Menu Design That Sells: Engineering, Layout, and the Mistakes Restaurants Make

A menu that sells is built in a fixed order. Work out which items make money and which ones actually move. Lay the page out so the reader meets the profitable ones first. Write descriptions that make those items sound like something. Print the price in a format that does not sting. Then choose a material that matches how often the prices change. Fonts and colors come last, and they only work if the first five steps happened.

Here is that method as of September 2026, with the research behind each step and, where an older menu of mine got something wrong, what I would change. The menus in the WBX portfolio are mine, from a seafood grill's two-page lunch menu to digital boards for a juice bar and a burger counter, several from my years at This Creative in Fort Myers and the vape shop boards from Nashville before that.

How to think about it

Start with the margin. The National Restaurant Association's 2026 State of the Restaurant Industry report (February 2026) puts 2024 median pre-tax income at 2.8% for full-service and 4.0% for limited-service restaurants, down from 4% and 6% in 2019, with food costs up 38% since 2019 and 42% of operators reporting they were not profitable in 2025. The menu is the one lever an operator can pull every day, and usually the one they touch last.

Menu engineering is the discipline for pulling it. Kasavana and Smith published the matrix at Michigan State in 1982, and it has not needed much revision since. Plot every item on two axes: popularity, meaning its share of the sales mix, and contribution margin, meaning price minus food cost. The popularity threshold is 70% of an even share, so on a 10-item section anything under 7% of orders counts as unpopular; the margin threshold is the section's average contribution margin.

Quadrant	Popular	High margin	What to do
Star	Yes	Yes	Protect it: first in its section, described, never buried
Plowhorse	Yes	No	Fix the cost or the portion before you touch the price
Puzzle	No	Yes	Reposition and describe it; it is a star nobody has noticed
Dog	No	No	Remove it unless it is there for a reason, like a kids' item or an allergen option

Everything a designer does after this is in service of that table. A beautiful menu that leads with a dog is a beautiful mistake.

Step 1: Do the math before you open the design file

Pull the sales mix from the POS for a normal stretch of weeks, not a holiday. Cost every item from the recipe, not from memory. Subtract to get the contribution margin, classify each item, and do it section by section, because a $9 appetizer and a $34 entree cannot share an average. Hand the designer the sheet with four labels on it. The brief writes itself from there: which items get position and words, which get quiet placement, which disappear.

Step 2: Where do diners actually look on a menu?

Not where the folklore says. The golden triangle, the sweet spot at the upper right, and the "109 seconds" attention window do not survive the research. Sybil Yang's 2012 study in the International Journal of Hospitality Management put diners in front of an infrared eye tracker and found they read a menu "sequentially like a book," found no evidence of a sweet spot, and did find a sour spot: the restaurant information and, in that study, the salads. The 109-second figure traces to a 1987 Gallup report whose method was never published and whose primary source I could not locate, so I do not build menus on it. A 2021 vendor eye-tracking study proposes a "T" pattern; it is not peer-reviewed, so treat it as a hypothesis.

Sequence, not sweet spots. So I lead each section with a star or a puzzle, the first thing a reader meets, and keep sections short enough to read as a list rather than a search. The story of the restaurant, the founder's grandmother, and the three addresses go somewhere other than the ordering surface. On the Little Caesars board in the portfolio, the top band is three store addresses and phone numbers, which earns its space on a piece that has to work as a directory for a multi-store operator. On a single-location table menu, that band is the sour spot; the address belongs on the back.

The burger board I designed for BWA does the other thing right: a "most popular" badge sits on the double patty, second in the burger list. Whatever the math said that month, the badge does a star callout's job: it tells an undecided guest what to order.

Step 3: Write descriptions that do work

The best-known study on menu language is Wansink, Painter, and van Ittersum's six-week cafeteria field study, published in the Cornell Hotel and Restaurant Administration Quarterly in 2001. Six items, about 140 customers, and labels that were geographic, nostalgic, or sensory raised sales of those items by 27% and improved attitudes toward the item and the restaurant. Diners did not tolerate a price premium for the label, and misleading labels backfired. Two cautions: it is a small study, and the lead author's later work has been retracted, though not this paper as far as I can find. Take the 27% as direction, not a promise.

Direction is enough. Describe the stars and the puzzles. Leave the plowhorses plain. Use three kinds of words: where it comes from, what it reminds you of, and what it tastes like. Never an ingredient list; that is a recipe, not a sale. The BWA board's "a holy union of two patties" is nostalgic and funny. On the FishTale lunch menu I designed in 2016, the item names are letterspaced capitals and the descriptions are small, light gray, and easy to skip. The names won and the descriptions lost. Today I would set the descriptions a size larger and a shade darker and let the names calm down.

Step 4: Should a menu show dollar signs?

Not on a table menu. Yang, Kimes, and Sessarego's Cornell Hospitality Report from May 2009 tested three price formats on 201 lunch diners at the Culinary Institute of America's restaurant: "$20.00," "20," and "twenty dollars." Diners who saw plain numerals spent $5.55 more on average, about 8%, than those who saw dollar signs. The authors' explanation is the "pain of paying": the symbol reminds you that money is leaving.

So: no dollar signs, no trailing zeros, and no right-aligned price column with dot leaders, which turns the page into a price list and invites the guest to shop by the cheapest number. I set the price at the end of the description in the same type, where it reads as information rather than a bill. The 2016 FishTale menu got this right: small italic numerals, no symbols on the items, right after the name.

The exception is a quick-service board where the price is the offer. The 3 Peppers burrito board says the base price for every style is $9, in big type with the dollar sign, on purpose, because the message is that every style costs $9. The BWA board sets prices as $9.5 and $2.7, which I would not do again: a dollar sign and a half-dollar decimal is the worst of both formats. Pick numerals or pick currency, then be consistent.

One legal note. Under 21 CFR 101.11, chains with 20 or more locations under the same name and substantially the same menu must list calories next to each item's name or price, in type no smaller than the smaller of the two and in the same contrast, with the statement that 2,000 calories a day is used for general nutrition advice. The rule covers electronic menus and menus on the internet, so a covered chain's QR and digital menus are in. A single-location restaurant is not covered by the federal rule.

Step 5: Photography, color, and type

Photos belong on boards and quick-service menus, where a guest decides in seconds and a picture of the pizza sells the pizza; the Little Caesars board leads with a stack of them. On a full-service dinner menu I leave photos out, because they pull the perceived price point down and the kitchen never plates it like the photo. That is my judgment, not a study.

Type has one rule: if a guest in their sixties needs a phone flashlight, it is too small. Dining rooms are dim by design. The menu is also a brand piece, so the typeface, colors, and logo should match the sign outside and the website; if they do not, run the brand audit before the reprint. Color on paper behaves differently from color on a screen, which is a separate piece; the short version is that a chalkboard-texture board and an uncoated paper menu will not show the same red unless somebody planned for it.

Step 6: Choose the material for the reprint cadence

Menujo's 2026 guide assumes three or four reprints a year, and the material has to match that rhythm. The cost guides from Menujo (updated September 15, 2026), Tabres (June 2026), and ColorCopiesUSA:

Format	Reported 2026 cost	How long it lasts (vendor figures)
Paper single sheet or insert	$0.40–$2.00 each; about $30 per 150 two-sided letter sheets	Weeks to months
Laminated	$2–$5 each	6–12 months
Synthetic, waterproof	$1.49–$1.70 each at 50 pieces (8 to 11 pt); blank stock $75–$80 per 25 sheets	2+ years
Tri-fold	$3–$8 each	—
Premium bound	$8–$25 each	—

Menujo's direct-cost estimate for a 50-table restaurant runs $600–$2,000 a year across those reprints, before design fees and rush jobs. The design answer is to separate what changes from what does not. A bound or synthetic cover carries the brand, the story, and the sections that never move; a cheap paper insert carries the prices and specials. Then a price change costs $30 of paper instead of a laminating run, and nobody has to write over the old number with a marker.

Step 7: Digital boards or print?

Several of the portfolio menus are digital boards: the juice bar's chalkboard-styled panels, the burger counter, and the vape shop boards from Nashville. Screens win when prices change often, when the menu changes by daypart, and when the counter is the whole restaurant. They lose to print when the room is bright enough for glare, when the viewing distance is long, and when the guest is seated and wants to hold something.

Type on a screen across a counter follows the same math as a sign, roughly 30 feet of legibility per inch of capital height per the United States Sign Council, so a board read from 15 feet needs half-inch capitals at minimum and more in practice; the wrap piece walks through the numbers.

The NRA's 2026 report says a solid majority of adults are comfortable ordering and paying through phones, websites, kiosks, and tablets, and that 62% would order from a live person on a video screen versus 39% from an AI persona. Guests accept screens; they still want a human behind them. The "73% prefer QR menus" statistic on vendor sites has no citation I could find, so it is not here.

Mistakes I see

• Dollar signs and a dot-leader price column on a dinner menu.
• Keeping the dogs because one regular orders one.
• The restaurant's story on page one, in the sour spot, ahead of the food.
• Descriptions that are ingredient lists.
• Laminated menus from 14 months ago with prices written over in marker.
• QR-only menus with no printed fallback for the guest whose phone is dead.

What it costs / what to expect

Print costs are in the table above; the number that matters is the annual one. Menu design is quoted per project through the creative services page; a menu system, meaning the table menu, the boards, and the online version built from one set of rules, costs more than a single sheet and less than three separate designs.

Where to go next

Look through the menu portfolio to see the examples above at full size, including the parts I would change now. My bio covers where the menus came from.