In 2021, 61,194 fatal workplace injuries were recorded in New Zealand. The next year, that number fell to 357. It's been roughly the same ever since: 315 in 2023, 312 in 2024.

That's a 99.5% drop in twelve months. (Source: Stats NZ / ACC, fatal-serious-injuries)

Numbers don't just collapse like that without reason. Something fundamental changed between those two years, and it wasn't that New Zealand workplaces suddenly became 200 times safer overnight.

The most likely explanation: a change in how these injuries are counted or categorised. Perhaps what was once classified as a fatal injury is now recorded differently. Perhaps the dataset was restructured. Perhaps the definition narrowed to focus on workplace-specific deaths rather than all ACC-covered fatalities.

Whatever the technical reason, the contrast matters because it shows how invisible methodology changes can be. If you looked only at 2024's figure of 312 fatal injuries, you might think that's normal. You might compare it to 2023's 315 and conclude things are stable, maybe even improving slightly.

But zoom out two more years and the entire picture changes. You're not looking at a stable trend. You're looking at a discontinuity so severe it suggests you're measuring two completely different things.

This is the danger of data without context. Three hundred and twelve deaths is still three hundred and twelve families who lost someone. That number deserves scrutiny, not just acceptance. But scrutiny requires knowing what you're actually counting.

The 2020 figure sits at 59,148, similar to 2021. Then the floor drops out. That two-year window is the hinge. Whatever happened between 2021 and 2022 redefined what this dataset measures.

For anyone trying to understand workplace safety in New Zealand, this matters enormously. You can't track progress if the baseline keeps shifting. You can't identify emerging risks if the counting method changes halfway through. You can't hold employers or regulators accountable if nobody knows what the numbers actually represent.

The dataset goes back to 2000, covering 24 years of history. That long view should tell a story about how New Zealand workplaces have become safer or more dangerous over time. Instead, it tells a story about how data definitions can obscure more than they reveal.

Three hundred and twelve people still died. That's the number that matters most. But understanding what it means, and whether it's comparable to previous years, matters too. Right now, we're working blind.