Imagine you run a restaurant review platform. A user searches for sushi in Manhattan and finds a place with a score of 9. They search the same evening for ramen in Los Angeles and see a different restaurant with a score of 94. Which one is safer to eat at?
The question is unanswerable without context - because those two numbers come from completely different measurement systems that are not even attempting to measure the same thing on the same scale. New York City's score of 9 is actually excellent (lower is better, and anything under 14 earns an A grade). Los Angeles's 94 is a percentage, and it might look great but it barely clears the threshold that nearly every restaurant in the county exceeds. They cannot be compared.
This problem - the fundamental incomparability of health inspection data across US jurisdictions - is the central challenge facing any application, platform, or enterprise that wants to use inspection data at national or multi-market scale. This article maps out exactly how the six largest and most distinctive inspection systems work, where their data lives, and why cross-system comparison requires a purpose-built normalization methodology rather than a raw data feed.
New York City: The DOHMH Point Deduction System
How It Works
New York City's Department of Health and Mental Hygiene (DOHMH) operates one of the most transparent and closely studied inspection programs in the country. The system is a point deduction format - inspectors start at zero and add points for each violation found. The counterintuitive result is that a lower score is better.
- Score 0-13: A grade (posted in window immediately)
- Score 14-27: B grade (triggers reinspection cycle)
- Score 28+: C grade (also triggers reinspection)
The specific point values for individual violations range from 2 to 28 points, calibrated to reflect the health risk associated with each type of failure. A critical violation that could directly cause foodborne illness carries far more points than a paperwork or labeling deficiency.
One distinctive feature of the NYC system is the "Grade Pending" card. When a restaurant scores in the B or C range on an initial inspection, they are not required to immediately post that grade. Instead, they receive a "Grade Pending" placard and the right to a reinspection. The grade from the reinspection (whichever is better - the initial or the reinspection score) becomes the official posted grade. This means the A card in a NYC restaurant window could represent either an initial A or a reinspection A after an initial B or C.
Violations are also classified as critical or general. Critical violations - those most likely to directly contribute to foodborne illness - carry higher point values and trigger different regulatory responses than general violations. A single critical violation of sufficient severity can push a previously A-grade establishment into B or C territory on its own.
Public data: data.cityofnewyork.us - updated continuously, full history available
Los Angeles County: The Percentage Placard System
How It Works
Los Angeles County's Environmental Health division uses a percentage-based scoring system that ranges from 0 to 100, with the score posted on a color-coded placard visible from the street. Higher is better. An A card (green) means 90-100%, a B card (yellow) means 80-89%, and a C card (orange) means 70-79%. Scores below 70 result in closure until corrections are made.
- 90-100%: A grade (green placard)
- 80-89%: B grade (yellow placard)
- 70-79%: C grade (orange placard)
- Below 70%: Closed / Conditional
The percentage is calculated by subtracting point deductions from 100. Critical violations carry heavier deductions than minor violations. However, the LA system has an important statistical property that substantially limits its utility for differentiation: the score distribution is heavily compressed toward the top end. The vast majority of inspected establishments - typically around 85-90% - score in the A range (90% or above). This means that an LA score of 91 and an LA score of 99 are both presented to the public as equivalent "A" grades, even though the underlying violation profile may be quite different.
This compression effect means LA scores are poor discriminators between establishments that are genuinely safe and excellent versus establishments that are technically compliant but have real issues. A place with two critical violations that were both corrected on-site may still score 91% - statistically indistinguishable from a spotless operation.
Critical violations in the LA system carry 4-point deductions, while minor violations carry 1-2 point deductions. Violations corrected immediately during the inspection receive partial credit, which can soften the score impact even for serious findings.
Public data: ehservices.publichealth.lacounty.gov - restaurant scores publicly searchable, bulk download available
Chicago: Pass/Fail with Violation Counts
How It Works
Chicago's Department of Public Health uses a fundamentally different model than NYC or LA. Rather than producing a numeric score, Chicago inspections result in a pass/fail determination, with violations categorized as critical or serious versus minor.
- Pass: No critical violations, minor violations within acceptable limits
- Pass with Conditions: Issues noted that require follow-up but establishment may remain open
- Fail: Critical violations present, reinspection required within 30 days
The "Pass with Conditions" category is an important middle ground. An establishment can pass with conditions when violations are noted but corrected during the inspection, or when the overall safety picture does not warrant closure but a follow-up is warranted. This category does not appear on a public placard - Chicago does not have a letter grade window posting requirement - so consumers interacting only with the posted result at the door have minimal information.
The Chicago open data portal publishes detailed violation records, meaning that while the top-level result is pass/fail, the underlying data contains each specific violation found, its risk classification, and whether it was corrected during the inspection. This is actually some of the richest inspection data available in the US - but consuming it requires parsing structured violation records rather than reading a single score field.
Establishments that fail are scheduled for reinspection within 30 days. Repeat failures can result in license suspension proceedings. Chicago also maintains a publicly searchable database of establishments with multiple recent failures, creating a de facto risk tier for persistent violators.
Public data: data.cityofchicago.org - detailed violation-level records, full history, excellent API
San Francisco: Numerical Score with Letter Grade Overlay
How It Works
San Francisco's Department of Public Health uses a 0-100 numerical score with letter grade thresholds. On the surface this looks similar to LA - both California cities, both using percentages, both posting grades publicly. In practice, the systems differ in important ways and scores are not directly comparable between the two cities.
- 90-100: Grade A
- 85-89: Grade B
- 70-84: Grade C
- Below 70: Closed or Conditional Permit
The grade thresholds in SF are lower than those in LA - an 87% in SF is a B, while that same 87% in LA would be a B as well, but the underlying inspection criteria, the specific violations that drive deductions, and the frequency of inspections differ between the two jurisdictions. The SF system also has a tighter score distribution in the B range than LA, making SF scores somewhat more informative for differentiating mid-tier establishments.
One important distinction in SF is the differentiation between routine inspections and complaint-driven inspections. Both appear in the public data, but they are not equivalent in terms of what they reveal. A routine inspection at a scheduled interval reflects the baseline operating condition. A complaint-driven inspection is almost always triggered by a specific reported incident, and violations found during complaint-driven inspections tend to be more severe on average simply because the inspection was prompted by a problem.
SF's public data includes inspection type, so consumers and analysts can distinguish between routine scores and complaint-inspection scores. This is a detail that gets lost when data is normalized without accounting for inspection trigger type.
Public data: data.sfgov.org - restaurant scores and inspection history with violation details
King County / Seattle: The Color Card System
How It Works
King County's Environmental Health division, which covers Seattle and surrounding areas, uses a color card posting system that is distinct from the letter grades used in NYC, LA, and SF. Inspected establishments post a placard color indicating their compliance status:
- Blue card: Excellent - no violations or all violations corrected during inspection
- Green card: Satisfactory - minor violations noted, no critical findings
- Yellow card: Unsatisfactory - violations found that require follow-up
- Red card: Imminent health hazard - immediate closure until hazards corrected
King County's system classifies violations into three risk categories: Priority violations (formerly called critical), Priority Foundation violations (systemic practices that could lead to Priority violations), and Core violations (general sanitation and maintenance). The three-tier risk classification provides more granularity than the standard two-tier critical/minor approach used in most jurisdictions.
The distinction between Priority Foundation violations and Core violations is particularly useful for identifying establishments that have underlying systemic issues - inadequate food handler training, gaps in written food safety plans, missing temperature logs - versus establishments that simply have a recurring physical maintenance problem like a damaged floor surface or inadequate lighting in a storage area.
King County's public data is available in multiple formats and the agency has historically been a leader in health department data transparency. The dataset includes not just violation types but specific description text, inspector comments, and the correction action noted during the visit.
Public data: kingcounty.gov - restaurant inspection reports, searchable database, annual data exports
Houston / Harris County: TFER-Based Scoring
How It Works
Houston and unincorporated Harris County fall under the jurisdiction of the Houston Health Department and Harris County Public Health, respectively. Texas food establishment inspections are governed by the Texas Food Establishment Rules (TFER), which aligns with federal Food Code guidance but is implemented at the local authority level with some jurisdiction-specific variations.
- Score 90-100: Excellent
- Score 80-89: Satisfactory
- Score 70-79: Needs Improvement
- Score below 70: Unsatisfactory - reinspection required
The Houston/Harris system starts at 100 and subtracts demerit points based on violation severity. Demerits range from 1 to 25 points per violation. The heaviest demerits - 20 to 25 points - are reserved for imminent health hazard violations that could directly cause foodborne illness: contaminated food supply, sewage backup in food prep areas, or lack of potable water.
Texas's TFER also requires that certain violations be marked as "In Compliance" or "Out of Compliance" rather than simply listing the finding. This binary coding of each item, combined with the numeric score, provides both a summary score and an itemized compliance record. However, not all Texas jurisdictions are equally consistent in their data publication practices - some maintain well-structured public datasets while others require public records requests for detailed reports.
Public data: houstontx.gov and Harris County Environmental Health - availability varies by entity
The Comparison Table
Here is a side-by-side summary of the six systems, structured to make the key differences visible at a glance:
| City / Jurisdiction | Score Format | Direction | Grade A Threshold | Public Posting | Open Data |
|---|---|---|---|---|---|
| New York City (DOHMH) | 0-28+ points (deductions) | Lower = better | 0-13 points | Letter grade card | Excellent |
| Los Angeles County | 0-100% (percentage) | Higher = better | 90-100% | Color-coded placard | Good |
| Chicago (CDPH) | Pass / Fail | Binary | Pass | Not required | Excellent |
| San Francisco (SFDPH) | 0-100 numeric | Higher = better | 90-100 | Letter grade card | Good |
| King County / Seattle | Color card (Blue/Green/Yellow/Red) | Categorical | Blue card | Color card required | Good |
| Houston / Harris County | 0-100 numeric (demerits) | Higher = better | 90-100 | Varies by inspector | Fair |
Why You Cannot Compare Raw Scores Across Cities
The table above makes the structural differences obvious, but the incomparability problem runs deeper than just different score formats. Even within systems that both use 0-100 scales, the underlying statistical distributions differ substantially.
Distribution Differences Change What Any Score Means
In Los Angeles, approximately 89% of all inspected food establishments score 90% or above - earning an A. In San Francisco, the figure is closer to 78%. This means that an LA score of 91 places you solidly in the bottom quartile of A-range establishments, while a SF score of 91 places you near the median. The same number conveys different information depending on where in the distribution it sits relative to that city's population of inspected establishments.
The distribution problem is even more severe for pass/fail systems like Chicago. "Pass" in Chicago covers an enormous range of underlying compliance quality. A perfectly clean kitchen and a kitchen with four minor violations that were corrected during the inspection both receive the same output: Pass. Any analysis that treats Chicago "Pass" as equivalent to an LA score of 94 is comparing incommensurable categories.
Inspection Frequency Changes What Scores Represent Over Time
New York City inspects each establishment significantly more frequently than most jurisdictions - typically 2-4 times per year with additional complaint-driven visits. Los Angeles also inspects high-risk establishments multiple times annually. But in many mid-size US counties, a single annual inspection is the norm for most restaurants.
This frequency difference changes the information content of any given score. A NYC score from a visit last month is fresh data. A county health score from a jurisdiction that inspects annually might be 10 months old. Treating these as equivalent signals in a restaurant safety ranking is misleading - the NYC score reflects current operating conditions while the county score reflects conditions almost a year ago.
A restaurant's "current" score means different things in different cities. In NYC, where inspections are frequent, last month's score reflects today's operation. In a jurisdiction that inspects once a year, last month's score might not exist - and the most recent score could be 11 months old. Any valid comparison must account for score freshness, not just score value.
Scheduled vs. Complaint-Driven Inspection Mixing
Most jurisdictions mix routine scheduled inspections with complaint-driven visits in their public data, and they don't always make it easy to distinguish them. A complaint-driven inspection at an establishment with a specific reported problem produces data that is not comparable to a routine inspection at the same establishment under normal operating conditions. Combining the two types into a single "current score" without flagging the inspection trigger produces a distorted picture.
Some jurisdictions handle this better than others. San Francisco explicitly tags inspection type in its open data. Many county-level systems do not, making it impossible to determine from the public dataset alone whether a given inspection was routine or complaint-driven.
The Case for a Unified Score
The practical implication of everything described above is that any application, platform, or enterprise tool that wants to display or act on health inspection data across multiple US markets needs either to build its own normalization layer or use an API that has already solved this problem.
The unified API approach handles this by applying a consistent scoring methodology regardless of the source system. For every jurisdiction, the same calculation applies: start at 100, subtract 25 for each critical violation, subtract 5 for each non-critical violation, and reduce each deduction by 2 if the violation was corrected on-site. This produces a score that is directly comparable across cities - a 78 in a Houston-based establishment means the same thing as a 78 in a Chicago establishment, because the underlying inputs have been normalized through the same lens.
The methodology has a specific design rationale. The 25-point deduction for critical violations is calibrated so that a single serious critical violation - temperature abuse, pest evidence, sewage backup - produces a score below 75, which is a B-range result that clearly signals a concern even if everything else in the establishment is perfect. Two critical violations in a single inspection push the score below 50, into F territory, which signals an establishment with systemic safety problems rather than a one-time oversight.
The 5-point deduction for non-critical violations is intentionally smaller, reflecting the real-world health risk difference between a temperature log not being maintained (paperwork violation, low direct risk) versus raw chicken stored above ready-to-eat salad (cross-contamination, high direct risk). The -2 point correction credit for violations fixed during the inspection rewards responsive management without eliminating the violation's impact entirely - finding and correcting a problem during inspection is better than not finding it, but it's still evidence of a lapse.
The result is a score that can be embedded in restaurant listings, franchise monitoring dashboards, delivery platform algorithms, or food inspection alert systems with a consistent interpretation: 85 and above is A-grade performance across the country. Below 70 is a B-range concern that warrants attention. Below 50 is an F that requires immediate action. The same scale, the same thresholds, every city.
For developers and data teams building on top of this normalized score, the FoodSafe Score API eliminates thousands of lines of jurisdiction-specific parsing code and the ongoing maintenance burden of tracking when individual health departments change their reporting formats, update their violation classification schemes, or migrate to new data portals. The normalization layer is maintained centrally, so downstream applications stay current without any engineering effort on the consumer side.
Summary: What to Take Away
Restaurant health inspection systems across the US are a patchwork of incompatible formats, scales, and methodologies - each individually defensible within its own regulatory context, but collectively impossible to use at national scale without normalization. New York counts up from zero. Los Angeles measures percentage of compliance. Chicago issues a binary verdict. San Francisco and Houston both use 0-100 scales but with different grade thresholds and statistical distributions. Seattle uses color cards with no numeric component at all.
Understanding these differences is essential for anyone building consumer-facing restaurant discovery tools, franchise compliance systems, insurance risk models, or public health analytics. The data is public, but it is not plug-and-play. The path from raw government inspection records to a usable, cross-market signal runs through a normalization methodology that accounts for score direction, grade thresholds, violation severity weighting, inspection frequency, and trigger type.
For most teams, building that normalization layer from scratch - and maintaining it against an ever-changing landscape of jurisdiction-level updates - is not the right use of engineering resources. The alternative is to use a purpose-built data layer that has already solved it.