CAPA management: running the system rather than closing the actions

Deciding what earns a CAPA and what is handled at the line

If every nonconformance opens a formal action, the system fills with trivia and the serious items lose their place in the queue. If almost nothing opens one, the same defect recurs for years under a stack of scrap tickets. The triage rule needs to be written and applied by a named person: severity of the effect, whether the issue has appeared before, whether it reached a customer, and whether the existing controls should have caught it but did not. Everything below that line gets fixed and recorded at the point of occurrence, with the recurrence data reviewed periodically so that a pattern of small events can promote itself into a formal investigation.

The backlog tells you more than the closure rate does

A closure rate can be manufactured in an afternoon. Ageing cannot. Sort open items by the date they were raised and look at the tail: actions that have been open for many months are usually ones where the owner has no authority to make the change, where the fix needs capital nobody has approved, or where the real cause was never found and the item is being nursed. Look also at ownership concentration. When a handful of engineers hold most of the open population, the system has quietly become their personal to-do list, and the plant has stopped treating problem resolution as line management work.

Effectiveness verification is the step that quietly disappears

Implementing an action and eliminating a defect are different claims, and only the second one matters. Before closing, someone has to state what will be measured, over what volume of production, and what result would count as failure — then wait for that evidence and record it. The common shortcut is to close on the day the new work instruction is issued. The consequence appears months later as a repeat complaint against an issue the records show as resolved, which costs far more credibility with a customer than the original defect did, because it demonstrates the system does not verify itself.

Preventive items never win an argument against reactive ones

In a combined system the corrective side is fed by angry customers and stopped lines, while the preventive side is fed by audit findings, near misses, read-across from a problem found on another product, and trends that have not yet broken through a limit. Given one engineer and both queues, the reactive work wins every week. The only defence is to ring-fence a portion of capacity, review the preventive queue separately, and require that a significant corrective investigation asks explicitly where else the same failure mode could exist. Without that read-across question, each plant relearns the same lesson product by product.

Governance: who reviews it, how often, and with what authority

A review that can only nag is theatre. The forum needs the standing to reassign an owner, release engineering time, escalate an item that has stalled for want of a decision, grant a documented extension with a reason attached, and reject a closure whose evidence is thin. Attendance matters as much as agenda: without production and engineering leadership present, the meeting becomes a quality department briefing itself. Record the decisions, not just the status, because the audit question a year later is rarely whether an item was open, but who agreed it could stay open and on what grounds.

Frequently asked questions

How long should a corrective action realistically stay open?

Long enough to gather effectiveness evidence, and no longer. Containment should be immediate, the investigation short, the permanent change bounded by whatever engineering or tooling work it requires, and the verification window set by production volume rather than by the calendar. What matters more than any target duration is that every extension carries a written reason and a named approver, so the ageing report distinguishes items genuinely waiting on evidence from items nobody is working.

Who should own a CAPA item — quality or the process owner?

The process owner, with quality facilitating and verifying. When the quality department owns the actions, it ends up negotiating with production for time to fix production's problems, and closure depends on goodwill. When the area manager owns the item, the fix competes for resource inside their own plan and the quality engineer can stay independent enough to refuse a weak closure. Ownership should be a named individual, never a department.

What evidence should a customer expect before we close an action raised against their complaint?

A description of what was contained and where, a cause supported by physical or process data rather than assertion, the specific change made to the process, control plan or instruction, and production evidence from after that change showing the defect absent across a meaningful quantity. Many customers also ask what else was checked for the same failure mode. Sending a closure without the verification data usually gets the action reopened and raises the scrutiny applied to the next one.