Get the free 8D/D4 checklist: https://8d-gate.vercel.app/?utm_source=devto&utm_medium=article&utm_campaign=capa-symptom-fixes_checklist#email-capture
There is a set of corrective actions that appears in a huge number of 8D reports:
- Operator retrained.
- 100% inspection added.
- Stock sorted.
- Reminder posted at the station.
- Supplier informed.
None of these is wrong. Several are exactly what you should do in D3 containment. The problem is when they show up in D5/D6 as the permanent corrective action. Then the 8D treats the symptom, the cause stays in place, and the complaint comes back - sometimes on the same part, sometimes on the next one.
This post is about telling the two apart and writing a D5-D7 that actually removes the cause.
Containment vs. correction vs. prevention
It helps to be strict about the three layers:
| Layer | Purpose | Typical 8D step | Example |
|---|---|---|---|
| Containment | Protect the customer now | D3 | Sort stock, add temporary 100% check |
| Corrective action | Remove the verified root cause | D5/D6 | Change fixture so part cannot be loaded wrong |
| Prevention | Stop it recurring elsewhere | D7 | Update PFMEA, apply to similar fixtures/lines |
A symptom-only CAPA is one where the D5/D6 entries are really still containment. A quick test: if you removed the action tomorrow, would the defect come back? If yes, it is containment.
Why "retrain" is almost never a permanent fix
Retraining assumes the cause was missing knowledge. Sometimes it was. But if the operator already knew the right method and the process still allowed the mistake, retraining does not change the process. It relies on people being perfect on every cycle, every shift, forever.
Stronger alternatives, roughly from strongest to weakest:
- Eliminate the possibility (design change, fixture that only accepts the right orientation).
- Detect automatically before the part moves on (sensor, interlock, in-line gauge with stop).
- Make the error obvious (visual standards, colour coding, go/no-go gauges).
- Procedure / training - only when the above are not feasible, and with a stated reason.
If your action is at level 4, a reviewer will often ask why levels 1-3 were not possible. Having an answer ready helps.
Why "100% inspection" is containment wearing a costume
Adding inspection catches the defect; it does not stop it being made. It also has known limits - human visual inspection in particular is not fully reliable over long runs. It may be a reasonable escape corrective action (if your D4 found the detection gap), but it should be paired with an occurrence action, and you should state when or whether it will be removed.
Every action needs to map to a cause
A simple discipline that fixes a lot: in D5, write each action next to the D4 cause it addresses.
- Occurrence cause -> occurrence action
- Escape cause -> escape action
If there is an action with no cause, why are you doing it? If there is a cause with no action, the 8D is incomplete. Reviewers notice both.
Verification (D6) is not "implemented"
"Implemented on [date]" proves the action exists, not that it works. Effectiveness evidence looks more like:
- a run of production with the characteristic monitored and no recurrence, with the quantity and period stated,
- a capability or measurement result after the change,
- a trial where the old failure condition was deliberately introduced and the new control caught or prevented it.
Do not overstate. If you only have a short run so far, say that, and give the date when you will review again.
D7: where symptom fixes usually go to die
Even a good root-cause fix stays local unless D7 spreads it. Questions worth answering in D7:
- Which similar parts, fixtures, lines or sites have the same weakness?
- Has the PFMEA been updated (failure mode, cause, controls, ratings)?
- Has the control plan been updated to reflect the new or changed check?
- Is there a lessons-learned entry somebody will actually see on the next launch?
A 2-minute self-check for D5-D7
- Remove each action mentally. Does the defect come back? -> it is containment.
- Does every action point to a verified D4 cause (occurrence or escape)?
- Is at least one action above "training" on the hierarchy, or is there a reason why not?
- Is effectiveness shown with data, not just a completion date?
- Are PFMEA/control plan updates and horizontal deployment named?
Let a second reader check it
I built 8D Gate because the second reader is often missing at small suppliers - the same person writes and approves the report. Paste an 8D and it returns Accept / Conditional / Reject and flags things like containment listed as corrective action, actions with no linked cause, and missing effectiveness evidence. It is free and it will not fill in evidence you did not give it.
Free check: https://8d-gate.vercel.app/review?utm_source=devto&utm_medium=article&utm_campaign=capa-symptom-fixes
Advisory only. 8D Gate gives a pre-send readiness opinion (Accept / Conditional / Reject). It does not replace your customer's SQE, your own quality sign-off, or any IATF/OEM requirement. It will not invent evidence; if something is missing, it says so.
Question for practitioners: what is the most creative symptom-only "permanent" action you have seen in an 8D? Anonymised, of course.
Top comments (0)