page. Make sure your team knows which dishes they are.

Plow Horses: Popular but Draining Your Margin

Plowhorses sell well but earn little per plate. Kasavana’s move is to reprice them (source). Kimes also lists high-volume items as candidates for a price increase (source).

Repricing is not the only lever. You can also lower the item’s cost. Common ways include a smaller portion, a cheaper side or a different supplier. Any of these raises CM without touching the price. Before you pick one, do the break-even check from the worked example above. It tells you how many orders the dish can lose before the change backfires.

Change one plowhorse at a time. If you change three things at once, you won’t know which one moved the numbers.

Puzzles: Your Hidden Profit Opportunities

Puzzles earn well but sell slowly. Kasavana’s move is to reposition them (source). In practice that can mean a better spot on the menu, a clearer description, a new name or a mention from your servers.

Kimes suggests that question marks, her name for puzzles, may suit a price cut (source). A small cut can still leave a puzzle above the average margin. Check the math before you try it.

A puzzle may also be slow for a reason the menu can’t fix. It may be too expensive for your guests, or a style they don’t come to you for. If repositioning doesn’t move it after a period or two, treat it as a candidate for a rethink.

Dogs: What to Do When an Item Is Not Working

Dogs sell slowly and earn little. Kasavana’s move is to remove them (source). Removing items is also first on his list of common menu-engineering strategies, next to limiting the number of options (source).

Ask a few questions before you cut a dog:

• Is it the only option for a dietary need that brings a whole table in?
• Does it share ingredients with stars, so it costs little extra to keep?
• Could a new price or a cheaper build turn it into a puzzle?
• Is it a signature dish that guests link to your name?

If the answers are no, take it off. A shorter list means fewer ingredients to stock and less to prep. Kasavana points to less food waste as one of the benefits of the method (source).

Menu Psychology: How Placement and Design Drive Orders

The matrix tells you which dishes to push. Menu design is how you push them. Kasavana lists the tactics that come up most often. They include changing prices, editing dollar signs and decimals, and featuring items with graphics or bold type. They also include enticing descriptions, using eye movement, limiting options and limiting the number of panels (source).

Be careful with the research here. A 2009 Cornell report noted that consultants and designers use many layout and copy tactics. It said there was some evidence linking them to attention, but no proven link between attention and what people buy (source). So treat design moves as tests. Your own sales data is the judge.

The Golden Triangle and Eye Magnets

The “golden triangle” is the old idea that guests look at a fixed set of spots on a menu first, often said to be the center and then the upper corners. That is where dishes you want to sell are usually placed.

The best study we could open does not back it up. A San Francisco State University eye-tracking study by Sybil Yang found that diners read menus in order, like a book. It found no evidence of menu “sweet spots”, including the spot often claimed just above the center of the right-hand page (source). The study was published in the International Journal of Hospitality Management in 2012.

What does that mean for you? Don’t rely on a magic corner. If guests read in order, the first item in each section and the item you set apart are likely to get seen. That is a reasonable test, not a law.

“Eye magnets” are boxes, bold type, icons or photos used to pull attention to one item. Kasavana lists featuring items with graphics or bold type as a common strategy (source). We found no source we could open that measures how much they lift sales. One plain rule of design still applies: if everything is boxed, nothing stands out. Use them for one or two puzzles or stars per section, then check the counts.

Pricing Architecture: The Psychology of Numbers

The best-known study here is from Cornell. Researchers tested three price formats on lunch menus at St. Andrew’s Café, at the Culinary Institute of America in Hyde Park, New York. The formats were “$00.00”, plain numerals like “00.” and prices written out in words. The test ran in 2007, and the analysis covered 201 tables (source).

Guests with the numeral-only menu spent significantly more than guests whose menus showed a dollar sign or written-out prices. The researchers had expected the written-out format to do best. Read the limits, though. Compared with party size and time at the table, the price format did not significantly change total spending. The authors also said the findings may apply only to lunch at that one restaurant (source). The “pain of paying” idea behind the test was a hypothesis, not a proven cause.

Price endings are another common tactic. The same report cites Naipaul and Parsa. They found guests saw a quick-service menu as better value when prices ended in 9. They saw fine-dining menus as higher quality when prices ended in 0 (source). That study measured attitudes, not purchases.

Price anchoring is a third. You place one high-priced dish in a section so the others look moderate. It is widely repeated, but we did not find a menu study we could open that measures it. If you try it, the anchor should still be a dish you are happy to sell.

Descriptive Language That Sells

The 2009 Cornell report also cites work by Brian Wansink and colleagues in a college faculty cafeteria. Descriptive labels raised how often an item was bought and how satisfied people were with it. They did not raise what people were willing to pay. Guests liked “Grandma’s Zucchini Cookies” more than plain “zucchini cookies”, though the cookies were identical (source).

There is a caveat. In 2018, Cornell said a faculty committee had found Wansink committed academic misconduct in his research, including misreporting research data (source). The statement does not single out the menu-label paper. Still, treat that result as a weak signal and test descriptions on your own menu.

Kasavana lists enticing descriptions among the common menu tactics (source). A practical approach is to rewrite the description of one puzzle at a time:

• say how it is cooked, in plain words,
• name a specific ingredient or source only if it is true,
• keep it short enough to read on a phone,
• leave the price and portion unchanged, so the description is the only thing you are testing.

Then compare the item’s count in the next period with the one before.

Your Digital Menu Needs the Same Engineering

Your menu now lives in several places: the printed card, your website, your online ordering page and your listings. The matrix doesn’t care which one a guest uses. A star should be easy to find on all of them. A dog you removed from print should be gone online too.

Kasavana counts graphic design as one of the fields menu engineering draws on (source). On a website, the design choices are things like section order, which items get a photo and how prices are shown. Apply the same rules:

• put stars and the puzzles you are testing near the top of each section,
• use photos for high-margin dishes you want to sell, not just the easiest dish to shoot,
• write descriptions online that match the print menu,
• keep the online menu in HTML, so it is easy to update the day you change it.

Where orders arrive also matters. A dish’s contribution margin is calculated before any channel fees. If you send online orders through a marketplace that takes a commission on each order, part of the margin you engineered goes to the marketplace. A direct ordering system for your restaurant keeps those orders on a channel you control.

Online ordering also gives you clean data. Item counts from your own ordering page are easy to export and drop into the same spreadsheet as your POS data. Keep online and in-house counts in separate columns at first. Guests may order differently for delivery than at the table.

If your restaurant’s website still shows the menu as a PDF or an old photo of the printed card, fix that alongside the print work.

How to Run a Menu Engineering Analysis in One Afternoon

For one category, the arithmetic fits in an afternoon if your POS can export item counts. Costing every recipe from scratch takes longer, so do that first if your recipe cards are out of date.

1. Pick one category and one meal period. Entrées at dinner is a good start. Kasavana compares only competing items, by meal period (source).
2. Export items sold for a recent period. Kasavana notes data can be averaged over a four-week cycle (source). Avoid a period with a holiday or a closure unless that is what you want to study.
3. Cost each item at today’s prices. Include waste, extra labor, condiments and packaging where you can (source). Ingredient prices move unevenly, so old costs can mislead (source).
4. Work out CM per item and total CM. Price minus cost, then times the number sold.
5. Work out menu mix and the 70% line. Divide each item’s count by the total. The line is 70% of 1 divided by the number of items (source).
6. Work out the average CM and sort. Total CM divided by total items sold. Each item is high or low on each number, which puts it in one of four boxes.
7. Pick a small set of changes and check next month. Kimes notes that restaurants using menu engineering generally review their results each month (source).

For step 7, a sensible first round is one plowhorse to reprice or recost, one puzzle to reposition and one dog to cut. That is our suggestion, not a rule from the research. Keeping the round small makes the next month’s numbers easier to read.

Common Mistakes Independent Restaurants Make With Menu Engineering

Most of these mistakes come from skipping a step in the method, not from bad intentions.

Ranking dishes by food cost percentage. Kasavana’s chicken and steak example shows a dish with a higher food cost percentage can leave more money per plate (source). Use food cost percentage for purchasing. Use contribution margin to sort the menu.

Mixing categories. Comparing a side salad with a steak puts every cheap item in the “low margin” box. Analyze each competing category on its own (source).

Using stale recipe costs. Costs shift by commodity, and the shift depends on your own menu mix (source). Update costs each time you re-engineer (source).

Changing many things at once. If you reprice, rename and move a dish in the same week, you can’t tell which change worked. Change one thing per item per period.

Keeping dogs out of habit. “It has always been on the menu” is not a reason on its own. Run the questions in the dogs section. If the dish fails them, remove it.

Treating design tips as proven. The golden triangle, eye magnets and price anchors are widely repeated. The research we could open is mixed or thin (source, source). Test them on your menu and let the counts decide.

Engineering the print menu and forgetting the online one. Guests who order online never see your printed card. Apply the same changes to your website and ordering page.

Start with the Math, Then Make Your Menu Work for You

Menu engineering is an accounting exercise first and a design exercise second. Get the numbers right before you touch the layout. Pick a category. Cost each item at today’s prices. Work out contribution margin, menu mix and the two lines. Sort the items into four boxes. Make a few changes, then check the counts next month.

The design side matters, but the evidence for most layout tricks is weaker than the blog posts suggest. The Cornell price-format study is a careful test, and even its authors said it may apply only to one lunch service. Your own before-and-after numbers are better evidence than anyone’s rule of thumb.

Once the menu is earning more per guest, it helps to bring those guests back. Restaurant email marketing is a direct way to tell regulars about a new dish or a seasonal change. If you’d like help putting your engineered menu online, talk to us about your restaurant’s digital presence.

Related Posts

• How to Build a Restaurant Menu Page That Drives Reservations: putting your menu on your own site
• Restaurant Mobile Menu Design: What Actually Works on Phones: how the menu reads on a small screen
• The Real Math Behind Restaurant Delivery App Commissions: what commissions do to each order
• Prix Fixe Menu Marketing: pricing and filling a set menu

FAQ

What is menu engineering in a restaurant?

It is a method for judging each menu item on two numbers: contribution margin and popularity. Items are sorted into stars, plowhorses, puzzles and dogs, and each group gets a different action (source).

Who created menu engineering?

Michael Kasavana and Donald I. Smith developed it at Michigan State University’s School of Hospitality Business, around 1982 (source).

How do you calculate contribution margin on a menu item?

Take the menu price and subtract the item’s direct cost. Ideally the cost includes food with waste, extra labor, condiments and packaging (source). Multiply by the number sold to get the item’s total contribution.

What is the 70% rule in menu engineering?

An item counts as popular if its share of sales is at least 70% of an equal share. With four items in a category, that line is 17.5% (source).

How often should you re-run a menu engineering analysis?

Kimes notes that restaurants using menu engineering generally review results each month (source). Kasavana also says recipe costs should be updated whenever you re-engineer the menu (source).

Does removing the dollar sign from menu prices really work?

In one Cornell test at a lunch restaurant, guests with numeral-only prices spent significantly more than guests with dollar signs or written-out prices. The authors said the result may apply only to that setting (source). Test it on your own menu before relying on it.

The numbers tell you what to change. A restaurant website with an HTML menu you can update the same day lets those changes reach guests online too. Get in touch to talk it through.