Problem solving and 8D

Two things make an 8D work or fail, and neither is the format. The first is defining the problem so precisely that you could tell whether it had stopped. The second is refusing to close the report until the action has been shown to work — not shown to have been done.

A problem solving meeting at a shop floor board, with an 8D report displayed on a screen, a fishbone diagram and a five why chain drawn on a whiteboard, three rejected parts laid out on a table with measurement records beside them, and a red quarantine bin of contained stock in the background. — The rejected parts on the table matter more than anything on the screen. Analysis without the parts in the room drifts into speculation.

The eight disciplines

The two arrows that matter are the one back from D6 when verification fails, and the bracket showing how long containment has to live.

	Discipline	The trap
D1	Form the team	No authority to change anything
D2	Describe the problem	Rushed, and not quantified
D3	Interim containment	Mistaken for the fix
D4	Root cause	One cause instead of two
D5	Select corrective actions	Actions not tied to the confirmed cause
D6	Implement and verify effectiveness	Closed on implementation
D7	Prevent recurrence	Not fed back into the FMEA
D8	Recognise and capture learning	Skipped entirely

D2: quantify, or investigate the wrong thing

“Noise complaint” gives no basis for ever knowing whether it stopped. State the part, characteristic, measured deviation, quantity, dates and conditions — and a surprising amount of the time the act of quantifying it eliminates most of the candidate causes before anyone opens a toolbox.

D3: containment is not corrective action

D4: two root causes, not one

A diagram showing the two root cause branches for a single defect reaching a customer. A defective part is drawn at the centre. One branch traces back through the process to the occurrence root cause, asking why the defect was made, illustrated by a tool wear condition on a forming operation. The second branch traces forward through the inspection and control points to the escape root cause, asking why the defect was not detected, illustrated by a control plan row whose check frequency was once per shift so an entire shift of parts passed between checks. Beneath each branch the corresponding corrective action is shown, a tool change interval for the occurrence cause and a revised check frequency with an error proofing device for the escape cause. A note states that fixing only the occurrence cause leaves the detection system equally blind to the next different defect. — Two failures, two branches, two corrective actions. Fixing only the left branch leaves the right branch just as blind next time.

A defect reaching a customer is two failures:

Occurrence cause — why was it made?
Escape cause — why did the controls not catch it?

The escape cause almost always lands on a control plan entry that was inadequate, absent, or simply not being performed. Fix only the occurrence and the detection system stays exactly as blind for the next, different defect.

D6: implemented is not verified

	Proves
Implemented	Something was done
Verified effective	It worked

Verification means data from the running process over a meaningful period, under the conditions that produced the problem. Not a statement that the change was made.

And if verification fails, the flow goes back to D4 — because the cause was wrong.

D7: apply it everywhere it applies

The same tool, material, method or design feature is shared across several parts. A cause found on one is present on the others, so fixing only the part that generated the complaint guarantees a reappearance under a different part number in about eight months.

D7 also closes the loop backwards. A confirmed failure mode the PFMEA did not anticipate — or ranked too low — is direct evidence that the analysis was wrong, and it must be fed back. An organisation whose FMEAs never change after launch is not learning from its own failures.

D8: lessons that stay in the report are not lessons

Push the learning into work instructions, FMEA libraries, design guidelines, checking fixture standards and a lessons learned record that the next programme actually reads. Skip it and the same problem gets solved independently by three departments over five years.

What to take away

Quantify the problem — part, characteristic, deviation, quantity, dates, conditions.
Containment is not corrective action, and it covers transit, customer, stock and WIP.
Containment stays until D6 verification, not until the report is typed.
Find two root causes: why it was made, and why it escaped.
The escape cause usually lands on a control plan entry.
“Operator error” is not a root cause — ask what allowed it and let it through.
Verify each why against evidence; ideally reproduce the defect.
Implemented ≠ verified effective. Close on data, over time, under the real conditions.
Failed verification sends you back to D4.
Apply the action to every part sharing the cause.
Feed the confirmed failure mode back into the FMEA.
A recurrence means the analysis was wrong. Reopen it; do not add inspection.