In the root cause meeting somebody says, "I think it was the material," with the confidence of a person who has seen it before. Nobody asks for the lot certificate. The minutes close with that cause, and three weeks later the defect comes back with a different material.
Root cause analysis is the process of identifying, with evidence rather than opinion, the cause that, if removed, keeps a problem from coming back. It is the discipline of requiring a data point before accepting any cause, regardless of which technique produced it.
The cause that wins the argument
Almost every room has someone with more years, more rank, or just a louder voice. When that person proposes a cause, everyone else nods, because disagreeing costs more than the argument is worth. The cause that ends up written down is usually the one nobody felt free to question.
I have told this story with the fishbone diagram, when the team fills it in without verifying, and with the 5 whys, when the chain stops at the most comfortable answer. Both symptoms come from the same place. Nobody demanded evidence before accepting.
The question that changes the conversation
One question changes the dynamic of the room: what data supports this? It is the question that turns a round of opinions into a list of hypotheses that still need to be verified. The person who proposed the material cause no longer has to defend it with authority. They have to defend it with the certificate.
That question works because it forces the group to separate what is believed from what can be proven, before anybody commits to one explanation. Without it, the meeting argues over who has more experience. With it, the meeting argues over what needs to be measured.
Before you conclude, define what evidence you accept
I use three types of evidence, and nothing else counts as proof of a cause. Direct measurement: someone goes to the floor and measures the condition in the moment, not from memory. A dated historical record: a data point that already existed before the cause was proposed, not one generated afterward to support it. And a repeatable test: trying to reproduce the defect under the suspected condition and confirming that it actually shows up.
This same discipline is what holds up the AIAG MSA criteria for deciding whether a measurement system is trustworthy: under 10% the system is accepted, between 10% and 30% it is conditionally accepted, above 30% it is not accepted. If you do not trust the instrument that produced the data, you do not trust the data, and the entire chain of evidence collapses from there.
Where the fishbone diagram and the 5 whys fit in
The fishbone organizes candidates by category, but it does not test them. The 5 whys drive down to the origin, but only on the branch that already survived verification. The evidence discipline is what decides, before the first why gets chained, which branch deserves that attention and which one gets ruled out without spending time on it.
When I work with a team, every branch on the diagram ends up marked as proven, ruled out, or waiting on data. That last column becomes next week's measurement list, and nobody chains whys onto a branch that is still sitting there, without data, waiting.
Three ways a cause gets accepted without evidence
- Closing by consensus: if everyone agrees it feels proven, and nobody asked for the data behind it.
- Using the same data point to support two different hypotheses, when that data point actually only rules out one of them.
- Concluding without having tried to reproduce the defect under the suspected condition.
Start here
On your next fishbone diagram, before the meeting, write next to each branch what data would prove it or rule it out. If you cannot write that data point down, that branch goes back to investigation before anyone discusses it.
This is what I run with quality, process and floor teams inside problem solving work. If you want to apply it to a real case, tell me what problem you are working on.
What counts as evidence in root cause analysis?
Three things: a direct measurement taken on the floor, a dated historical record that existed before the hypothesis, and a repeatable test that reproduces the defect under the suspected condition. The opinion of the most experienced person in the room counts as none of the three.
How do I stop the loudest voice from winning the cause?
With one question before accepting any cause: what data supports this. That question forces the group to separate what is believed from what can be proven, regardless of who proposed it.
Is root cause analysis a tool or a process?
It is a process. The fishbone diagram and the 5 whys are tools inside it. The discipline of requiring evidence before accepting a cause is what holds both of them up.