The method is one question asked repeatedly. Something went wrong; ask why; ask why of the answer; continue until you reach something that can actually be changed.

Sakichi Toyoda's observation was that people stop at the first answer, which is almost always a symptom, and that five iterations usually gets past it. Five is a rule of thumb rather than a count — sometimes it is three and sometimes eight.

What it looks like when it works

The client deliverable was two days late.

Why? It went to review on the last day. Why? The draft was finished the night before. Why? The work started a week later than planned. Why? The kick-off waited on a contract signature. Why? The contract is drafted after the project is scheduled rather than before.

The first answer suggests working faster. The last one is a sequencing decision somebody can change this week, and it will fix every future instance rather than this one.

The failure that makes it useless

Stopping at a person.

Why was it late? Because Dave forgot. True, probably. It ends the inquiry at the point where the only available remedy is that Dave should remember better, which is not a remedy. Everybody forgets things; the interesting question is why this particular forgetting turned into a late deliverable rather than being caught.

Push through it. Why did Dave forgetting cause a failure? Because nothing else checks. Why does nothing else check? Because the step was added without a handover. That is a fact about the process, and it is fixable in a way that Dave's memory is not.

This is Hanlon's razor applied to operations: assume the arrangement, not the person, and you will be right more often and you will learn more when you are wrong.

The other three failure modes

One chain, taken as the answer. Most failures have several contributing causes and five whys follows one thread. Running it two or three times from the same symptom, deliberately taking a different branch, produces a much more honest picture.

Hindsight doing the reasoning. Each answer is constructed after the outcome is known, which makes the chain feel inevitable. It was not, and the test is whether the proposed cause would have predicted the failure in advance. Frequently it would not.

Landing on something unchangeable. A chain terminating in "because the market is competitive" has gone one step too far. Stop at the last link somebody in the room can act on.

Where it belongs

It is a method for recurring problems, not for one-off surprises, and not for anything that needs a decision today. Its output is a change to how work is arranged, which is why it sits naturally next to kaizen — the whys find the thing, the small change fixes it, and the next occurrence tests whether the fix held.

One caution before acting on the answer. A step that looks pointless in a whys chain may be there for a reason nobody present remembers, and removing it can produce the failure it was quietly preventing. See Chesterton's fence: find out why it was put there before taking it away.