Fix the Risk

Here's an uncomfortable question for every complaint: was this failure on your PFMEA? For the scratched housings, the answer was no. Not because the team missed it, but because packing wasn't on the PFMEA at all.

The process FMEA is the plant's list of what can go wrong at each step, how bad it would be, and what stops it. A complaint is a test of that list. Every escape tells you one of two things: the failure mode was there and the controls didn't work, or the failure mode was never there.

In 8D terms, this chapter and the next three are D5 through D7: choosing the corrective action, putting it in place, and locking it in so it can't happen again.

Find where the PFMEA went quiet

Walk the process in order, from receiving to the truck, and hold the PFMEA next to it. In this plant, the PFMEA ended at final inspection. Packing, labeling and staging for shipment weren't listed. As far as the PFMEA knew, the part stopped existing once it passed inspection.

That's common, and it's exactly where complaints hide. Steps after the last inspection can damage a good part, and nothing downstream will catch it.

Add the step, then rate it honestly

The team added a packing step with a failure mode of parts contacting each other in the tote. They rated it the way it really ran, not the way they wished it ran:

1. Severity 5. A cosmetic defect the customer rejects at incoming inspection.
2. Occurrence 6. It was already happening with the new totes.
3. Detection 8. Nothing after final inspection could catch it.

You don't multiply those three numbers. The current AIAG and VDA method uses an Action Priority table instead. You look up severity, occurrence and detection together, and it returns High, Medium or Low. Severity 5, occurrence 6 and detection 8 come back Medium.

Why not multiply? The old risk priority number could give the same score to very different situations. A rare failure with a safety effect could score the same as a minor defect that happens every day. Action Priority looks at severity first, then occurrence, then detection, so a serious failure can't hide behind a low multiplied score.

Here is what the three levels ask of the team. High means act, or write down why the current controls are enough. Medium means you should act or justify. Low means you may. Medium is not a pass. This team acted because a customer was already holding 14 bad housings, and an occurrence rating of 6 says it keeps happening.

After the fixes in the next two chapters, divider totes and a check at packout, the team expected occurrence to drop to 2 and detection to 4. That lookup returns Low. Those new ratings only count once the controls are in place and working.

Some customers and older procedures still ask for RPN. Use whatever your customer requires. If you aren't sure, ask which one they expect.

Ask what else is missing

Don't stop at the one line that failed. If packing was missing, what else was? Labeling, staging and returnable dunnage all sat in the same blind spot. A reverse PFMEA, where you stand at each station and compare what's happening with what's written, finds these gaps faster than a meeting does.

The mistake to avoid

Changing the ratings without changing anything on the floor. If occurrence drops from 6 to 2 on paper but the totes are the same, the PFMEA is now wrong in a more dangerous way. Ratings follow actions, never the other way around.

The opposite mistake is reading a Medium as "no action needed." The table tells you where to look first. It doesn't tell you what the customer will accept.

The plant, the part and the complaint in this book (RMA-031) are a made-up example. The tool sidebars show ShopFloorOps tools, which I make and sell. Every chapter teaches the step itself, so you can use your own forms.