Key takeaways
  • Build a factual timeline before analysing anything. Analysis contaminated by an early theory finds evidence for that theory.
  • 'Human error' is a starting point, not a conclusion. The useful question is why that action made sense to that person at that time.
  • Investigate near misses with the same method. They contain the same causes at no cost.
The finding that stops investigations
Operator error

Once an investigation concludes someone made a mistake, it stops. The conditions that made the mistake likely go unexamined, remain in place, and produce the next incident with a different person.

Immediately after

  • Make the area safe and care for anyone involved. Everything else waits.
  • Preserve the scene and the physical evidence, including settings, positions and the state of guards.
  • Photograph before anything is moved or reset.
  • Capture accounts early and separately, while memory is fresh and before people converge on a shared version.
  • Record equipment data, shift patterns, and what else was happening at the time.

Build the timeline first

Sequence what happened, with times, from well before the incident. The period before the event is where the conditions accumulated: a changed schedule, an absent colleague, a machine that had been playing up, a rushed changeover. An investigation that starts at the moment of injury sees only the final step.

Then analyse conditions, not just actions

  • What was the state of the equipment, including guards, tools and lighting?
  • What was the workload and time pressure?
  • Was the procedure available, accurate and followable as written?
  • Had this deviation happened before without consequence, making it normal?
  • What information did the person have at the moment they acted?
  • What would a competent, well-intentioned person have done in the same conditions?

Interviewing

Interview to understand, not to establish fault. Open questions, one person at a time, and an explicit statement that the purpose is prevention. If people believe the outcome is discipline, you will get an account designed to be safe rather than accurate, and the investigation is then working from fiction.

Corrective actions that hold

Test each proposed action against the hierarchy of controls. If every action is retraining and a toolbox talk, the investigation has concluded that the person was the cause. Ask what physical or organisational change would make the incident impossible or obvious, and record why higher controls were rejected if they were.

Near misses

A near miss has the same causes and no injury. Investigating them is the cheapest safety learning available, and the volume of near misses reported is a direct measure of whether people trust the process. A sudden drop in near-miss reports is bad news, not good news.