Where an item sits on a menu changes how often it gets ordered, and the effect is large enough that researchers have measured it in cafés, labs and online panels. The measured gap is not the difference between a hit and a flop. It is closer to the difference between a dish that gets noticed and the same dish sitting in the middle of a list where eyes slide past it. I have not run a layout test on the Heritage Diner menu that I can report here, so everything below comes from published studies, each linked, along with what those studies could and could not show.

Psychology of Restaurant Menu Design: Writing Menus That Sell, Donald Burns on YouTube

What the research covers and what it does not

Three differently laid-out printed menu cards fanned out on a wooden table with a pencil.

Four pieces of published work do most of the lifting in this post. Two of them measured where an item sits in a list. One tracked where people look. One looked at fonts, boxes and the names of dishes. They were done in different places, on different people, with different foods, and none of them was run on a Long Island diner menu.

That matters because menu research has a reputation problem. Nearly every article on the topic quotes the same handful of studies, often without saying who ran them, how many people took part, or whether the people were ordering real food. Some of the studies used hypothetical choices from a printed page. Others used a real café. A few used online panels choosing between tacos and wine they would never eat. Each design answers a slightly different question, and a menu is only one of the places where a customer’s decision is made. Price, habit, hunger and the person across the table are all pulling at the same time.

So the reasonable way to read this research is as a list of things worth checking on your own menu, in your own room, with your own regulars. It is a list of hypotheses with some evidence behind each. None of it is a rule.

Position inside a list

In 2011 two researchers at the Hebrew University of Jerusalem, Eran Dayan and Maya Bar-Hillel, tested whether the position of an item within its category changes how often it is ordered. Their paper appeared in the journal Judgment and Decision Making, and the abstract reports that items at the beginning or end of a category were “up to twice as popular as when they were placed in the center of the list.”

The design is worth describing, because it is what makes the result believable. In the first study, 240 students were handed one of four versions of a pizzeria-style menu and chose an appetizer, an entrée, a soft drink and a dessert. The versions rearranged the same items, so the dishes themselves were held constant and only the order moved. In the second study a café in Tel Aviv alternated its standard menu with a rearranged one, 15 days each, and waiters recorded the orders. Across the two studies, items at the edges of a category drew about 55 percent of the choices against about 45 percent for the middle.

Then comes the part that gets left out of most summaries. The authors note that individual items mostly did not reach statistical significance on their own, and that in the café only the soft drinks category did at the category level. The first study was hypothetical. The waiters in the second were not blind to the change. The authors also wrote, of why menus behave differently from other position effects, “We cannot offer a satisfying explanation.” An honest reading is that edges seem to beat middles, by a margin that is smaller than “twice as popular” suggests in the typical case.

A second study complicates the picture in a useful way. In a 2018 paper in Cornell Hospitality Quarterly, Jungkeun Kim and four co-authors compared horizontal and vertical displays. When options were laid out in a row, people chose the middle more often. When they were stacked in a column, people leaned toward the first and last. In one experiment with 202 U.S. adults choosing among four tacos and four wines, non-edge tacos were chosen 60.4 percent of the time in the horizontal layout against 46.2 percent in the vertical one. In a second experiment with 155 adults and five cocktails, the figures were 70.9 percent against 55.3 percent. Both groups were online panel members, not diners, and the choices were hypothetical.

Put the two papers side by side and the lesson is that “put the item you want to sell first” is a half sentence. Which spot is the strong one depends on whether the eye is traveling across or down.

The center of the page

Order pad, pen and a stack of folded paper menus on a diner counter.

The reason the middle wins in a row has a name, the central gaze cascade, and it comes from eye tracking. A. Selin Atalay, H. Onur Bodur and Dina Rasolofoarison published the work in the Journal of Consumer Research in 2012. According to the release announcing it, participants increasingly looked at the central option in the five seconds before they decided, and that gaze pattern predicted what they chose. They did not remember doing it. The authors concluded: “Being unaware that our attention is focused on the center can lead to poor choices.”

The products in that study were vitamins, meal replacement bars and energy drinks, low-involvement purchases among brands the buyers did not know well. A person scanning a menu for a place they have eaten at for twenty years is not in that situation. Someone who orders the same omelet every Sunday is not looking at the center of anything. The effect is likeliest to show up for the customer who has never been in before, which for a restaurant on a busy road is a real share of the room, and for a neighborhood place on a Tuesday is a small one.

Boxes, bold type and the names of dishes

The best known work on emphasis is from Cornell’s Food and Brand Lab. Brian Wansink and Katie Love analyzed 217 menus and the choices of more than 300 diners, and Cornell’s news release on the study, published in the International Journal of Hospitality Management in 2014, reports two findings. Items that stood out through bold or colored type, highlighting or a box were ordered more often than the neighbor beside them. And items with more descriptive names sold better, with one cited renaming test lifting sales by about 28 percent while the recipe stayed the same. TIME’s coverage gives “seafood filet” becoming “Succulent Italian Seafood Filet” as the example.

A caution goes with this one, and it is the sort a careful reader should want stated. In 2018 Cornell’s faculty committee found that Wansink had committed academic misconduct in his research and scholarship, and he resigned; the JAMA journals retracted six of his studies the day before the announcement, and Retraction Watch counted 13 retracted papers overall. That coverage concerns other work. I have not checked whether the 2014 menu paper itself has been corrected or retracted, so I am treating it as a suggestion about what to try and not as settled fact. The other findings in this post stand on different authors and different data.

One line from the release is useful regardless of the paper’s standing. Wansink said of the visually prominent dishes, “In most cases, these are the least healthy items on the menu.” That is a statement about what restaurants tend to highlight, which is usually the high-margin plate. It is a reminder that emphasis is a choice the owner makes, and that customers who read menus closely will sometimes notice which dishes got the box.

What a small restaurant can test

Back-office desk with a notebook of tally marks, a calculator and a printed menu page.

None of the work above comes with a recipe, but it points to a short list of changes that cost nothing to try and are easy to undo.

  • Move one item. Take a dish that deserves more orders than it gets, shift it to the top or bottom of its section, and leave everything else alone for a few weeks. Moving five things at once tells you nothing about which move worked.
  • Keep the layout orientation in mind. A menu printed in columns is read down the column. A card with items in a row is read across. The Kim study suggests the strong positions are different in each case.
  • Rename, don’t reinvent. A plainer name and a more specific one can describe the same plate. The test is whether the longer name tells the customer something true, such as the cut, the preparation or where the ingredient came from.
  • Count something. A point-of-sale report or a tally on a notepad of how often the item sells before and after is enough. Do not read much into one slow week.
  • Write down what changed and the date. Menus drift, and a layout that worked in October may be sitting beside a price change from November.

Be wary of any single study that promises a lift of some percentage on your own menu. The café in the Dayan study changed 20 items across three categories and ended up with only one category showing a clear effect. Expect something modest.

Keeping every copy of the menu the same

A layout choice only works if the customer sees the layout you chose, and a restaurant’s menu now lives in a lot of places: the printed card, the website, the Google listing, the third-party pages nobody remembers creating. The Heritage Diner learned that lesson the plain way. The diner’s Google Business Profile once showed stale menu prices because the menus were updated by hand and the Google copy was forgotten, and people called to order from the old menu. Today one menu is synced to Google, the diner’s website and other places, so a change happens once.

The same logic applies to a test. If a dish moves to the top of the printed menu and stays buried on the website, half the customers see the old version and the numbers are muddy. Change every copy at once, or change one place and compare against the others on purpose. Till Post has a longer piece on keeping a printed menu in line with the online one, and another on how a menu is priced, which is the other half of what a customer is weighing when reading the page.

For informational purposes only. The research cited here comes from lab studies, online panels and a small number of restaurants, and results on a particular menu will differ.

Sources