NYC Clean Eats

The thirteen point cliff

One point separates an A with no fine from no grade, a return visit and a bill. In the city's records, every grade boundary leaves a mark the size of what it costs to cross.

What one point is worth

Violations carry points, and a lower score is better. The letter bands themselves are clear enough: 0 to 13 is an A, 14 to 27 is a B, and 28 or more is a C.

However, on the first inspection of the year, the boundary between 13 and 14 points is not really about the letter grade. It is about three major financial and operational consequences at once. The Health Department's official guide for business owners states it plainly: score under 14 points and "any and all penalties from the inspection are waived". Score 14 or more points and "you will be re-inspected".

OutcomeScore 13Score 14
Placard in windowA, immediatelyNo new grade issued
FinesWaivedStand
Next stepNothingA re-inspection

A single point is the entire difference between a clean result at zero cost and an open city case with a bill attached. That is substantial weight to place on one number, and boundaries carrying that much consequence leave unmistakable patterns in the data.

The mark it leaves

Plotting every initial inspection score from published city records reveals a massive cliff right at the threshold.

0 2k 4k 6k A ends at 13 0-151013142027283540 inspection score (lower is better)
A0 to 13 B14 to 27 C28 or more n = 55,795

A score of 13 is 22 times more common than a score of 14. Scores of 12 and 13 together make up 21% of every initial inspection across the city, while scores of 14 and 15 together make up just 1%.

Testing the stakes: Initial visits versus re-inspections

We can test if the scoring threshold itself causes this cliff by comparing lines where consequences are high against lines where consequences are low. Look at the boundary between 27 and 28 points, which separates a B from a C grade.

On an initial inspection, that boundary carries almost no penalty. At 14 points or more, no grade is posted either way, so scoring a 28 instead of a 27 changes nothing on the day. The data reflects that lack of consequences:

ScoreInitial inspectionsChange
26888−8%
27937+6%
28840−10%
29752−10%

The 10% drop at 28 points sits alongside normal baseline fluctuations. Nothing dramatic happens because nothing is at stake.

On a re-inspection, however, that same line determines the actual letter placard posted in the window. When real consequences attach to the score, the pattern shifts immediately:

ScoreRe-inspectionsChange
26438+16%
27643+47%
28126−80%
2988−30%

Scores build steadily up to 27 points, which is the absolute limit for a B grade, before dropping by 80% at 28. That shape, climbing to the edge and plunging past it, is what happens when a boundary carries real stakes.

0 2k A ends at 13 0-151013142027283540 inspection score (lower is better)
A0 to 13 B14 to 27 C28 or more n = 19,705

Both cliffs appear clearly on re-inspections. The drop at 13 points is a wall, while the drop at 27 points is a steep step. Neither cliff exists on an initial inspection except at 13 points, which is the only boundary that carries an immediate financial penalty during a first visit.

How penalties shape score distributions

Comparing all four boundaries shows how closely score patterns mirror real-world consequences:

BoundaryWhat crossing it costsDrop
13 to 14, initial The A grade, waived fines, and a required return visit −96%
13 to 14, re-inspection An A becomes a B −95%
27 to 28, re-inspection A B becomes a C −80%
27 to 28, initial Nothing, no grade is posted either way −10%

Avoiding a C grade matters, and the 80% drop reflects that. However, avoiding a B grade or a re-inspection penalty matters far more, and those drops are even larger. The single boundary where nothing is at stake shows no cliff at all.

Administrative hearings and score adjustments

An inspection score recorded on site is not always the final number. Restaurant owners have the legal right to challenge citations at administrative hearings held through OATH, the city's Office of Administrative Trials and Hearings.

This appeal process impacts how the city records public data. The official open dataset includes "every sustained or not yet adjudicated violation citation". When an administrative judge dismisses or reduces citations during a hearing, those points are deleted from the database entirely. An inspection that originally scored 16 points on site but lost 3 points in tribunal ends up listed as a 13 in the permanent public record, completely indistinguishable from a kitchen that scored a 13 on inspection day.

Inspector discretion at the margin, formal tribunal appeals, and operators fixing cheap violations first all contribute to this distribution. Together, they cause scores to stack up directly on the pass side of every major boundary.

Why this matters when reading a letter grade

A letter grade placard is a summary of a range, not a precise health verdict. A restaurant with 2 violation points and a restaurant with 13 violation points both display an identical A card in the window, but they represent very different kitchens.

Those scores are not spread evenly inside the A band. Of every A earned on an initial inspection, 36% scored 12 or 13, and one in six scored exactly 13, a single point from losing it. An A placard tells you a business passed. The score behind the letter tells you whether it passed cleanly or scraped by.

Sources

Look up a restaurant · The year-long A · More insights