Most corrective actions do not fail at execution. They fail at the point they are written — because what is written down is not a corrective action at all.
The single most common entry in the root cause field is some version of "operator error — staff retrained." It appears across sites, sectors and audit schemes. It closes quickly, it satisfies the paperwork, and it has close to zero probability of preventing recurrence. Twelve months later the same nonconformity is raised, a different operator is retrained, and the cycle repeats.
Writing a corrective action that genuinely closes is a discipline with a defined structure. Here it is.
Start by separating correction from corrective action
These are different things, and conflating them is where most closures go wrong.
Correction eliminates the detected nonconformity. The product is placed on hold. The record is completed. The equipment is cleaned. It deals with what has already happened.
Corrective action eliminates the cause of the nonconformity so it does not happen again. It deals with what would otherwise happen next.
Almost every prematurely closed CAR is a correction recorded in the corrective action field. Cleaning the equipment is a correction. Changing the cleaning frequency because the validated interval was wrong is a corrective action.
Both are necessary. Only one of them prevents recurrence.
Write the nonconformity so it can be acted on
A corrective action can only be as good as the finding that triggered it. A well-written nonconformity contains four elements:
- The requirement — the specific clause, procedure or specification that was not met
- The observation — what was actually seen, where, and when
- The objective evidence — the record, measurement, photograph or statement supporting it
- The significance — why it matters in food safety terms
Compare these two:
"Cleaning records not adequate."
"The documented cleaning procedure CP-04 requires post-production cleaning of Filler 2 to be verified by ATP swab at three defined points, with results recorded. ATP verification records for Filler 2 were absent for 14 of 30 production days reviewed in August 2026 (records CL-08-01 to CL-08-30). Cleaning sign-off was completed on all 30 days. This means cleaning effectiveness was not verified on 47% of production days for equipment in direct contact with ready-to-eat product."
The first produces a defensive conversation and a vague action. The second produces a specific investigation, because the facts are not in dispute and the significance is stated.
Do the root cause analysis properly
Root cause analysis is where the discipline lives. Two techniques cover most situations:
Five Whys — ask why the nonconformity occurred, then why that occurred, iterating until you reach a cause that is within the system's control to fix. The discipline is in not stopping at the first answer involving a person.
Worked example, using the ATP finding above:
- Why were ATP swabs not performed on 14 days? Because the swabs were not available on those days.
- Why were swabs not available? Because stock ran out and was not reordered.
- Why was stock not reordered? Because there is no minimum stock level or reorder trigger for consumables used in verification activities.
- Why is there no reorder trigger? Because verification consumables are not included in the critical spares and consumables list.
- Why not? Because the critical consumables list was built around production materials and never reviewed against the food safety verification program.
The root cause is a procurement system that does not recognise food safety verification consumables as critical. Retraining the cleaner would have achieved nothing — the cleaner did not have swabs.
Fishbone analysis is more appropriate where multiple contributing factors are likely, examining method, machine, material, manpower, measurement and environment as categories.
Two tests for whether you have reached a real root cause:
- The control test — is this cause something the organisation can change through its system? If the answer is "people need to be more careful," you have not reached it.
- The prevention test — if this cause were eliminated, would the nonconformity be prevented? If it would only be made less likely, keep going.
Write the action against the cause
Once the cause is identified, the corrective action should follow almost mechanically. Against the example above:
- Add food safety verification consumables to the critical consumables list, with defined minimum stock levels and an automatic reorder trigger
- Amend CP-04 so that unavailability of verification materials is itself a recordable nonconformance requiring escalation, rather than a silent omission
- Review the verification program for any other activity dependent on a consumable with no stock control
Each of these addresses the system. None of them depends on someone remembering.
The action must also carry:
- A responsible owner — a named individual, not a department
- A due date — realistic and specific
- The verification criteria — defined now, not at closure
That last element is the one most often omitted, and it is the one that makes closure objective rather than negotiable.
Define effectiveness verification before you close
Effectiveness verification asks a different question from implementation verification.
Implementation verification asks: was the action done? Is the consumables list updated? Is the procedure amended?
Effectiveness verification asks: did it work? Has the nonconformity stopped occurring under real operating conditions, over a period long enough to prove it?
For the example above, effectiveness verification might be: review ATP verification records for three consecutive months following implementation; effectiveness is demonstrated if verification completion is 100% across all production days, with any gap traceable to a recorded escalation.
That is objective. Either the records show it or they do not.
The interval matters. Verifying effectiveness one week after implementation, while attention is still high, proves very little. The period must be long enough to cover normal operating variation — shift changes, staff absence, peak production, the conditions under which the original failure occurred.
The failure modes to watch for
- Root cause restates the nonconformity. "Root cause: swabs were not performed." That is the finding, not its cause.
- Retraining as the sole action. Valid only where the genuine cause is a competence gap — and even then, it should be paired with a system change that makes the correct action easier than the incorrect one.
- Closure on submission. The CAR is closed when the response is received rather than when the evidence is verified.
- No effectiveness check. Implementation confirmed, effectiveness never tested.
- Scope too narrow. The action fixes the one instance found and does not consider whether the same cause affects other lines, shifts or products.
That last point deserves emphasis. If a systemic cause was found on Filler 2, the corrective action must address whether the same weakness exists on Fillers 1 and 3. An auditor samples; the site is responsible for the population.
Practical takeaways
- Use separate fields for correction and corrective action. If your CAR form has one field, it will be filled with the correction.
- Ban "operator error" as a terminal root cause. Treat it as a prompt to ask two more whys.
- Define verification criteria at the point of raising, not at the point of closing.
- Set the effectiveness review interval to cover real operating variation — typically one to three months, not one week.
- Ask the extension question every time: where else could this same cause be operating?
- Track recurrence as a metric. A corrective action process that produces repeat nonconformities is itself a nonconformity.
A corrective action is closed when there is evidence the cause is gone and the control is holding. Until then, it is open — regardless of what the due date says.