What the raw data actually says
CY2023–CY2025OSHA publishes establishment-level injury data every year. It is the only public source granular enough to answer "is our rate any good?" - and it is close to unusable straight out of the download. These are the things we found that change the answer.
Finding 1 - the 200,000-hour trap
Finding 2 - zero is not excellence
Finding 3 - injuries cluster inside the shift
Benchmark my site
percentile against real peersEnter one establishment's numbers. You are compared against establishments in the same NAICS industry and size band that filed for CY2025 - not against a national average. A lower percentile is better.
Company search
Every employer that filed a 300A summary is here, resolved to a corporate entity by EIN where the filed names agree. Search by name - results show the rolled-up rate across all that company's establishments, with a per-state and per-industry breakdown underneath.
Largest filers by reported hours
How people actually get hurt
688,367 OIICS-coded casesRates tell you how much. Case detail tells you what. Every case below is coded to the Occupational Injury and Illness Classification System - event, body part, nature, source. Expand any row to see what it breaks down into and how many days it cost.
Event - what happened
Body part injured
Hours into the shift
Month of incident
Sector rates and trend
hours-weighted, 2023–2025Hours-weighted aggregate rates by NAICS sector. Expand a sector to see the industries inside it - sector averages hide enormous spread, and the sub-industry view is where the actionable comparison lives.