Turnover and subcontractor layering don't just thin your workforce—they silently collapse the near-miss data your triangle depends on.
Herbert Heinrich published his accident ratio in 1931—300 near-misses for every serious injury, 29 serious injuries for every fatality. Frank Bird updated it in 1966 with data from 1.75 million accidents across 297 companies: 1 major injury, 10 minor injuries, 30 property damage incidents, 600 near-misses. Both models rest on the same assumption: that you're actually capturing what's happening at the base of the triangle. In a 200-person stamping plant with 40% annual turnover, three active staffing agencies, and a rotating cast of maintenance contractors, that assumption collapses before the end of Q1.
The triangle is not a description of reality. It's a description of your reporting system. When a press operator on their third week from a temp agency slips on coolant near a Heim joint and catches themselves on a guard rail, the probability that incident makes it into your near-miss log is close to zero. They don't know the reporting procedure. They don't know the form. They don't know whether reporting it gets them sent home or flagged as a problem. So it doesn't get reported.
Your near-miss count stays flat. Your safety coordinator sees a flat triangle and reads it as a stable risk profile. That's not stability—that's silence. And silence at the base of the triangle is exactly what a serious injury or fatality looks like just before it happens.
This is the counterintuitive reality most safety managers miss: a dropping near-miss report rate is not good news. Most plant safety managers assume a low near-miss count means fewer hazards. The real problem is that new hires and subcontractors have no psychological safety, no procedural fluency, and no reason to trust the reporting system—so the hazards are still there, they're just invisible to you now.
Here's a scenario that happens in mid-size general industry plants constantly. A Tier 1 automotive parts supplier runs three production lines. They're pulling from two staffing agencies because they can't keep direct hires—exit interviews say the same things about shift scheduling, but nobody upstream acts on them. On top of that, their predictive maintenance program relies on a contracted millwright crew that rotates every 90 days based on the prime contractor's project schedule.
The millwrights do LOTO isolation on a hydraulic press. They follow the procedure—mostly. One millwright on day three of his rotation has been verbally walked through the energy isolation points but hasn't had formal hands-on demonstration. He isolates four of five energy sources. Nothing bad happens. He doesn't report it as a near-miss because he's not sure he did anything wrong, and he's not going to raise his hand in week one.
Your LOTO audit shows 94% compliance. Your near-miss log shows three entries for the month. Your Bird's Triangle looks fine. The hazard is real, it's active, and it's completely invisible to your data.
When that press cycles unexpectedly and a maintenance tech loses a hand, the OSHA 300 recordkeeping failure that follows is almost secondary to the fact that every leading indicator you were tracking was being fed corrupted data for months before the event.
Bird's ratio was derived from companies where workers had enough tenure and familiarity to recognize what a near-miss was and trust the system enough to report it. That's not a 90-day staffing agency worker on a press line. That's not a subcontracted electrical crew that rotates between three clients and answers to a foreman they've known for a week.
What actually happens in these environments is that the triangle inverts at the base. The SIF-potential events are still occurring at the ratio Bird described—but they're surfacing as recordables and hospitalizations before they ever show up in near-miss reports, because the near-miss reports never existed. You're not working from a triangle. You're working from a spike.
OSHA serious violations run up to $16,131 per citation. Willful or repeat violations reach $161,323. But that's before the workers' comp claim, the DART rate spike that follows you for three years in your EMR, and the customer audit that quietly pulls your preferred supplier status because your TRIR no longer qualifies you for their contractor threshold. The fine is the smallest number on the invoice.
Three things that actually work in high-turnover manufacturing environments:
The OSHA incident investigation guidance is clear that near-miss reporting is a foundational element of any injury prevention program—but it can only function in an environment where reporters trust the system and understand how to use it. Turnover destroys both conditions simultaneously, and it does it quietly.
Heinrich's original 1931 ratio was 1 fatality : 29 serious injuries : 300 near-misses, based on insurance claim analysis. Bird's 1966 update used direct accident data from 1.75 million events and added property damage as a distinct layer: 1 major : 10 minor : 30 property damage : 600 near-misses. Bird's model is generally considered more empirically grounded, but both depend on consistent near-miss reporting to produce a meaningful picture of underlying risk.
New workers—especially temp or contract workers—rarely report near-misses. They don't know the reporting system, they don't know whether reporting creates negative consequences, and they often don't have enough operational context to recognize a near-miss as reportable. The result is that the base of the triangle shrinks artificially while the underlying hazards remain constant or grow.
With suspicion, especially during periods of high turnover, rapid hiring, or subcontractor use. A flat near-miss rate in a stable, tenured workforce can indicate genuine risk reduction. The same number in a high-turnover environment almost always indicates underreporting—which means the triangle base is hollow and the next recordable or SIF event is not predicted by the data you're looking at.
A SIF precursor is a near-miss or incident that had the potential to produce a serious injury or fatality but didn't—usually due to luck rather than a control that actually worked. Most near-miss programs count all near-misses equally. SIF-prevention frameworks separate out high-energy events (LOTO failures, fall exposures, struck-by near-misses, confined space entries without atmospheric testing) because those carry disproportionate fatality potential and need separate tracking and investigation, not just a row in the same near-miss log.
Tracking near-miss segmentation, maintaining OSHA 300 logs across a workforce that turns over every quarter, keeping subcontractor training records current, and running LOTO compliance audits that are actually documented—that's a full-time job on top of a full-time job. EHS, Inc. takes the documentation and training administration off your plate so you can spend your time in the field where the hazards actually are. Talk to EHS, Inc.
Aaron West
Founder, EHS, Inc. — 18+ years in EHS compliance and contractor safety
Aaron West has spent over 18 years helping contractors and businesses navigate OSHA compliance, ISNetworld® certification, and workplace safety management. He founded EHS, Inc. to make enterprise-level EHS accessible to companies of all sizes — serving contractors and businesses nationwide — without long-term contracts or enterprise overhead.
Our team handles the complexity so you can focus on running your business. No long-term contracts, no learning curve.
Talk to EHSMost refinery turnaround contractors self-assess at Independent. OSHA's inspection results tell a different story.
Two risk visualization tools, one budget cycle, and a GC who can't afford to pick wrong — here's what each one actually costs you.
When OSHA shows up during a refinery turnaround, they don't grade your program on paper — they grade what they find in the field.
Framework to achieve zero incidents
Stop hitting paywalls
54 topics in English & Spanish