Your safety stats look great. That’s the problem.

Safety in numbers. That’s the simplest way to understand the decades-long reliance on a handful of health and safety metrics that, while well-intentioned, may be masking the biggest risks to employees. Lost time injury frequency rates (LTIFR), total recordable injury rates (TRIR) and their variations have become the common language of safety across geographies and industries.

Their appeal is easy to understand: they provide a simple, widely recognised and comparable way to track performance and communicate results in complex operating environments. As a result, they have become deeply embedded in board reports, annual disclosures and stakeholder communications. Many organisations may rely on them without questioning whether they still provide the most meaningful picture of risk.

The widespread reliance on these metrics is also bolstered by encouraging progress on health and safety since their introduction in the 1970s. In the United States, data from the Bureau of Labor Statistics show that the rate of nonfatal workplace injuries and illnesses fell from 10.9 cases per 100 full-time workers in 1972 to 2.8 in 2018, a reduction of almost 75 per cent. Similar trends can be seen across many developed economies. Eurostat data show that incidence rates have fallen over time, from around 1,700 per 100,000 workers in 2013 to around 1,400 in 2023, while Australian safety statistics point to significant long-term reductions in workplace injuries despite year-to-year fluctuations.

While nonfatal incidents have declined, serious incidents and fatalities continue to occur across high-risk sectors such as infrastructure, construction and engineering. Organisations can report low injury rates while remaining exposed to events with the potential for catastrophic consequences.

In an area where the only acceptable number is often zero, the continued occurrence of life‑altering and fatal incidents raises an uncomfortable, but necessary question: Are we measuring what really matters?

Injury metrics were not designed to measure exposure to serious risk. They measure past harm, often treating very different events in the same way. A minor injury and a potentially catastrophic near-miss can sit side by side in the data, while the conditions that can lead to fatal incidents remain largely invisible.

A 2020 study by the Construction Safety Research Alliance, based on more than 3 trillion worker-hours of safety data, found no meaningful relationship between TRIR and serious injuries or fatalities. It also showed that, without extremely large exposure volumes, these metrics are not statistically reliable for comparing teams, projects or organisations, raising questions about how and why they are used to inform decisions.

So why do these metrics persist as the dominant signal of workplace safety?

Part of the answer lies in how they shape attention. When safety is defined by what can be counted, organisations tend to focus on reducing visible, recordable events.

Research in safety science shows that sustained operational success can conceal growing risk, and the absence of incidents does not necessarily indicate a well‑functioning system. Where performance is judged on maintaining low numbers, weaker signals such as near misses or emerging hazards are more easily overlooked, and incremental trade-offs made under pressure can accumulate unnoticed until a triggering event occurs.

In some cases, the heightened focus on injury rates can actively distort behaviour. When performance, incentives, or reputation are tied too tightly to “zero harm” type metrics, underreporting, claim management and classification debates can take precedence over learning. The metric becomes the objective, rather than the safety outcome it was meant to represent.

If safety is about preventing harm, risk deserves to be at the centre of how performance is measured.

A risk-based approach shifts attention to the presence and effectiveness of controls for material hazards such as energy isolation, working at height, confined spaces, vehicle use and other critical risks specific to the organisation. The focus moves from outcomes to prevention, and from individual incidents to system performance.

This approach aligns far more closely with how other enterprise risks are managed. Cybersecurity, financial integrity and operational resilience are not judged by how many losses occurred in a reporting period, but by the strength of controls, the quality of assurance and the visibility of emerging risks. Safety should be no different.

Moving beyond recordable injury rates does not mean abandoning measurement. It means measuring performance differently. Leading organisations are placing greater emphasis on indicators such as critical control performance, exposure to high-risk work, the quality of hazard reporting and the ability to learn from high-potential events. These measures provide a clearer view of whether risks are understood and actively managed.

They also shift executive and board discussions from target achievement to preparedness to prevent serious harm.

If the current model holds, most organisations will continue to improve what they measure rather than what matters. Injury rates may stabilise or even improve while exposure to high‑consequence risk remains unchanged.

Over time, strong numbers can create confidence that is not matched by underlying risk control, increasing the likelihood of infrequent but high‑impact events that challenge leadership assumptions and erode trust.

A different path is already emerging. Organisations that focus on risk exposure and control effectiveness gain earlier visibility of where serious harm may occur and why controls might fail. This enables intervention before harm, not after. It is reinforced by more forward‑looking indicators such as the effectiveness of critical controls, exposure to high‑risk work and the quality of learning from near misses.

Standard safety metrics such as TRIR and LTIFR still have a role, but they should not define how safety performance is understood or managed. Organisations that distinguish between managing minor injuries and controlling fatal risks are better positioned to prevent serious harm. Injury metrics can provide useful context, but attention should be directed towards critical risks, control performance and prevention.

Ultimately, people are not protected by statistics. They are protected by effective controls for the risks most likely to cause serious harm. If we want to improve safety outcomes, our measures of success must reflect that reality.