The site review that couldn’t find a safety failure — and the warning signs nobody acted on

The conversation usually starts the same way. There has been a death. The review was thorough and found no failed physical controls — no missing barrier, failed lockout, or skipped permit. The system did what it was designed to do. And yet, when people looked back, the warning signs were obvious: missed shifts, uncharacteristic irritability, and concerns that colleagues had noticed but had nowhere to take. Nobody acted on them. As one operations leader put it: “Our physical-safety culture is mature and accountable. The psychosocial side barely exists.”

That is not a culture failure. It is a design gap. Being precise about the difference is what makes the problem solvable.9

Why couldn't a good safety system find this?

A physical-safety system is built around a specific logic: identify a hazard, assess the risk, apply the hierarchy of controls, verify the control held. When something goes wrong, the investigation traces backwards to the control that failed. That logic is powerful precisely because it is observable and auditable — you can verify whether a control existed, whether a procedure was followed, and whether a sign-off occurred.

Psychosocial harm does not present as a failed control. It presents as a person, and it presents slowly. There is no barrier that was removed, no permit that lapsed. By the time it is obvious that someone is struggling, the warning signs have often been visible for weeks — but they sat in a blind spot between what crews could see and what the system could record. However, the investigation comes back clean because it is asking the questions it was built to ask, and those are the wrong questions for this hazard.

For boards, the critical question is not how the outcome occurred, but how much was visible before it did. In construction, a worker is around eight times more likely to die by suicide than in a workplace accident (MATES in Construction, 2025). Yet the industry's most mature safety systems currently have been built to prevent the event that is far less likely to occur.

None of this displaces physical safety. The point is narrower and more uncomfortable: the system is highly effective at preventing the risks it was designed to see, and far less effective at recognising the ones unfolding in people.

Is this a wellbeing problem or a risk problem?

It is a risk problem, and the distinction matters for who owns it. Psychosocial hazards now sit inside WHS duties. They belong on the risk register, inside the psychosocial safety framework, subject to officer due diligence and the reasonably-practicable test — the same machinery already applied to a fall hazard or a confined space. The moment a pattern of warning signs is foreseeable and there is no control in place for it, the exposure is not theoretical. This is not primarily a wellbeing issue. It is a legal, regulatory, and insurable risk that lands in operations.

This is the point that often shifts the conversation in safety-mature organisations, because it speaks their language. No organisation would accept "we have an EAP" as a control for working at height. An EAP is reactive support that activates after harm — it is the bottom of the cliff. It is necessary. It is also downstream of every decision that determines whether someone ends up there. The missing control sits at the top of the cliff, upstream of the EAP, in the weeks when distress is visible, concerns are being noticed, and nobody has a clear way to act on them.

Why does "see something, say something" keep failing here?

Because people have been asked to act, without giving them anything to act with. Right now, the warning signs are carried by whoever happens to notice them — a leading hand, a foreman, a peer. They become accidental counsellors, carrying a concern they are not equipped to hold and have no route to escalate. So they do the rational thing and say nothing, or they jump straight to the EAP referral, which for someone two steps before crisis feels like an overreaction and reliably gets refused.

On paper, the answer is simple: tell the supervisor. In practice, there are structural reasons why many concerns never take that path. Across the research, workers are often more willing to raise suicide and mental-health concerns with trusted peers than with supervisors, while peer-led approaches have demonstrated positive outcomes (Gullestrup et al., 2023). This is not a criticism of supervisors. It is a recognition that reporting lines and trust networks are not the same thing. Effective controls account for both.

This is where Dunbar's number matters. Research suggests people can maintain only a limited number of close, meaningful relationships (Dunbar, 1992; Dunbar, 2010). A site leader cannot know 500 workers well enough to spot subtle changes in behaviour, and it is unrealistic to expect them to. But someone knows their fifteen — the people they work alongside every day. That is where the signal lives. The question is whether the organisation has built a way to capture it.

What does the missing layer actually look like?

A trained champion network embedded where people are actually known. Its role is simple: recognise concerns early, start conversations, and escalate when needed. It sits between "looks okay" and "needs support" — providing a clear pathway for action before a situation becomes a crisis.

This is what makes the invisible risk visible. It gives early distress a name, a pathway, and a response — much like a physical-safety system does for a hazard. Done well, it reduces crises while increasing appropriate EAP uptake, because people are connected with support earlier, by someone they trust and at the point it is most likely to help.

The case for action is grounded in risk, not sentiment. Construction recorded the third-highest fatality rate of any industry in 2023 — 3.4 deaths per 100,000 workers against a national average of 1.4 (University of Melbourne / The Conversation, 2025). When psychosocial harm is considered alongside physical injury and fatality, the sector's largest blind spot is no longer what it controls, but what it does not.

How does this fit with existing safety and wellbeing programs?

Alongside the safety system, not inside the EAP. It is the operating model for psychosocial risk — giving wellbeing the same structure, escalation pathways, and accountability that physical safety has long relied upon. The Pathway builds that capability through a champion network, clear governance, and a repeatable response process. The result is a control you can point to in a due-diligence review, rather than a hope you can point to in a coronial one.

The next review will almost certainly be thorough. The real question is whether, by then, the warning signs have somewhere to go.