EMR vs. TRIR: Why the Two Safety Scores Disagree
One number comes from your insurance data, the other from your OSHA logs. They measure different years, count different things, and routinely tell prequalification reviewers two different stories.
EMR and TRIR measure safety from different records. TRIR counts OSHA-recordable incidents per 200,000 hours worked in a single year. EMR compares three years of dollar-weighted workers' comp claims against expected losses for your size and classification. Because the inputs differ, a contractor can hold a low TRIR and a high EMR at the same time.
A prequalification reviewer opens your file and sees two safety scores. Your TRIR is 1.8, better than most of the industry. Your EMR is 1.24, worse than average. One of these numbers must be lying.
Neither is. EMR vs. TRIR isn't a contest between a right number and a wrong one. They're built from different records, over different time windows, counting different things. Understanding the gap between them is worth real money, because one of the two can be corrected and the other can only be outlived.
What TRIR measures
TRIR (Total Recordable Incident Rate) comes from your OSHA logs. The formula is recordable incidents multiplied by 200,000, divided by total hours worked. The 200,000 is a normalizing constant: 100 full-time employees working 40 hours a week for 50 weeks. It converts your raw incident count into a rate per 100 full-time workers so a 40-person framing crew can be compared against a 4,000-person industrial contractor.
What counts as recordable is defined by OSHA's recordkeeping rules: work-related injuries and illnesses involving medical treatment beyond first aid, days away, restricted duty, job transfer, or loss of consciousness. Every recordable counts as one incident, whatever it cost. A laceration that took three stitches and a shoulder reconstruction that took $250,000 in surgery and eighteen months of indemnity each add exactly one to the numerator.
TRIR is also short-memory. It's typically reported for a single calendar year, so a bad stretch rolls off quickly.
What EMR measures
Your experience modification rate comes from a different pipeline entirely: the claims data your workers' comp carrier reports to NCCI or your state rating bureau. It spans three completed policy years, not one. And it's dollar-weighted, with the formula splitting each claim into a heavily weighted primary portion and a discounted excess portion, which is why frequency moves the mod more than severity.
EMR is slow-memory. A claim sits in the calculation for three rating years, valued at whatever the reserve said on the valuation date, whether or not the claim later settled for less. TRIR forgets last year's incident. Your mod is still billing you for it two renewals later.
Why the two scores disagree
Put the two designs side by side and the disagreement stops being mysterious. Five small medical-only claims barely dent a TRIR average but can stack real points onto a mod, because the formula punishes frequency. One catastrophic claim does the opposite: it wrecks neither number proportionally, since TRIR counts it once and the mod discounts its excess portion.
Then there's the input-quality difference. TRIR is your own count of events that either happened or didn't. EMR is a calculation on carrier-reported dollars: reserve estimates, classification assignments, payroll allocations. Events are hard to misreport. Dollars are easy to misreport, and nobody audits them on the way into the formula.
That's why the low-TRIR, high-EMR contractor is such a common sight in prequalification files, and why platforms that grade on both, ISNetworld and Avetta among them, can show a contractor passing one bar and failing the other in the same submission. When incident counts say safe and claim dollars say risky, the honest explanations are severity, bad luck, or bad data. In our reviews of Southeast contractor worksheets, bad data carries more of that gap than most owners assume.
Which number you can actually fix
You can't revise last year's incident count. TRIR improves one way: fewer recordables going forward.
The mod is different. It's arithmetic on reported data, and when the data is wrong, the number is wrong, and a wrong mod can be disputed and corrected, sometimes with retroactive refunds. A contractor staring at a 1.24 next to a clean TRIR shouldn't treat the mod as a verdict on their safety culture. They should treat it as a claim about their data, and claims can be checked.
What an audit would check
An audit checks whether the story your mod tells matches the story your loss runs and OSHA logs tell. That means reconciling worksheet claim values against current carrier records, checking how claims were coded, and confirming the classifications behind your expected losses. When TRIR and EMR point in opposite directions, the worksheet usually explains why. Send us your NCCI worksheet and we'll review it for free.