Heinrich's safety triangle: is it still relevant in 2026?

If you've spent any time in workplace safety, you've seen the triangle. It's probably on a poster somewhere in your office. It might be in your induction slides. It's almost certainly come up in a toolbox talk at some point.

Heinrich's safety triangle (also called Bird's triangle, the accident pyramid, or the safety pyramid) is one of the most recognisable concepts in occupational health and safety. It's been shaping how we think about risk for over 90 years.

But is it still useful? Or has it become an oversimplified myth that's doing more harm than good?

The honest answer (as with most things in safety) is: it's complicated.

What the triangle actually says

Let's start with what Heinrich actually proposed. In 1931, based on a review of more than 75,000 accident reports from insurance company files, Herbert William Heinrich suggested a ratio of 1 major injury to 29 minor injuries to 300 no-injury incidents. His conclusion: by reducing the number of minor incidents, you'd see a corresponding reduction in serious ones.

Frank Bird expanded the model in 1969, analysing 1.75 million accident reports from 297 companies. Bird's updated ratio was 1 serious injury to 10 minor injuries to 30 property damage events to 600 near misses. A 2003 ConocoPhillips Marine study pushed it further, finding approximately 300,000 at-risk behaviours for every fatality.

The core idea is simple and intuitive: serious injuries don't happen in isolation. They sit at the top of a much larger base of less severe events and unsafe conditions. Address the base, and you reduce the peak.

That core idea has been described as a cornerstone of 20th-century workplace safety philosophy. And for good reason. It shifted the entire profession from reactive (wait for the serious injury, then investigate) to proactive (look for the near misses and fix the underlying causes before someone gets hurt).

What the critics get right

The triangle has always had critics, but the criticism has sharpened significantly in recent years. Here's what they argue, and they make some genuinely strong points.

The ratios aren't universal

Heinrich's original data cannot be verified. His files have been lost, so nobody can review his methodology. Bird's data, while more substantial, was based on reports submitted by organisations (not independent field studies), and a 1991 study of confined space incidents found a ratio of just 1.2 minor injuries per serious injury or death, which is nothing like the 29:1 or 600:1 that the triangle suggests.

The ratios clearly vary by industry, by activity, and by how incidents are reported. Using a fixed ratio as a predictive tool is, as Manuele argued in his 2002 book Heinrich Revisited: Truisms or Myths, misleading.

Reducing minor incidents doesn't necessarily prevent fatalities

This is the most damaging criticism. Manuele's analysis concluded that fatality and serious injury events often occur without prior evidence from the analysis of less severe incidents. In other words, the causes of fatal accidents and the causes of paper cuts may be completely different.

The Campbell Institute has been particularly influential in advancing this argument through the concept of SIF (serious injury and fatality) potential. Their position: not all near misses are equal. A dropped pen and a forklift-pedestrian close call both sit at the base of the triangle, but only one has the potential to kill someone. Treating them as equivalent, or assuming that reducing one reduces the other, is a fundamental error.

As TapRooT's Mark Paradies puts it: "Stopping paper cuts won't prevent major process safety-related accidents or industrial safety fatalities."

It over-blames the worker

Heinrich's original work suggested that 88% of accidents were caused by "unsafe acts of persons." Modern safety science has moved well beyond this view. W. Edwards Deming, whose work on quality management transformed manufacturing, argued that it was poor management systems, not individual behaviour, that caused most accidents. Manuele agreed, noting that Heinrich placed an inordinate emphasis on individual failure and gave insufficient attention to systemic causes.

When you combine a model that blames the worker with a metric system that rewards low injury counts (like TRIR-based incentive programmes), you get workplaces where people stop reporting incidents rather than workplaces that are actually safer. That's a real and documented harm.

What the triangle still gets right

So should we throw it out entirely? I don't think so. And here's why.

The underlying principle is sound

Even the harshest critics of Heinrich's specific ratios generally agree on the underlying principle: there are always more low-severity events than high-severity events, and low-severity events contain valuable safety information. Safeopedia's balanced assessment makes this point well: although Heinrich's proportionality doesn't exist as a universal law, there are many opportunities to learn from non-accidents that we miss when we limit ourselves to learning only from serious injuries.

It makes the invisible visible

The triangle's greatest contribution was making people aware that the events below the injury line (near misses, unsafe conditions, at-risk behaviours) exist, matter, and deserve attention. Before Heinrich, most organisations only paid attention when someone got hurt. The triangle argued that by the time someone gets hurt, you've already missed hundreds of warning signs.

That insight is as true today as it was in 1931. The challenge has always been capturing those warning signs at sufficient volume and quality to act on them.

It gives safety a language

For better or worse, the triangle gave the profession a shared mental model. When a safety manager says "we need to focus on the base of the triangle," everyone in the room understands what that means. It's an imperfect model, but it's a model that communicates a proactive mindset, and that matters.

What the triangle looks like in 2026

Here's where I think the conversation gets genuinely interesting. The triangle's biggest limitation was always practical, not conceptual. The principle (learn from near misses to prevent serious injuries) is sound. The problem was that organisations had no reliable way to capture near misses at scale.

Manual reporting captures a fraction of actual events (we've covered why that gap is so large in detail elsewhere). When the base of your triangle is built on 10 to 15 manually reported near misses per month, the data is too sparse to reveal patterns, too inconsistent to compare across shifts or zones, and too unreliable to use as the foundation for any meaningful analysis.

Computer vision AI changes this. When a system is continuously detecting safety events across your highest-risk areas, you're not building the base of the triangle from voluntary reports. You're building it from comprehensive, timestamped, location-specific data that captures events whether anyone reports them or not.

And here's the critical evolution: with this volume of data, you can do what the Campbell Institute has been advocating for years. You can separate the events with SIF potential (a forklift-pedestrian near miss at speed) from the events without it (a minor housekeeping issue). You can focus on the portion of the triangle that actually predicts serious harm, rather than treating all base-level events as equivalent.

This is the modern version of the triangle: not a fixed ratio, but a data-driven system that captures leading indicators at scale, filters them by severity and SIF potential, and feeds them into coaching workflows that address the specific risks that matter most.

At inviol, we see this playing out every day. Customers don't use our heatmaps and trend reports to count every minor event equally. They use them to identify where the highest-severity events concentrate, which shifts and zones need attention, and whether coaching interventions are reducing the events most likely to cause serious harm.

The compliance angle

Under the Health and Safety at Work Act 2015, New Zealand PCBUs are required to proactively identify and manage risks. WorkSafe NZ expects organisations to look beyond injury data and understand the broader risk landscape. The ASSP's 2025 trends analysis reflects the same shift at an international level: the profession is moving from TRIR toward SIF and PSIF metrics, recognising that total recordable injury rates don't adequately capture the risk of catastrophic events.

Whether you call it "the base of the triangle" or "leading indicators with SIF potential," the regulatory and professional direction is clear: organisations need visibility into the events below the injury line, and they need to act on what they find.

So, is it still relevant?

Heinrich's specific ratios? No. They were never verified, they vary by industry, and using them as a predictive model is unreliable.

Heinrich's underlying principle that serious injuries are preceded by far more numerous lower-severity events, and that capturing and learning from those events is the most effective way to prevent harm? Absolutely yes. That principle is more actionable in 2026 than it has ever been, because the technology to capture those events at scale finally exists.

The triangle isn't wrong. It's just incomplete. And the missing piece was always the data.

If you'd like to see what a data-driven version of the safety triangle looks like for your operation, book a demo with inviol and we'll show you what the base of your triangle actually contains.

Frequently Asked Questions

What is Heinrich's safety triangle?

Heinrich's safety triangle (also called the accident pyramid or Bird's triangle) is a model proposed by Herbert William Heinrich in 1931, showing a ratio of 1 major injury to 29 minor injuries to 300 no-injury incidents. Frank Bird expanded it in 1969 to 1:10:30:600. The core idea is that serious injuries are preceded by a much larger number of less severe events, and addressing those events can prevent serious harm.

Is the safety triangle still accurate?

Heinrich's specific ratios have been widely challenged. His original data cannot be verified, and research shows the ratios vary significantly by industry. Manuele (2002) argued that the causes of fatal accidents are often different from the causes of minor incidents. However, the underlying principle that there are more low-severity events than high-severity events, and that they contain valuable safety information, remains widely accepted.

What is SIF potential and how does it relate to the triangle?

SIF (serious injury and fatality) potential is a concept developed by the Campbell Institute that addresses a key weakness of the triangle. It argues that not all near misses are equal: only those with the potential to cause serious harm should be prioritised. This means focusing on the portion of the triangle's base that actually predicts catastrophic events, rather than treating all low-severity events as equivalent.

How does AI change the way we use the safety triangle?

Computer vision AI solves the triangle's biggest practical limitation: capturing the base-level events at scale. Manual reporting systems typically capture a small fraction of near misses. AI detection captures events continuously and automatically, providing the data volume needed to identify patterns, filter by severity and SIF potential, and target coaching at the risks most likely to cause serious harm.

What is the ASSP's position on the safety triangle in 2025?

The ASSP's 2025 trends analysis reflects a significant shift from total recordable injury rates (TRIR) toward SIF and PSIF (potential serious injury and fatality) metrics. This aligns with the modern view that the triangle's principle is sound but its application needs to focus on the events with serious harm potential, not all events equally.