The card is taped inside the glass at eye level, between a delivery sticker and a handwritten note about the bathroom code. Most people read the letter, decide it is fine, and walk in. A few pull out a phone and open the underlying report, which is where the useful material sits, because a single letter is a compression of a long document and compression throws things away.
I have read enough of those reports to have formed opinions. They are worth checking, and they are misread constantly, in both directions, by people who assume the letter measures either everything about a kitchen or nothing at all.
Fifty states and no single rulebook
There is no national restaurant inspection. The Food and Drug Administration publishes a model called the Food Code, now in its tenth edition dated 2022, and every line of it stays advisory until a state, county, or city writes it into local law. The agency's 2024 adoption report counted 30 state agencies across 24 states using either the 2022 or the 2017 edition, covering roughly 52 percent of the population. Oversight of restaurants and retail food stores runs through state and local agencies rather than the federal government.
That has a consequence people rarely account for. Two counties an hour apart can inspect against different editions of the same model, score them on unrelated scales, and publish the results in different formats. Some places assign letters. Some assign numbers where high is good. Some publish a pass or fail.
So a grade is a local unit of measurement. An 88 in one county and an A in another are not two readings of the same instrument, and comparing them the way people compare hotel stars produces nonsense. The first question about any score is which scale produced it.
The five behaviors an inspector is actually hunting
Under the Food Code, an inspection is organized around risk factors, meaning the practices most often reported to the Centers for Disease Control and Prevention as contributors to foodborne illness outbreaks. There are five of them, and they carry more weight than anything else on the form.
- Food arriving from an unsafe source, meaning a supplier outside the approved chain.
- Improper holding temperatures, the single most common finding in most jurisdictions, covering both hot holding and cold holding.
- Inadequate cooking, where a product never reached the internal temperature its pathogen profile requires.
- Contaminated equipment, which includes cross contamination between raw and ready-to-eat food as well as dirty food contact surfaces.
- Poor personal hygiene, most often handwashing failures and bare hand contact with food that will not be cooked again.
Layered on top are five interventions the code treats as the defenses against those risks: a person in charge who can demonstrate knowledge, employee health controls that keep sick staff away from food, barriers that stop hands from carrying contamination, time and temperature parameters, and the consumer advisory on undercooked items.
Everything else on the report is real but secondary. A cracked ceiling tile, an unlabeled squeeze bottle, a mop stored in the wrong place: these matter, and they signal how a kitchen is run, but none of them is how people get sick. A report with eleven minor findings and no risk factor violations describes a different kitchen than a report with one violation for chicken held at 52 degrees.
What a letter grade compresses
New York City has required restaurants to post a letter grade based on the sanitary inspection score since July 2010. Points are assigned per violation according to severity, and unlike a school grade, lower is better. The bands are fixed.
| Grade | Point range | What it hides |
|---|---|---|
| A | 0 to 13 | A spotless kitchen and one that missed the B by a single point look identical |
| B | 14 to 27 | Points may come from one serious finding or from many small ones |
| C | 28 or more | The specific failure that produced the score, which is often fixable in a day |
The band is the whole trick. A restaurant sitting at 13 points and a restaurant sitting at zero both hang the same blue A in the window, and the difference between them is invisible until you open the report. Grades near a boundary are the least informative numbers on the card.
The evidence that the card changed something
Public posting was not adopted on a hunch. Wong and colleagues, writing in the American Journal of Public Health in 2015, tracked New York inspection results across the transition and found the share of restaurants scoring in the A range rose from 28 percent in July 2008 to 46 percent five years later. Survey work in 2012 found 91 percent of residents approved of the program and 88 percent said the grades factored into where they ate.
A later study went further. Firestone and Hedberg, publishing in the CDC journal Emerging Infectious Diseases in 2018, compared Salmonella infection rates in the city against the rest of New York State. After grading began, city rates fell 5.3 percent per year faster than the comparison area between 2011 and 2015, a 32.6 percent decline against 14.1 percent elsewhere.
Read that as strong evidence that public posting changes kitchen behavior at scale. Read it carefully as evidence about any single restaurant, which it is not. The finding is about a population of thousands of kitchens over five years, and the card in one window carries none of that statistical power.
What the report cannot see
The date on the card is the most underrated field on it. An inspection is a few hours in a kitchen that operates thousands of hours a year, and the report describes those hours only. Between visits, a chef leaves, a walk-in compressor starts failing, or a new crew is trained by whoever happens to be free.
Several real hazards also fall outside most scoring systems entirely. Allergen handling is inconsistently covered. Delivery and third-party courier handoffs sit past the point where the inspector stops looking. Whether an employee felt sick and came in anyway is captured only if someone admits it. Staff turnover, which drives most of the drift between good inspections and bad ones, appears nowhere on the form.
None of that makes the card worthless. It makes it a floor rather than a verdict, and floors are still useful.
Looking up one place before you order this week
Search your county or city health department for a restaurant inspection database, since almost every jurisdiction now publishes one and most are searchable by name. Skip the summary line and open the full report for the most recent visit.
Then read three things. Check the date, and treat anything older than a year as a weak signal. Check whether any violation is flagged as critical, priority, or a risk factor, since that flag separates the findings that make people sick from the housekeeping. Check the visit history, because one bad inspection followed by a clean reinspection tells a story of a problem fixed, while three years of the same repeated violation tells a story of a manager who has decided not to.
If the report worries you, the useful move is not a review site post. Call the health department's environmental health line and ask whether the violation you read was corrected on site or at a follow-up visit. That question is answerable, they answer it, and the answer is usually the thing the letter left out.