A materials recovery facility in Gainesville finished 2024 with a TRIR of 3.1, an EMR trending toward 0.9, and a safety record his operations VP was comfortable presenting to clients. Then five weeks into January 2025, three recordables in a row — including a DART case that put one of their back-line sorters on restricted duty for six weeks.
Nothing in the lagging data had flagged it coming. Nothing in the prior year's numbers suggested the program was drifting.
When we went through the last eight months of records together, the signal was there. The facility's near-miss log had gone quiet starting in August. Not because near misses stopped happening on the sorting line, but because the foreman on the back end had stopped filing them. Workers had figured out that reports generated paperwork and not much action. The corrective action queue had 23 items open, some of them four months past due. The safety observation rate, which supervisors had been running at about three per week in Q1, had dropped to near zero by Q3. By the time the recordables started in January, the program had been drifting for most of a year. The prior year's TRIR had made it invisible.
Leading safety indicators measure the conditions and activities that precede incidents before they happen. Lagging indicators measure outcomes after the fact. TRIR, DART rate, and EMR are all lagging: they're accurate, but backward-looking. The three leading indicators with the strongest track record in light industrial operations are near-miss reporting rate, safety observation completion rate, and corrective action close rate. Organizations that actively track leading indicators see measurably lower incident rates over time, according to research from the Campbell Institute, the research arm of the National Safety Council. A practical starting point: pick one of the three and track it consistently for 90 days before adding the others.
What Lagging Indicators Can't Tell You
TRIR counts every recordable case per 100 full-time-equivalent workers. DART narrows that to cases serious enough to cause missed work, restricted duty, or a job transfer. EMR applies a three-year claims history to adjust your workers' comp premium. They're all measuring what already happened.
That's useful. You need to know your rates. You need to compare them to BLS benchmarks by NAICS code. You need your EMR before a renewal cycle and your TRIR before a prequalification review. But every one of those numbers is a report on the past 12 to 36 months of your safety program, not a forecast of next month.
For a long time, we'd tell clients that staying on top of the lagging numbers was enough if those numbers were good. We don't say that anymore. The Gainesville account above had good lagging numbers. The program was still drifting. You can't see that in a TRIR until a recordable lands on your 300 log.
The lag isn't just semantic. It's operational. A pattern of risky behavior on a sorting line, a backlog of unresolved corrective actions, a foreman who stopped doing safety walkthroughs because he's short-staffed: all of that can develop over months before a single incident hits your recordable count. By the time the lagging metric moves, you've lost the opportunity to intervene. Most light industrial recordable clusters we've seen follow a period of program drift that was visible in leading data 60 to 120 days before the first incident.
Near-Miss Rate: The Signal Operations Skip
A near miss is an unplanned event that didn't result in injury, illness, or property damage but, under slightly different circumstances, could have. A forklift that brakes hard and stops two feet from a picker who stepped into the aisle without looking. A drum that tips and spills but doesn't catch anything. A conveyor jam cleared without lockout because the foreman was in a hurry and nobody was watching.
OSHA recommends near-miss tracking explicitly and has published a template near-miss reporting policy on its website. The guidance is clear: a high near-miss reporting rate reflects an engaged workforce identifying hazards before they escalate. A low rate doesn't mean the hazards aren't there. It means they're not being reported.
Safety management research suggests that mature safety programs typically see 50 to 100 near-miss reports filed for every recordable injury. Most light industrial operations we work with are nowhere near that ratio when they first set up a near-miss system. A typical starting point looks more like three to five near misses per recordable. Getting toward a 20:1 or 30:1 ratio over time usually tracks with a corresponding drop in recordables, because the reports are surfacing hazards that would otherwise go unaddressed until someone gets hurt.
The number itself isn't the goal. An MRF in Conyers that reports 40 near misses per recordable and closes the corrective actions within 30 days is doing something categorically different from an operation reporting 40 near misses and parking them in a queue. The rate tells you your workers are identifying hazards. What you do with the reports tells you whether that identification is preventing anything.
Operations where near-miss reporting is discouraged, either directly or through indifference, don't have fewer incidents. They have fewer reports. This was what happened at the Gainesville MRF. The foreman wasn't suppressing reports maliciously. Nobody was following up. So workers stopped filing them. The hazards were still there; the system had just gone silent.
Safety Observation Rate
A safety observation is a structured, documented review of how a worker is performing a specific task against a specific standard. Not a general walkthrough or a "things look okay" check, but a deliberate observation of a defined procedure. How the pallet jack is being operated in the narrow aisle. Whether the picker is following PPE protocol at the conveyor infeed. Whether the forklift operator is checking mirrors before reversing in a pinch zone.
Safety observation rate is typically measured as the number of completed formal observations per supervisor per month. Some operations track it by shift instead. The specific cadence matters less than the consistency.
What makes this a useful leading indicator is that it catches behavioral drift before it produces a recordable. A line where supervisors are running three to four formal observations per month is a line where someone is paying deliberate attention to how tasks are being executed. When that rate drops to near zero, as it did in Gainesville by Q3 2024, it's a flag that supervisors are too stretched, too disengaged, or have stopped treating safety observations as a real part of the job rather than optional paperwork.
For operations with a heavy new-hire flow, observation rate matters especially in the first 60 days of a placement. We've documented this pattern in our post on reducing new-hire safety incidents in light industrial settings: most recordables in warehousing and manufacturing cluster in the first 60 to 90 days because new workers are learning tasks under productivity pressure in an environment they don't know well enough to recognize what a hazard looks like. Supervisors running structured observations during that window catch the behavior before it produces a DART case. Supervisors who aren't doing structured observations don't catch it until after.
Corrective Action Close Rate
Every safety observation, near-miss report, and audit finding should generate a corrective action: a specific task assigned to a specific person with a specific due date. Corrective action close rate measures the percentage of those tasks completed on time.
The benchmark cited most often in safety management literature is 85% on-time closure. If you have 20 corrective actions due in a given month, 17 of them are completed by the assigned date. The other three are escalated or rescheduled with a new owner, not just left sitting.
In practice, corrective action close rate is the leading indicator most operations find hardest to maintain because it requires someone to own the queue and follow up. Near-miss reports are filed by workers. Observations are done by supervisors. But corrective actions often span functions: maintenance, operations, HR, procurement. Without a clear owner for the queue and a consistent review cadence, the pile builds.
An overdue corrective action backlog is one of the clearest advance signals of an upcoming recordable cluster we've seen in Georgia accounts. The Gainesville MRF had 23 open corrective actions by September, some dating back to June. A few were for physical hazard corrections in the sorting area. Two were for updated lockout/tagout procedures on the glass line. None of them were addressed before the January cluster.
If you only track one leading indicator, make it this one. A corrective action queue that's consistently at 85% or better on-time closure is a program that closes its own loops. A queue where items are perpetually overdue is a program that identifies problems and doesn't fix them. The difference shows up in your recordable count. Usually within 6 to 12 months.
A Dashboard for Monday Morning
You don't need a software platform to start tracking leading indicators. A shared spreadsheet works for most light industrial operations under 200 workers. What you need is a review cadence that puts the numbers in front of the right people at the right time.
Track near-miss reports weekly. Assign each one to an owner within 48 hours. Review the open list at the Monday morning safety brief. Any report more than two weeks old without an assigned corrective action gets raised to the plant manager that week, not next month.
Track safety observations monthly by supervisor. Each supervisor submits a count at the end of the month. Any supervisor at zero for the month gets a conversation with their plant manager before the next shift cycle. That conversation should be brief and direct: not punitive, but the absence shouldn't go unaddressed.
Track corrective action close rate monthly. Everything due in the prior month, percentage completed on time. Anything more than 30 days past due gets escalated to the operations VP. A queue review that happens monthly and surfaces the oldest open items is usually enough to keep the rate above 80%.
Three metrics on one page, reviewed monthly, tells you whether your safety program is active or just documented. Most operations we work with track TRIR carefully and check near-miss and corrective action data once a quarter at best. The TRIR isn't the problem. The quarterly cadence on the metrics that would have predicted the TRIR is the problem.
These three leading indicators connect directly to the broader KPI framework we outlined in Staffing KPIs That Actually Predict Client Retention, where we cover how to build a safety and operations data dashboard that gives an ops VP a single view of account health. Safety program activity isn't just an EHS function. It affects fill rates, turnover, and the on-site reputation that determines whether your best workers stay through a full placement cycle.
For context on where your TRIR and DART rate land against sector peers, the TRIR benchmarks by industry post covers 2024 BLS data at the four-digit NAICS level for warehousing, food manufacturing, recycling MRFs, and light manufacturing. Your lagging number matters most when you read it against the right comparison group.
We staff warehousing, recycling, food manufacturing, and light industrial accounts across Gainesville, Lawrenceville, Conyers, Smyrna, and the Atlanta MSA. We track safety KPIs alongside fill rate and turnover for every account we manage. If you want to talk through how a leading indicator system fits into your current safety program, or if you're evaluating how a staffing partner should contribute to your safety data, Get Started and tell us what you're working with.
