Stop Counting Injuries. Start Building Capacity.

For forty years, the safety profession has trusted a number that is mostly statistical noise. Here is why TRIR misleads, and what to measure instead.

Twenty years ago, in a shipyard in Singapore, I watched a system absorb a death.

A worker fell from height. I did not see the fall. I saw the aftermath. The silence. The stretcher. The body carried away. Then a whistle blew, the area was cordoned off, and the paperwork began. Five minutes later another whistle blew, and every one of us went back to work.

I have spent twenty years trying to understand that second whistle. The machines restarted. Production resumed. And a few weeks later, a number was updated on a dashboard, and that number still looked fine.

That is the problem I want to put in front of you. Not the fall. The dashboard.

The bucket is the wrong instrument

Neil deGrasse Tyson has a way of teaching the limits of measurement. Imagine I asked you to count the whales in the ocean, he says, and I gave you a bucket. You could dip that bucket a million times over a million years and never catch a whale. The whales are real. The bucket is simply the wrong instrument.

The safety profession has been holding a bucket for forty years. We call it the recordable injury rate. TRIR, TRIFR, LTIFR, the family of lagging indicators we put in front of boards and write into contracts. We have been dipping it into an ocean full of things it cannot catch.

Here is what the research actually says. The Construction Safety Research Alliance examined seventeen years of data across more than three trillion worker hours. Their finding: the occurrence of recordable injuries is between 96 and 98 per cent random. A recordable rate cannot be represented as a single point estimate. It is statistically meaningless to compare two organisations by it. And there is no correlation between recordable injury rates and fatality rates. None.

Let me make that concrete, because the abstraction lets people off the hook. Picture an operation that expects five recordable injuries a year. By the simple mathematics of low-frequency events, going from five injuries one year to eight the next, a sixty per cent jump that will get a safety leader called into a room, is statistically indistinguishable from nothing having changed at all. How many quarterly reviews have you sat through where exactly that kind of swing drove the entire conversation?

It gets worse. The safer an operation becomes, the noisier its injury rate gets in relative terms. The metric is least informative precisely where it is most celebrated.

The metric is not a passive instrument

If the recordable rate were merely uninformative, that would be one thing. It is not merely uninformative. It is active.

When a measure becomes a target, it stops being a good measure. The moment injury rate became the number that decided bonuses, contracts and reputations, it began to be managed. Minor injuries get reclassified as first aid. Modified-duty programmes keep a recordable from becoming a lost-time case. Higher-risk hours move onto a subcontractor's books. None of that is fraud. It is rational behaviour for people whose standing depends on a number. And every piece of it suppresses the information the organisation most needs to see.

The pattern shows up in the wreckage. Texas City refinery, 2005. Macondo, 2010. Pike River, 2010. Three different industries, three jurisdictions, and in each case a personal-safety record that looked acceptable, or was actively being celebrated, in the period before the catastrophe. If a metric can fall reliably in the years before a disaster, then a falling metric cannot be evidence that the conditions for disaster have reduced. That is a point of logic. More data will not fix it.

Safety was never the absence of injuries

So what do we measure instead?

We change the question. Safety is not the absence of injuries. It never was. Safety is the presence of capacity: the ability of a system to absorb the variability that real work always produces, and to recover when something fails.

That is a measurable idea. Sidney Dekker and Michael Tooma made the academic case for it in 2022. Capacity is built deliberately, in advance of the operational surprise that will inevitably arrive. You can measure whether your people have the competence, the resources and the time. You can measure whether your barriers are holding, by tracking potential severity alongside actual severity and watching the gap between them. You can measure whether reporting is flowing or freezing.

And here is the part the safety profession has been slow to accept. Most of the capacity in your organisation lives on the human side. Fatigue. Cognitive load. Recovery time. Psychological safety. Role clarity. These are not feelings to be managed quietly inside HR. They are the leading indicators of whether a person can absorb a bad moment or be overwhelmed by it. When a system depletes its people, it depletes its own capacity to fail safely.

This is why I argue that wellbeing is not a soft initiative. It is the human-side measurement of capacity. It is operational intelligence. And it is the one leading indicator of human reliability that you can act on before the chain runs from stress to fatigue to error to incident.

The whistle has blown

The worker in that shipyard was not unlucky. He was predictable. He was six days a week into a four-month stretch in the heat. His reserves were gone before the shift began. The investigation almost certainly closed with two words: human error. But the human was not the error. The system was.

The whistle has blown for our profession. The recordable rate is no longer enough, and most of the people I speak to already know it. The harder question is what to build instead.

Ian Collins is the Founder and Managing Director of Wellbeing Daily, a consultancy that turns wellbeing into measurable operational intelligence for organisations in safety-critical industries. He is the author of The Wellbeing Imperative.