Corrective action and preventive action: a loop that actually closes

Corrective and preventive action, defined

Corrective action is the work done to remove the cause of a non-conformance that has already occurred, so that it does not recur. Preventive action is the work done to remove a potential cause before a non-conformance occurs. The two share a method but start from different evidence.

In everyday supplier quality, most requests are corrective. A customer finds a wrong dimension, a mixed lot or a failed plating test, and asks the supplier what happened and what will change. The preventive side is the same supplier looking at near misses, trends and process data without waiting for a complaint.

The distinction matters because a corrective action has a real failure to investigate, with parts and records. A preventive action often starts from a trend or a risk assessment, which makes it easier to drift into general statements about training and awareness.

A single event can need both. A cracked bend may require a corrective action on the tool and a preventive action on every other part that runs on the same tool. Treating them as one ticket often means the wider risk is never assessed.

Start with containment, not cause

The first job is to protect the customer. Containment means identifying the suspect stock, quarantining it, checking what has already shipped and telling the customer what to do with the parts they hold. It should happen within hours or days, not after the investigation.

Containment is temporary. It does not answer why the failure happened, and it should not be presented as the fix. A supplier that responds to a complaint with a sorting report and no cause analysis has contained the problem, not corrected it.

Record the containment scope precisely. Which lots, which dates, which serial numbers, which warehouse. A vague containment statement forces the customer to sort everything, which is the outcome the supplier was trying to avoid.

Find a cause you can act on

A useful root cause is specific enough to change something. "Operator error" is not a cause, because it does not say which decision, which instruction or which condition made the error likely. "The bend gauge was set from the previous revision because the work instruction still showed the old dimension" is a cause you can fix.

Common methods are the five whys, a fishbone diagram and a fault tree. The method matters less than the discipline of testing each explanation against evidence. Look at the parts, the machine settings, the material certificate and the inspection records. If the explanation does not fit the physical evidence, it is a guess.

Ask why the failure escaped as well as why it occurred. A defect that reaches the customer has at least two causes: the one that made it, and the one that let it pass. Both belong in the CAPA, because fixing only the first leaves the detection gap open.

Write an action plan with an owner and a date

Each action should name what will change, who owns it, when it will be done and how it will be verified. Actions that change a document, a tool, a setting, a supplier or a test are stronger than actions that ask people to be more careful.

Sort the actions by whether they remove the cause or only reduce the chance. A poka-yoke fixture that prevents the wrong orientation is stronger than a note in the work instruction. A drawing change that removes an ambiguous tolerance is stronger than a reminder at the morning meeting.

Set dates that reflect the risk. Containment is immediate. A permanent tool change may take weeks, and that is acceptable if the interim control is defined. What is not acceptable is an open action with no date, because it will still be open at the next audit.

Keep the plan short enough to manage. Three to six actions with clear owners are easier to close than twenty vague ones. If the list is long, separate the cause-removing actions from the supporting ones and track them at different levels.

Verify effectiveness with data

A CAPA is not closed when the actions are done. It is closed when the evidence shows the failure has not returned. Choose the evidence before you close, so the test is not shaped to fit the result.

For a dimensional failure, that may be a run of consecutive lots with the characteristic in tolerance and a capability figure above the agreed target. For a mixed-lot failure, it may be a period with no mix incidents and an audit of the label and changeover process. For a plating failure, it may be thickness results across several racks and a bath analysis.

The verification period should be long enough to cover the conditions that caused the failure. If the fault appeared on a night shift or with a particular material lot, the evidence must include those conditions. A week of good parts on day shift proves very little.

The stages and the evidence for each

The table sets out a CAPA route and the record that should exist at each stage. It works for a supplier response and for an internal investigation.

Close the loop across similar parts

A single part fix leaves the same risk on every similar part. Before closing, ask where else the cause could exist. The same tool, the same operator, the same plater or the same drawing note may appear on other part numbers.

Update the control plan, the FAI record and the risk file where the change affects them. If the cause was an ambiguous drawing, revise the drawing and tell the customer. If the cause was a supplier process, update the supplier scorecard and the audit plan.

Share the lesson in a form the next project can use. A short note in the design guide or the supplier onboarding pack is worth more than a closed ticket that nobody reads. That is how a corrective action becomes a preventive one.

Why CAPA stalls

The most common failure is a cause that describes a person rather than a system. The second is an action that cannot be verified, such as retraining the team. The third is a closure date driven by an audit rather than by evidence.

Another pattern is the repeat finding. The same non-conformance returns a year later because the action fixed the symptom and left the process unchanged. A good review looks for repeat findings across part numbers and suppliers, not just within one ticket.

Keep the CAPA file small enough to read. A long narrative with no data hides the logic. A short record with the failure, the cause, the change and the evidence is easier to defend and easier to reuse.

Assign one owner for the whole loop, not one per department. A single owner can see the containment, the cause, the action and the evidence together, which is what stops a ticket from being closed before the process has changed.