In 2021, New Zealand recorded 61,194 fatal injuries. The next year? 357. By 2024, it was 312.

That's not progress. That's a definitional earthquake.

The numbers tell you immediately: something changed in how Stats NZ and ACC count workplace deaths between 2021 and 2022. The pre-2022 figures likely captured something broader. Claims lodged. Historical cases revisited. Administrative recategorisations. The methodology shifted, and with it, our understanding of what a "fatal injury" actually means in official data.

But here's what the contrast reveals: 312 deaths is still 312 families. When the counting method changed, the human toll didn't. What changed was visibility.

The 2024 figure of 312 fatal injuries represents the current measurement baseline. It's holding roughly steady with 2023 (315) and 2022 (357). That consistency matters. It means we now have three years of comparable data to track whether workplaces are getting safer or more dangerous. (Source: Stats NZ / ACC, fatal-serious-injuries)

Before 2022, the numbers were so inflated by definitional scope that real trends were invisible. If fatal injuries rose by 5%, was that more deaths or more paperwork? If they fell by 10%, were fewer people dying or were claims being processed differently?

You couldn't know. The data was counting something, but not necessarily the thing you thought it was counting.

Now we can know. The 312 deaths in 2024 tell us where the baseline sits. They give us a number to measure against. When next year's figure comes out, we'll be able to say with confidence: are New Zealand workplaces becoming safer?

The pre-2022 numbers weren't lies. They were answering a different question. Possibly a broader one about injury-related mortality captured in ACC's administrative systems over time. Possibly including historical claims finalised in a given year. The exact reason for the break matters less than recognising it happened.

What matters now: 312 people died in workplace-related incidents last year. That number has stayed within a narrow band for three years. It hasn't spiked. It hasn't plummeted. It's stable, which means it's measurable, which means we can finally start asking the right questions.

Is 312 acceptable? What would it take to get it to 250? To 200? Which industries account for the majority? Which regions? What interventions actually work?

Those questions were impossible to answer when the numbers moved from 59,148 to 61,194. They're possible now. The contrast between the old methodology and the new one doesn't just show a data revision. It shows the moment we started being able to see clearly.