An Oxford comma once cost a Maine dairy $5 million. Drivers for Oakhurst Dairy sued over unpaid overtime, and the fight came down to a missing comma in a state law listing which work was exempt from overtime pay. The First Circuit found the punctuation ambiguous, and in 2018 the company settled the case for $5 million. Your menu is not a labor statute. But the lesson travels: the words on the document you put in a customer’s hands are not decoration. They get read literally.
Here is the uncomfortable part for anyone running a kitchen. The National Restaurant Association’s 2025 Restaurant Operations Data Abstract put the median full-service restaurant’s pre-tax income at 2.8 percent of sales in 2024. Full-service places doing under $2 million in sales kept 1.1 percent. On that margin, a menu that quietly turns diners off is not a proofreading issue. It is a revenue leak with nowhere to hide.
What actually changes when a diner spots a mistake
Researchers have tested the menu-typo question directly, and the answer is messier than the grammar police would have you believe. Ricardo Fernandes’ 2021 study at the University at Albany showed 566 people a Chinese restaurant menu. Some saw a clean version. Others saw one seeded with 36 grammatical errors. The bare presence of errors changed nothing: the tidy menu and the error-strewn menu drew the same quality ratings and the same intent to order.
Volume moved the result. Participants who reported spotting nine or more errors rated the restaurant significantly lower on quality and were significantly less likely to order. Those who caught about six behaved almost exactly like those who caught none. Damage from bad copy is not a switch you flip. It is a running total, and the total is what eventually trips a diner’s patience.
Surveys point the same direction, though stated intent is a softer signal than a receipt. In a 2017 SurveyMonkey Audience study of 1,042 U.S. adults, 81 percent of women and 77 percent of men said they would be less likely to buy a product advertised with spelling or grammar mistakes, a number that climbed to 85 percent among millennials. A 2022 UK survey from Accuracy Matters and Opinium found 68 percent of adults less likely to purchase from a brand with errors in its messaging, and 73 percent saying their opinion of it would drop. Neither study watched anyone walk out of an actual restaurant, which is the honest limit of survey research. The pattern still holds, and it bites hardest in the age group most restaurants are chasing.
The evidence at a glance
| Study | What it tested | What it found | Year |
|---|---|---|---|
| Fernandes, University at Albany | 566 diners shown a menu seeded with 36 grammatical errors | Noticing nine or more errors cut quality ratings and purchase intent; noticing a few did not | 2021 |
| Poundstone, national survey | A fictitious sandwich menu, clean versus deliberately misspelled | No measurable difference in appeal, willingness to try, or price people would pay | 2016 |
| Accuracy Matters / Opinium, UK | Adults shown brand messaging containing errors | 68 percent less likely to buy; 73 percent said their view of the brand would worsen | 2022 |
| SurveyMonkey Audience, US | 1,042 adults shown ads containing errors | 81 percent of women and 77 percent of men less likely to buy; 85 percent of millennials | 2017 |
| Al-Batineh, delivery menus | Talabat app menus and a user survey in the UAE | 58.7 percent would pay more for a clearly translated dish; 64.2 percent said poor translation deters them from new places | 2024 |
Figures as published in the sources above, checked in September 2026. Survey results measure what people say they would do, not what they did.
The case that diners don’t care, and where it actually holds up
There is a real counterargument, and it deserves a hearing. William Poundstone’s 2016 experiment for his book Head in the Cloud put a fictitious sandwich shop menu in front of a national sample, one version clean and one stuffed with errors like “Sandwhitchs” and “vegitarian.” He asked how appealing the food looked, whether people would try the place, and what they would pay. The misspelled version scored the same on every measure, inside the margin of error. No penalty at all.
My read is that the finding is genuine and narrower than it sounds. His sandwich shop did not exist. Participants had no relationship to it, no loyalty, and no real money on the line, and the errors were the entire point of the test rather than something they stumbled onto while hungry. A stranger grading a hypothetical has little reason to care about a typo. A regular deciding where to spend $60 on a Friday has more.
What would settle the question is a field experiment inside working restaurants, measuring repeat visits rather than survey answers. I could not find one, and neither the surveys above nor Poundstone’s study replaces it. So treat every number in this article, including the ones that back the point I am arguing, as weaker than a POS report.
A menu sells and testifies at the same time
Operators already spend money engineering the sales side of a menu. Cornell’s Brian Wansink showed in 2001 that sensory descriptions, the gap between a plain dish name and copy that names texture and cooking method, lifted both sales and diners’ ratings of the food even when the dish was identical. Descriptions are sales copy. A description a reader has to decode is sales copy that has stopped selling.
The testifying side is where it gets expensive. In Williams v. Stellar Restaurant Group, a Massachusetts jury found a burrito chain negligent in 2024 after a child whose order was flagged for a peanut allergy was served a burrito with peanut sauce; the judgment, interest, and legal fees came to about $220,000. That same year, a grocery chain faced a wrongful death suit after a mislabeled cookie containing peanuts killed a 25-year-old dancer. In a stranger turn, a fired Disney employee was federally charged with hacking the company’s menu software and falsely marking items as safe for people with peanut allergies.
Allergen markers, dietary icons, and price fields are the parts of a menu customers act on. A wrong icon is not a stylistic slip. It is a claim, and claims carry liability.
The mistakes a grammar checker will sail right past
Running the menu through a grammar checker before the file goes to print is worth the ten minutes, and it will catch a fair share of the obvious slips. It will also wave through the errors that cost restaurants the most, because most of those are not grammatical. Here is what to hunt by hand.
- Translation errors. The 2024 Talabat study found that 58.7 percent of users would pay more for a dish with a clear, well-translated description, and 64.2 percent said poor translation keeps them from trying new restaurants. A separate analysis of menus in Geneva’s Old Town found mistranslation to be the single most common error type across every menu studied. Machine translation usually gets the dish list mostly right and the allergen line dangerously wrong.
- Prices that drifted. Compare every printed price against the POS export after each update. A decimal in the wrong place turns a $12 sushi roll into a $120 one, and the customer spots it before you do.
- Inconsistent units and formatting. “8 oz” on one line and “8oz” on the next, a gluten-free marker on a dish that shares a fryer, a section header in title case above items in sentence case. Consistency reads as competence.
- Ghost items. Dishes you dropped months ago that are still on the dining room copy, the website, or the delivery app. The app version drifts fastest and gets read most.
A proofing routine that catches what actually costs you
You do not need an English degree. You need a process that treats the menu like the operational document it is. Assign one person, give them an hour, and run this list.
- Read it out loud, slowly. Your ear catches what your eye skips.
- Hand it to someone who did not write it and does not cook. They read as a customer, which is the only reading that counts.
- Reconcile every price against the POS on the same day you update either one.
- Check every allergen and dietary marker against the current recipe, not last season’s menu.
- Print it and hold it. Formatting problems hide on backlit screens.
- Update the website and the delivery apps in the same pass, then check them again a week later. They are the copies most likely to fall out of sync.
None of this requires talent. It requires deciding a menu is worth the same attention as a portion scale, then doing the boring part on schedule. The restaurants that get this right are rarely the best writers. They are the ones with the shortest gap between a dish changing and its description changing with it.
Menu grammar questions, answered
Do typos on a menu actually reduce sales?
Not on their own, based on the best evidence available. What reduces sales is diners noticing a pile of them. In Fernandes’ study, and in a separate 2019 experiment on commercial websites, the effect showed up mainly among readers who detected the errors, and it grew with the number they caught.
How many errors are too many?
In Fernandes’ study, the penalty appeared at nine or more noticed errors and not at roughly six. Treat that as a rough threshold, not a rule. The safer reading is simpler: you have no margin for error in the fields customers act on, and a little tolerance for minor slips in prose.
Will a spell checker catch menu mistakes?
Some. Spell check will not flag a correctly spelled wrong word, a mistranslated dish, a price that no longer matches the POS, or an allergen icon on the wrong item. Those need a human who knows the current recipes.
Should I list allergens on my menu?
If you print them, you own the accuracy of every marker. Courts have found restaurants liable when a stated allergy accommodation failed, and labeling rules vary by state and country. Some places require allergy awareness materials at food establishments even where menu labeling is not mandatory. Check your local rules before you design the icons.
How often should a menu be checked?
Every time a price changes, a dish is added or dropped, or a supplier swaps an ingredient. At minimum, once a quarter. Menus that only get read when they are reprinted are the ones that stay wrong the longest.
How this article was put together
This piece set out to answer one question: does bad spelling and grammar on a restaurant menu cost sales, and if so, how. I read the original studies rather than summaries of them, including Fernandes’ 2021 menu experiment, Poundstone’s 2016 sandwich shop survey, and the 2024 Talabat delivery-app research, and checked the National Restaurant Association’s 2025 operations data for margin context. Where the evidence is thin, mainly the absence of a real-restaurant field experiment, the article says so. Survey figures describe stated intent, not observed spending. Margin and menu-price numbers move yearly, so recheck them before you quote them. No one paid for the inclusion of any product or service mentioned here.