AI menu engineering: pricing and placement with your own data

AI menu engineering is the classic stars, plowhorses, puzzles and dogs analysis with software doing the sums and suggesting moves. The result rests on two numbers you supply: what each dish costs to plate and how many you sold. With those right, a spreadsheet does most of the work. AI tools help most when ingredient prices change every week.

If you searched "AI menu" for help writing descriptions or laying out a printed menu, general AI writing and design tools can draft both. Check every price and allergen by hand before it prints. The rest of this guide covers the harder part: deciding what to price up, move, promote or cut.

What is menu engineering?

Michael Kasavana and Donald Smith developed menu engineering at Michigan State University around 1982, Kasavana writes in a 2025 series for the American Hotel & Lodging Educational Institute. It judges each dish by contribution margin, the menu price minus the item's direct cost, instead of food cost percentage. He says that cost should include food with waste, in-house prep labor, condiments and packaging.

In part two, Kasavana sets the two cut-offs. An item is popular if its share of items sold is at least 70% of an equal share, which is 70% × 1 ÷ (number of items). Its margin is high if it is at least the menu's average contribution margin: total contribution margin divided by total items sold. Compare items within one section and meal period, such as dinner entrees.

Group	Popularity	Margin	Kasavana's move
Star	High	High	Keep it as it is
Plowhorse	High	Low	Reprice it
Puzzle	Low	High	Move it to a better spot on the menu
Dog	Low	Low	Remove it

What do AI tools add?

They keep plate costs current, answer questions about your sales in plain language, and put chosen items in front of guests on screens. None of them knows your recipes or portions until you enter them.

Tool	What it says it does	What you still supply
xtraCHEF by Toast	Recipe costing from invoices; a Product Mix Report with margins by item and by modifier, and which items are discounted most	Recipes and product mix mapping. Toast notes that retroactive product mix reporting is not possible, so map early
Toast IQ	Flags top sellers by daypart, analyzes the menu against sales and can suggest retiring a dish that is not selling; edits menus from a chat prompt	Plate costs, so the answer covers margin as well as sales
Square AI and Order Guide	Answers questions on item and category performance; Order Guide turns an uploaded menu into ingredient lists and compares supplier prices per unit	Portion sizes and your judgment on which answers to act on
MarginEdge	Updates ingredient prices as invoices are processed; shows theoretical menu costs from those prices and product mix. Its AI help page describes a built-in assistant and a connector to ChatGPT, Claude and Gemini	Recipes and invoices
Toast Menu Upsells	AI suggestions of high-margin items and add-ons on handhelds, kiosks and guest-facing displays, added in May 2026	The list of items you are willing to push
Deliverect	Delivery-app menu management with dynamic pricing and A/B testing	Your price rules for each app

Voice ordering is a placement channel too. A phone or drive-thru agent offers one add-on at the moment of the order. Maple Pro lists intelligent upselling on phone orders, and Maple's drive-thru unit offers an upsell the restaurant chooses on every order. Pick that item from your stars and puzzles, then measure it with the upselling pilot guide.

What does a worked example look like?

The numbers below are made up for an eight-item burger counter over four weeks. They show the method and do not come from a real restaurant. Plate cost includes the bun, sides, sauce and packaging.

Item	Price	Plate cost	Margin	Sold	Share of items sold	Group
Classic burger	9.50	3.40	6.10	620	31.0	Plowhorse
Double bacon burger	13.00	4.90	8.10	310	15.5	Star
Chicken sandwich	10.00	3.10	6.90	410	20.5	Star
Salad bowl	10.50	3.90	6.60	230	11.5	Star
Veggie burger	11.00	3.30	7.70	90	4.5	Puzzle
Kids meal	7.00	2.80	4.20	150	7.5	Dog
Loaded fries	7.50	1.90	5.60	120	6.0	Dog
Fish sandwich	12.00	5.60	6.40	70	3.5	Dog

Prices, costs and margins are in dollars; share is a percentage. The popularity cut-off is 70% × 1 ÷ 8 = 8.75% of items sold. Total margin is $13,083 across 2,000 items, so the average margin is $6.54.

Loaded fries have the lowest food cost here, about 25% of price, and still land as a dog, because each order earns $5.60 and few people buy them. Kasavana warns that promoting low food-cost items often lowers the average check and gross profit.

Plowhorse: the classic burger sells most but earns less than average. A $0.50 rise takes its margin to $6.60, but the burger's own gain lifts the menu average to about $6.70, so it stays a plowhorse. A $1.00 rise takes its margin to $7.10 against a new average of about $6.85, if cost and sales held. Recalculate the average after every price change, because the item you reprice moves it. Test the rise for four weeks and recount.

Puzzle: the veggie burger earns well but few people order it. Move it to the top of its section, give it a better description, and make it the phone or kiosk suggestion for a month.

Dogs: the fish sandwich costs 47% of its price and sells least, so reprice it or cut it. The kids meal is a judgment call: it may bring in families who buy stars. Keep it only if you believe that, and check its cost.

Where does AI pricing go wrong?

Guests notice price moves. In February 2024 Wendy's CEO said the chain would use digital menu boards to test "dynamic pricing" and AI-enabled menu changes. After a backlash, Wendy's said it would not raise prices when customers visit most, Restaurant Dive reported. It said the boards could offer discounts in slower times of day.

Keep price changes few and deliberate. Make one change per item, give it four weeks, and count before the next one. If a tool suggests a price, ask it which sales and costs it used, and check them against your own matrix.

What should you check before changing the menu?

1. Pick one section and one meal period, such as lunch entrees.
2. Export four weeks of item counts from your POS.
3. Cost every plate with this week's invoices, including sides, sauces and packaging.
4. Work out each item's margin and share of sales, the 70% popularity cut-off and the average margin.
5. Group every item as a star, plowhorse, puzzle or dog.
6. Choose one move per item: keep, reprice, move or remove.
7. Make the change on every channel that sells the item (printed menu, website, kiosk, delivery apps and phone agent) unless you price a channel differently on purpose.
8. Recount after four weeks and compare the groups.
9. Only then let an AI tool's suggestions drive changes, and check each one against your matrix.

If a phone agent will sell the new menu, the menu preparation worksheet helps you load items, modifiers and the upsell you picked.

Published by Maple. This AI-assisted guide combines Michael Kasavana's published method, vendor pages and help centers, company press releases and a trade news report with an original worked example and checklist. The example uses made-up numbers. It does not report a test we ran or promise any sales or profit result.