Menu Engineering for Restaurants: Matrix, Math and Template

Menu engineering is a method for sorting every menu item by popularity and contribution margin into Stars, Plowhorses, Puzzles and Dogs, then repricing, repositioning, promoting or cutting each one accordingly.

This guide is for an owner or manager who wants to run the analysis this week: the matrix, the math, an 8-item worked example, where the data comes from, what to change in each quadrant, how commissions change the answer, and a Google Sheets template. It is not a food cost tutorial (see the restaurant food cost percentage guide), not a guide to growing the check (see how to increase restaurant average order value), and not about layout or photos.

What is menu engineering?

Menu engineering is a way of scoring every item in a menu category on two numbers, how often it sells and how much money it leaves after ingredient cost, then placing each item in one of four boxes so you know whether to keep, reprice, promote or cut it.

The method comes from Michael L. Kasavana and Donald I. Smith of Michigan State University's School of Hospitality Business, who published Menu Engineering: A Practical Guide to Menu Analysis in 1982; the American Hotel and Lodging Educational Institute's summary of the method credits the same authors, school and year. A 1997 paper in the International Journal of Contemporary Hospitality Management, hosted by Cornell's hotel school, calls it a tool that "labels menu items within a competing menu group using their respective popularity and contribution margin". That last phrase matters: compare entrees with entrees, never a $4 side against a $32 steak.

Three things it is not:

• It is not menu design. Item order, boxes, descriptions and photos come after the analysis says which items deserve them.
• It is not food cost percentage control. A $12 salad with $3 of ingredients runs 25% and leaves $9; a $30 steak with $12 of ingredients runs 40% and leaves $18. As the 1997 paper puts it, "it is dollars that are deposited in the bank, not percentages."
• It is not a one-time project. Prices, plate costs and habits move, so the labels expire.

One blind spot: the matrix ignores labor, so an eight-minute plate and a ninety-second plate can share a box. The authors of that 1997 paper had added a labor axis in 1995; for most independents, keeping prep time in mind when choosing what to promote is enough.

What are the four menu engineering categories?

The four menu engineering categories are Stars (popular and high margin), Plowhorses (popular but low margin), Puzzles (high margin but unpopular) and Dogs (unpopular and low margin), and each has a standard first move: keep Stars, reprice Plowhorses, reposition Puzzles, remove Dogs.

AHLEI's second article on the method puts the four moves in one line: "star items should be retained, plowhorse items repriced, puzzle items repositioned (on the menu), and dog items removed." The rest of this guide is about getting the labels right.

How do you calculate the menu engineering matrix?

The menu engineering matrix uses four formulas: contribution margin is menu price minus plate cost, menu mix is an item's units divided by the category's units, the popularity threshold is 70% of an even split, and the margin threshold is total contribution dollars divided by total units sold.

For one category with N items:

Contribution margin. Menu price minus plate cost, where plate cost is the recipe cost of one portion including garnish and, for takeout, packaging. Not labor, not rent.

Menu mix. The item's units divided by the category's total units over the same period, as a percentage.

Popularity threshold. 100% divided by N, times 0.70. With 8 items an even split is 12.5% and the threshold is 8.75%; with 12 items it is 5.83%. The 0.70 factor is the Kasavana and Smith convention, used by the 1997 paper and AHLEI's worksheet alike: an item is popular when its share "exceeds 70 per cent of the average popularity for the group".

Average contribution margin. Total contribution dollars (each item's margin times its units, summed) divided by total units sold. It is a sales-weighted average, and the weighting matters: a simple average lets a $32 steak that sold 90 times move the bar as much as a $14 burger that sold 520 times.

Popularity is High if the mix is at or above the threshold; margin is High if the contribution margin is at or above the average. High on both is a Star; High popularity with Low margin is a Plowhorse; Low popularity with High margin is a Puzzle; Low on both is a Dog.

Worked example: an 8-item entree category

A made-up 8-item entree category over four weeks. Round, illustrative numbers, not benchmarks.

Category totals: 1,890 units and $19,560 of contribution, so:

• Popularity threshold: 100% divided by 8 is 12.5%, times 0.70 is 8.75%, or about 166 of the 1,890 units.
• Average contribution margin: $19,560 divided by 1,890 is $10.35.

A simple average of the eight margins would be $11.00 and would put the fish tacos exactly on the line; the weighted figure reflects what sold.

What it tells this owner:

• Three Plowhorses carry 65% of the volume at below-average margins. A $1 rise on each, if volume holds, adds $1,230 every four weeks, roughly $16,000 a year, without touching a recipe. (The average rises with the prices, so they may still read as Plowhorses next quarter; that is normal.)
• The fish tacos are a Star by a whisker, 9.0% against 8.75%. Leave the recipe alone and give them the first slot in the category.
• The ribeye is a Puzzle: $18 a plate, sold 90 times. Forty more in four weeks is another $720. It needs a better description, a server who mentions it, and a check on whether $32 fits the neighborhood.
• Pasta primavera is a Dog that earned $570 in four weeks. Cut it. The veggie bowl is also a Dog, but if it is the only vegetarian entree, keep it and rebuild the plate until it clears $10.35.

Where do the numbers for menu engineering come from?

The units come from your POS product mix report over 4 to 8 full weeks, the prices come from the menu on the channel you are analyzing, and the plate costs come from recipe costing; rerun it quarterly and after any large ingredient price change.

Units sold. In Toast it is the Product Mix (PMIX) report under Reports, then Menus, then Product mix. Toast's help article says it shows quantity sold and net item amount, filters by date range and order source, exports to Excel or CSV, and adds cost of goods sold and gross profit per item if you have loaded recipe costs. Square's equivalent is Reports, then Custom, then Custom Item report; Square's help center says it lists quantity sold, gross and net sales per item for any date range, with CSV export. Other POS systems have an equivalent; the column you want is units sold, not dollars.

How much history. Four to eight whole weeks. Fewer and one busy Saturday skews the mix; more and you blend seasons and old prices. Skip weeks with a holiday, a closure or one large catering order, and use net quantities so voids are out.

Plate cost. A cost per portion for each item, from the recipe and current invoice prices. The item-level walkthrough is in the food cost guide. POS cost columns are only as good as the recipes you loaded, so spot check the top five sellers against a current invoice.

How often. Quarterly. AHLEI pairs "quarterly analysis of effectiveness" with menu revisions "done twice per year": the numbers move every quarter, the menus change about twice a year. Also rerun after any large ingredient price move, because a $2 jump in the cost of a patty changes the burger's margin and the category average with no change in sales. The 30/30/30 rule post shows how far food cost has drifted above the old 30% line; that drift is what a quarterly rerun catches.

What do you do with each quadrant?

The standard moves are to keep and feature Stars, reprice or cost-trim Plowhorses, promote or rework Puzzles, and cut Dogs, using four levers (price, plate cost, menu position and staff suggestion) in a different mix for each box.

Three of those levers need a note.

Position. Menu consultants have long claimed a "sweet spot" above the center of the right-hand page. The best-known eye-tracking test of that claim, by Sybil Yang at San Francisco State University (International Journal of Hospitality Management, 2012), found no such spot: diners "read menus sequentially like a book", left to right, top to bottom, and slowly. Do not pay for a redesign built around a sweet spot. The safe version: the first item in a section is read first, on paper and on a phone, so that slot belongs to a Star or a Puzzle, never to a Plowhorse that sells anyway.

Price. Raising a Plowhorse is the highest-value move on most menus and the easiest to overdo. Small steps, one category at a time, and read the next product mix report before the next step; if volume falls faster than the price rose, you found the ceiling. Pairing a Plowhorse with a high-margin side is the other route, and that is check-building work, covered in the average order value guide.

Staff. The 1997 paper lists "train the staff on the principles of contribution margin" as a core technique: guests ask what is good, everything is good, so servers might as well name the items that pay. Give them two Stars and one Puzzle to suggest by name, and rotate the Puzzle monthly.

Does menu engineering work for online menus and delivery apps?

Menu engineering works on online menus and delivery apps, but you have to run a separate matrix for each channel, because the sales mix differs by channel and a marketplace commission comes off every item's price before you calculate contribution margin.

Two things change off-premise. The mix: what people order at a table is not what they order for delivery, so a dining-room Puzzle like the ribeye may be a Dog on a delivery app where it travels badly. Pull the product mix report filtered by order source (Toast has that filter) and build one sheet per channel: dine-in, your own website, each marketplace.

The margin: DoorDash's published merchant plans charge 15% (Basic), 25% (Plus) or 30% (Premier) on delivery orders and 6% on pickup; the other apps' rates are in the delivery app commission rates post. Here is what that does to the fish tacos, at $16 with a $5 plate cost:

A direct order still pays card processing, a much smaller line, and off-premise plate cost should include packaging on every channel. The commission is a percentage of price, so it takes the most dollars from your highest-priced items, and the category average falls with it. With a different mix and a different average, labels change: a dine-in Star can be a Plowhorse or a Dog on the delivery sheet. Whether or not you raise marketplace prices to offset the fee, build each channel's matrix on that channel's prices, mix and fee. Restaurant profit margins run in the single digits for most independents, so losing $4.80 of an $11 margin on one channel is not a rounding error.

A menu engineering spreadsheet you can rebuild in 20 minutes

Eleven columns and two helper cells in Google Sheets. Type the first four from your product mix report and recipe costs; the rest is formulas.

The two helper cells, off to the right so they stay out of the data:

• N2, popularity threshold: =0.7/COUNTA($A$2:$A$9)
• N3, average contribution margin: =SUM($G$2:$G$9)/SUM($B$2:$B$9)

Change the 9 in every range to your last row. Then:

Step 1. Export the product mix report for one category and one channel, whole weeks only, as CSV, and paste names and units into A and B.

Step 2. Type menu price and plate cost into C and D. If the POS holds recipe costs, export those too and check the top five against an invoice.

Step 3. Fill E through J down, then sort by J and by G from high to low, so Stars and the biggest Plowhorses sit at the top.

Step 4. Write one action per item in K with a date: small price steps on Plowhorses, a description and slot change on Puzzles, a cut date on Dogs.

Step 5. Duplicate the tab per category and per channel, and again next quarter. Keep the old tabs; the quarter-to-quarter comparison is where you learn whether the changes worked.

Where DirectOrders fits

Every commission row in the channel table disappears on an order placed through your own website, which is what DirectOrders is built for: commission-free online ordering for independent restaurants on a flat monthly fee (Pro $249 per month, Pro + Voice $349 per month with 500 AI phone minutes included, 14-day free trial). On that channel an item's margin is the full menu price minus plate cost, less card processing, so your dine-in labels mostly hold. Menu Brain, the menu search layer with nutrition and allergen clarity, helps a customer searching for a gluten-free or high-protein dish find the Puzzle you want found. It works alongside Toast, Square, Clover, Lightspeed and Revel, so the product mix report you already pull still feeds the sheet.

Sources

All sources checked September 2026.

• Kasavana and Smith, Menu Engineering: A Practical Guide to Menu Analysis, Hospitality Publications, 1982 (Google Books record)
• AHLEI, The Power of Menu Engineering, Part One (origin, categories, cadence)
• AHLEI, The Power of Menu Engineering, Part Two (70% threshold, average margin, four moves)
• LeBruto, Ashley and Quain, Using the Contribution Margin Aspect of Menu Engineering to Enhance Financial Results, International Journal of Contemporary Hospitality Management 9(4), 1997 (competing group, 70% rule, weighted average, staff, labor)
• EurekAlert, "Do menu 'sweet spots' really exist?", San Francisco State University, 2012 (no sweet spot, sequential reading)
• Toast Central, Product Mix (PMIX) Report Overview (report path, filters, export)
• Square Support, View item, category and modifiers sales reports (report path, columns, export)
• DoorDash merchant pricing (commission by plan, pickup rate)

Bottom line

Menu engineering is two numbers per item, menu mix and contribution margin, against two thresholds, 70% of an even split and the sales-weighted average margin. Every item lands in one of four boxes with a standard move: keep Stars, reprice Plowhorses, promote Puzzles, cut Dogs. Pull 4 to 8 weeks of your product mix report, cost the recipes, build the sheet above, and rerun it quarterly and after any big ingredient price move. Run it once per channel, because a 15% to 30% commission changes every margin before you have made a single decision.