A guest decides what to order in under two minutes, usually without reading more than half the page. Menu engineering is how you make sure the dishes they land on are the ones worth the most to your restaurant.
Two restaurants serve the same roast chicken. Same recipe, same plating, same $24 price. One sells 40 orders a week. The other sells 110. The difference almost never comes down to the chicken.
It comes down to the menu: where that dish sits on the page, what it's called, how it's described, whether a price has a dollar sign in front of it, and whether a guest's eye even lands on it before they order something else out of habit. Menu engineering is the discipline of making those decisions on purpose instead of by accident, and it is one of the few changes a restaurant can make that raises average check without touching a recipe, a price, or a staffing schedule.
None of what follows requires a menu consultant or proprietary software. It requires your own sales data, an afternoon of math, and a willingness to redesign a document most restaurants haven't genuinely rethought since opening week.
Run the numbers before you touch layout
List every item currently on the menu with three columns: selling price, plate cost (what the ingredients actually cost you, including garnish and sauce, not just the protein), and units sold over a representative period, four to eight weeks is usually enough to smooth out one slow Tuesday or one Instagram-driven rush. Subtract plate cost from price to get contribution margin per dish. That's the number that matters, not the price on the menu and not the food cost percentage by itself.
A dish with a 68% food cost can still be a star if it sells constantly and the dollar margin is healthy. A dish with a pristine 22% food cost can be a dog if nobody orders it. This is the single most common mistake restaurants make when they try to fix a menu without data: they chase food cost percentage instead of dollars earned, and end up pushing guests toward cheap-to-make dishes that don't actually move the average check.
Once every dish has a margin number and a volume number, plot them against your own averages, not an industry benchmark. A dish above your average margin and above your average volume is a star. Below both, a dog. The two mixed categories are puzzles and plowhorses. This exercise takes an afternoon with a spreadsheet and your POS sales export. It's the only step in this entire process that genuinely requires data instead of judgment, and it's the step most restaurants skip, which is exactly why most restaurant menus are laid out by instinct instead of evidence.
Place items where eyes actually land
Eye-tracking studies on printed menus consistently find the same pattern: attention lands first near the center of a page or panel, then sweeps to the upper right, then the upper left, and only reaches the bottom of a long list after a guest has mostly already decided. Few menus are designed with this in mind. Most are organized the way a kitchen thinks about food, appetizers at the top, entrées in the middle, dessert at the end, with every item given roughly equal visual weight regardless of what it earns.
Once you know which dishes are stars, give them the real estate the eye-tracking data says actually gets seen. That means the center of the first panel a guest opens, or the top of the entrée section, not wherever the dish happened to land alphabetically or by course. Within any list longer than four or five items, also protect the first and last position: readers register those more than anything in the middle, the same primacy-and-recency effect that shows up in surveys, playlists, and ballots.
- Limit each section to five to seven items. More than that and guests stop reading carefully and start skimming for something familiar, which usually means a plowhorse wins by default.
- Put one star in the visual center of each page or panel, and a second strong item at the top or bottom of each section list.
- Signal a featured dish with a subtle box, a short label like "house favorite," or slightly larger type, not an icon-heavy callout that reads as advertising rather than guidance.
- Keep total page count low enough that a guest can hold and scan the whole menu under normal restaurant lighting without hunting for a second panel.
Write descriptions that sell without sounding like copy
Description length has a real sweet spot. Very short labels ("Grilled Salmon") give a guest nothing to react to. Long paragraphs slow decision-making and get skipped entirely under time pressure. Somewhere between 12 and 25 words, with one or two sensory or sourcing details that match what the kitchen actually plates, consistently outperforms both extremes.
Compare two real approaches to the same dish:
- Weak: "Chicken Sandwich. Served with fries. $16."
- Working: "Buttermilk-brined chicken, pickled jalapeño slaw, house mayo, toasted brioche. Served with fries. $16."
Price presentation matters almost as much as the words
The second description above isn't longer by much, but it gives a guest something concrete to picture, and concrete language is what moves a dish from "maybe" to "yes." Write every description in the same sentence rhythm across the menu so it reads as one considered document instead of a patchwork of whoever wrote each section. If a dish carries multiple allergen notes or modifier options, move those to a smaller line underneath rather than stacking them into the main description, where they compete with the language actually meant to sell the dish.
Price placement matters too. Put the price directly after the description, or in a clean, unobtrusive column, and drop the dollar sign. Multiple studies on menu pricing have found that removing the currency symbol measurably reduces how much guests associate ordering with spending money, which shows up as a small but real increase in average check. A price formatted as "24" reads differently than "$24.00," even though it's the exact same number.
Use photography selectively, and make it earn its place
A menu with a photo next to every item reads as a chain, not a kitchen with a point of view, and it dilutes the one tool photography is actually good for: directing attention to dishes you want guests to choose. Photograph only your stars and a small number of dishes that genuinely need a visual explanation, an unfamiliar preparation, an unusual presentation, something a guest can't picture from the name alone.
Shoot every photo under the same lighting setup, from a consistent angle, on a neutral surface that doesn't compete with the food, and show the actual portion a guest receives, not a stylist's exaggerated plating. Consistency across a handful of photos reads as intentional; inconsistent photography, even of great individual shots, reads as improvised. Export the same files at the sizes you need for print, web, and any delivery or online ordering platform, so the dish looks identical everywhere a guest might see it before they decide to come in.
Keep print and digital versions in sync
A surprising number of restaurants redesign their printed menu and leave the website menu untouched for months, which means the engineering work only reaches guests who physically walk in. Publish the website menu as real HTML text, not a scanned PDF. Search engines can't read dish names inside a PDF image, mobile guests have to pinch and zoom to read one, and you lose the ability to update a single item without re-exporting the whole document.
When a dish changes, update the description, price, and photo everywhere at once, the printed menu, the website, any delivery platform, and anywhere servers reference it verbally. A guest who sees one version online and a different one at the table reads that gap as sloppiness, even when the actual dish is fine.
Print, train, and measure
Specify paper weight and finish that survives repeated handling and matches your other printed materials, and give the printer a locked template file with type styles, spacing, and logo placement already set, not a loose document to be assembled at the print shop. Brief servers on the two or three dishes you most want to move so they can speak about them in language that matches the menu; a server who can describe a star dish in one confident sentence sells more of it than any layout choice can on its own.
Track average check and the mix of star versus other orders for four to six weeks after a new menu launches, and treat that window as the real test of whether the redesign worked. If a specific placement or description underperforms, adjust it on the next printing rather than waiting for a full annual redesign. Once the template and photography standards exist, each future seasonal update takes a fraction of the time the first one did.
A realistic rollout plan
Trying to overhaul pricing, layout, photography, and copy all in one pass is how these projects stall. A phased plan keeps it moving and front-loads the work that actually affects average check first.
- Week 1: Pull sales and cost data, run the margin-and-volume classification, and identify your actual stars, puzzles, plowhorses, and dogs, not the ones you assumed before looking at the numbers.
- Weeks 2-3: Rebuild the layout around page geography, rewrite descriptions to a consistent length and rhythm, and reformat pricing to remove currency symbols and break up strict price-ordered columns.
- Week 4: Shoot photography for stars and any dish that needs visual explanation, then publish the new menu in print and as real HTML text on the website at the same time.
- Weeks 5-10: Track average check and the star-versus-other order mix, brief staff on what changed and why, and adjust any underperforming placement before the next printing.
Frequently asked questions
What is menu engineering?
Menu engineering is the practice of classifying every dish on a menu by contribution margin (price minus food cost) and sales volume, then using that classification to decide layout, description length, pricing presentation, and photography. The goal is to make high-margin, well-loved dishes easier to notice and order, without changing the recipes or raising prices.
What are stars, puzzles, plowhorses, and dogs on a menu?
They're the four categories menu engineering sorts dishes into: stars are high margin and high volume (your best dishes by every measure), puzzles are high margin but lower volume (profitable dishes guests aren't finding), plowhorses are high volume but lower margin (popular but not very profitable), and dogs are low in both (candidates to cut or rework).
Does removing the dollar sign from menu prices actually work?
Multiple studies on menu pricing have found that formatting a price without a currency symbol (writing "24" instead of "$24.00") reduces how directly guests associate ordering with spending money, which shows up as a small but measurable increase in average check. It's a low-cost change worth making regardless of what else you redesign.
How long should a menu item description be?
Roughly 12 to 25 words tends to outperform both very short labels and long paragraphs. That length gives a guest one or two concrete, sensory details to react to without slowing down decision-making, which matters since most guests spend well under two minutes actually reading a menu.
Should every dish on a menu have a photo?
No. A photo next to every item dilutes the technique and reads as a chain menu rather than a kitchen with a point of view. Photograph only your star dishes and anything genuinely hard to picture from the name alone, shot consistently under the same lighting and angle.
How often should a restaurant redesign its menu?
Most independent restaurants benefit from a full engineering pass once a year, using a fresh four-to-eight-week sales window, plus smaller adjustments each time a seasonal menu changes. Once a template and photography standard exist, seasonal updates take far less time than the first full redesign.
How soon will a redesigned menu affect average check?
Most restaurants can see a measurable shift in average check and in the mix of star-versus-other orders within four to six weeks of launching a redesigned menu, which is also the window worth tracking closely before judging whether a specific change worked.
Does menu engineering work for a menu that changes seasonally?
Yes, and it gets easier each time. The first redesign takes the most work because it requires a full data pull and a new layout template. Once that template, description format, and photography standard exist, seasonal updates only require rewriting and reshooting the items that actually changed.
