That 'A' on the Window? It's a Snapshot, Not a Report Card

You've done it. We've all done it. Standing on a sidewalk, squinting at a laminated piece of paper taped next to a restaurant's front door, reading a number like it's a credit score for burritos.

92. Okay, that seems fine. 87? Hmm. Is 87 good? Is that a B? Is a B kitchen going to give you food poisoning, or is that just... Tuesday?

Here's the uncomfortable truth about health inspection scores: that number isn't a comprehensive audit of the kitchen. It's a sample. A small, one-shot, eyeball-based sample of a food operation that runs twelve hours a day, seven days a week. And the way most of us read it — as a stable, meaningful grade — is one of the most quietly misunderstood statistics in American daily life.

What an Inspector Actually Sees

A typical health inspection lasts somewhere between one and three hours. The inspector walks through, checks refrigeration temperatures, looks for cross-contamination issues, peeks at handwashing stations, and jots down whatever violates the local code. Then they leave.

That's it. That's the dataset.

Think about what that means in sampling terms. A restaurant might serve 300 meals on a busy Friday. Over a year, that's tens of thousands of food-handling events — every chop, every wash, every temp check. The inspector observes a fraction of one percent of that activity, once or twice a year, on a day the kitchen may or may not have known was coming (in many jurisdictions, inspections are unannounced, but staff can usually spot the clipboard).

If a pollster called 40 people and predicted a national election, you'd laugh. But a couple of hours of observation becomes a laminated grade we treat like gospel. N equals many in a kitchen — the inspector's n equals, well, one visit.

The Same Kitchen, Three Different Grades

Here's where it gets genuinely weird: the number itself doesn't mean anything consistent across state lines. Or sometimes, county lines.

There is no national health inspection grading standard in the United States. The FDA publishes a model food code, but enforcement is local, and scoring is all over the map:

• Some places use letter grades (Los Angeles County's famous A/B/C system, which it pioneered in 1998 and much of the country copied).
• Some places just use numbers with no letter attached, leaving you to guess whether an 80 is great or grim.
• Some places post nothing at all — the score exists in a database you've never opened.
• Some places weight violations differently, so a minor paperwork slip in one county costs the same points as a serious temperature problem in another.

A kitchen scoring 92 in one jurisdiction might score 74 two counties over, not because the kitchen changed, but because the rubric did. Violation categories, point deductions, and grade cutoffs are all locally determined. That 'A' isn't a universal constant. It's a local dialect.

The Myth of the Stable Grade

The deeper statistical problem is the one we fall for constantly: treating a single measurement as a fixed trait.

We do this with restaurant scores, but also with a one bad performance review, a single bad Yelp rating, or one bad night's sleep telling you your health is collapsing. One observation tells you something — an inspection that finds raw chicken above ready-to-eat salad is real information, and it matters. But it's information with enormous uncertainty attached, and the letter on the window strips all that uncertainty away.

Kitchens also change. Staff turnover in food service is famously brutal. The crew that earned a 98 in March might be largely gone by October. The score on the door could be describing a team that no longer exists.

So Should You Ignore the Grade?

No — but read it like a statistician instead of a judge.

A very low score, especially one with repeated critical violations, is genuinely useful signal. Studies after LA's grading system launched found measurable drops in foodborne illness hospitalizations, which suggests the scores do capture something real, even with tiny samples. A single observation can still be damning if what it observes is bad enough.

But the difference between a 93 and an 88? That's mostly noise. It might reflect the inspector's mood, what time of day they showed up, whether the dishwasher was on break, or whether the kitchen had just survived a slammed lunch rush. Two inspectors on two days could plausibly hand the same kitchen different grades.

The healthiest way to read that laminated paper: it's one draw from a distribution, not the distribution itself. A kitchen is a process, not a number. And processes — like all the messy, repeated, everyday systems we actually live inside — only reveal themselves over many, many observations.

Which is a roundabout way of saying: the score on the window is a hint, not a verdict. The real grade gets written one shift at a time, by a sample size the inspector will never see.

And the 3 a.m. line cook who washes his hands every single time? He's the data point that never makes the window.