Root cause analysis, without blaming the operator

The operator-blame trap

Here is how most defect investigations end. A batch comes back with chipped edges. Someone asks what happened, an operator gets named, everyone agrees to "be more careful," and the report is closed. Three weeks later the exact same chip is back, because nothing about the process changed. Being more careful is not a fix, it is a wish.

Real root cause analysis starts from a different assumption: good people produce defects when the process lets them. The operator who chipped the edge was working with the same fixture, the same speed, the same handling method as everyone else. If the chip is repeatable, the cause is in the system, not the person. Blame ends the investigation early and guarantees the defect comes back. Curiosity keeps it open long enough to actually find the cause.

5 Whys, done properly

The 5 Whys is exactly what it sounds like: you ask why, and you keep asking it, following the chain from the symptom down to something you can actually change. Five is a rough number, not a rule. Sometimes it is three, sometimes seven. You stop when the next why stops being useful. A worked example from a glass shop:

1. Why did the customer reject the batch? Several lites had chipped corners.
2. Why were the corners chipped? They were knocking against the rack during transport from cutting to tempering.
3. Why were they knocking? The interleaving between lites had been skipped, so glass touched glass.
4. Why was it skipped? The interleaving powder ran out mid-shift and nobody flagged it.
5. Why did nobody flag it? There is no reorder trigger for consumables, so it is only noticed once it is already gone.

Notice where that landed. The chip looked like a handling problem at step two, which is where a lazy investigation stops and tells the operators to be careful. Five questions in, the real cause is a missing reorder system for consumables, which is fixable and which was quietly causing other problems too. Fix the trigger and you fix every defect that missing powder would ever have caused, not just this batch.

Two cautions. Keep the chain honest, do not steer it toward the answer you already wanted. And a single line of whys assumes one cause, which is fine for a simple problem. When a defect has several tangled causes at once, reach for the fishbone.

The fishbone, for tangled problems

The fishbone diagram (also called Ishikawa, after the man who popularized it, or a cause-and-effect diagram) is for when "why" has more than one answer at each step. You draw the defect as the head of a fish, then draw bones for the major categories of cause and brainstorm possibilities onto each. The classic categories are the 6 Ms, and they map neatly onto a glass line:

The value of the fishbone is that it forces you to consider causes you would otherwise skip. A team convinced the chip is a handling problem gets made to at least look at the fixture, the incoming glass, and the measurement method before settling. Once the likely bones are on the board, you test them, and you use the 5 Whys to drive down whichever one holds up.

Contain, correct, prevent

These three get muddled constantly, and keeping them straight is most of what separates a shop that improves from one that firefights forever:

• Containment is the immediate stop-the-bleeding action: quarantine the suspect batch, check what already shipped, protect the customer today. It buys time. It fixes nothing.
• Correction is fixing the defective product itself: rework or remake the chipped lites. Necessary, but it only deals with the units in front of you.
• Corrective action (prevention) is changing the process so the defect cannot recur: the reorder trigger, the fixture redesign, the recipe change. This is the only one that makes the problem go away for good, and it is the one shops skip because the fire is already out.

If every investigation ends at containment and correction, you will be scrapping the same defect next year. The corrective action is the point of the whole exercise.

Proving the fix held

An investigation is not finished when you make the change. It is finished when you have evidence the change worked. That means deciding, before you close it, how you will know: the chip rate on that line should drop to near zero over the next month, and you go back and check the number rather than assuming. Sometimes the fix does not hold, or it moves the problem somewhere else, and only the follow-up measurement tells you that. Write down what you changed, why, and what the result was, so the next person who sees this defect starts from your answer instead of from scratch.