Incident investigation: finding the root cause with the 5 whys method

When do you investigate an incident?

Not every incident needs a full investigation. Look at two things: how serious was the outcome, and how serious could it have been? A near miss where a heavy load just missed someone deserves more attention than a graze. Repetition is also a reason to dig deeper.

Have the investigation led by someone who knows the work but is not directly responsible for it, and involve someone from the shop floor. For a minor incident, a short conversation and a few lines on paper are enough. For a serious accident you want a full file.

Management systems such as ISO 45001 expect you to investigate incidents, establish causes and follow up on measures. You decide the depth of the investigation yourself, based on the risk.

An incident investigation in six steps

A fixed order stops you from jumping to an explanation too soon. The investigation need not be heavy: for a minor incident the whole route fits on a single sheet of paper.

• Make the area safe and take the immediate measures that are needed
• Record what you see: photos, location, equipment, weather conditions and who was there
• Talk to those involved and to witnesses separately, using open questions
• Put events on a timeline, from shortly before the incident until it was dealt with
• Look for causes, for example with the 5 whys method
• Choose measures, assign owners and check later that they work

How do you talk to those involved?

A conversation after an incident is not an interrogation. Choose a quiet place, ask the person to tell the story in their own words and do not interrupt. Only then ask deeper questions: what did you see, what did you do, what did you expect to happen and what was different from normal?

Talk to people separately, so that accounts do not colour each other, and have the person show on site how the task normally goes. It often turns out that the agreed method and actual practice differ, and that is exactly what you want to know.

Direct cause versus root cause

The direct cause is what directly caused the incident: a badly placed pallet, a slipping ladder, a missing guard. The root cause is the reason that situation could arise and was not stopped. Only there can you change something so it does not come back.

A simple test: if you remove this cause, can the same incident still happen? If so, you are still at a symptom. Note that there is rarely just one root cause. Usually a few circumstances come together, and then you address all of them.

How the 5 whys method works

You start with a factual description of the problem and then, for each answer, ask again why that was so. Every answer must be a fact you can verify, not an assumption. If you cannot back up an answer, go back to the witnesses or the site.

Five is a rule of thumb. Sometimes you reach something you can change after three rounds, sometimes you need seven. Stop when you arrive at a cause the organization controls. If the chain branches, follow each branch. Do it together with people who do the work, for example on a whiteboard.

Worked example: a pallet falls from the racking

In a warehouse a pallet of material falls from the racking and lands half a metre from a colleague. Nobody is hurt. The investigation goes like this:

• Why did the pallet fall? It was sitting crooked on the beams.
• Why was it crooked? The driver placed it in a hurry, because the truck with the next delivery was already waiting.
• Why was that truck waiting? Several suppliers arrived at the same time, because they decide for themselves when to come.
• Why do they decide that themselves? There is no agreement on delivery times and no loading schedule.
• Why is there no schedule? Nobody owns the incoming goods flow.

From root cause to measure

In the example the direct cause is the crooked pallet, and the root cause is that the incoming goods flow has no owner, which creates time pressure. Speaking to the driver solves nothing: the next driver faces the same choice. A measure at the level of the root cause is an owner for incoming goods and fixed time slots for deliveries.

Also choose a second layer, because one measure is seldom enough: a check on placement at high racking, or fall protection at the rear. Record how you will know it works, for example by looking after a few months whether reports are still coming in with time pressure as the cause.

Pitfalls: the blame question and other shortcuts

The biggest pitfall is the blame question. If those involved feel accused, they say less and will not report next time. Deliberately ignoring rules does happen and calls for a separate conversation, but even then you ask why it was possible: was the rule unworkable, or was nobody looking?

• Stopping at human error or inattention: that is the start of the question, not the answer
• Asking "who" questions instead of asking how it could happen
• Writing down assumptions as if they were facts
• Looking for only one cause when several come together
• Questioning only the person directly involved and skipping the planner and supervisor
• Choosing measures that consist mainly of an instruction or a talking-to

How BrydgeOne supports this

In Incidents & reports you record, investigate and close out accidents, near misses and hazardous situations. You determine the risk with Kinney & Wiruth, set out measures and the reporter automatically receives status updates.

The measures go into Actions with an owner and deadline, and you close them with an explanation and evidence. You assess structural risks further in Risks & opportunities, and you share the lesson from the investigation through a toolbox with attendance recorded.