How the Safety Score works
The score is one fixed method applied the same way to every employer. No part of it can be paid for, edited, or removed.
1. Citation history (the core of the score)
Every citation counts according to how serious OSHA classified it. Willful, repeat, and failure-to-abate citations weigh far more than routine "other-than-serious" ones, because they reflect a different order of risk and employer knowledge.
Citations are also discounted by age: an older citation counts for less than a recent one, so a site that has since cleaned up isn't condemned forever by a lapse from years ago. The weighted, time-discounted total is measured against how often the establishment has been inspected.
2. Injury rate vs. the employer's own industry
Where an employer reports to OSHA's Injury Tracking Application (ITA Form 300A), we use its DART rate (days away, restricted, or transferred) and TRC rate (total recordable cases), computed with the standard OSHA formula cases × 200,000 ÷ hours worked, and compare them to the average for its NAICS industry.
An establishment sitting exactly at its industry average lands in the middle of this component, not the top; "average" isn't "clear". When an employer isn't required to report to ITA (most small ones), this component is simply absent and the verdict is labeled limited data. Missing injury data is never treated as zero incidents.
3. Bands and the severe-citation rule
Scores are banded into Clear Record, Some Concerns, and High Risk.
A recent willful, repeat, or failure-to-abate citation means the record can never read as "Clear", regardless of how the rest of the arithmetic lands. A pattern of recent severe citations pushes the verdict to High Risk.
4. Confidence, not false precision
Every verdict carries a confidence label. Strong means multiple inspections or reported injury data. Limited data means a thin record; read it as "little evidence", not a clean bill of health. Unverified means the record couldn't be confidently matched to a single establishment, so no score is published.
5. Matching records to the right establishment
OSHA records are keyed loosely on employer name and address, and multi-location employers have to be kept apart. A record only attaches to an existing establishment when the normalized name is a strong match and the physical address corroborates it.
A same-name business in a different city or ZIP is treated as a separate establishment, not merged. Uncertain matches go to a human review queue rather than being guessed, because mistaken-identity attribution is the risk this product takes most seriously. Each physical location is scored on its own; company rollups are shown as an explicitly separate view.
Data & vintage
- OSHA enforcement (inspections, citations, penalties): U.S. Department of Labor Open Data Portal, updated daily upstream.
- Injury & illness rates: OSHA Injury Tracking Application (ITA) Form 300A public dataset, updated annually.
- Industry benchmarks: BLS Survey of Occupational Injuries and Illnesses (SOII) by NAICS. Current reference: BLS SOII 2023 (sector fallback).
Raw source data is archived on every ingest so any past score can be reproduced. HazardLedger is independent and is not affiliated with, endorsed by, or sponsored by OSHA, the U.S. Department of Labor, or BLS.
Corrections
If a record is attributed to the wrong employer or is out of date, use the correction link on any verdict page. Corrections go to a moderation queue and are checked against the source records. No employer can pay to change, improve, or remove its score.