What metrics demonstrate CAPA effectiveness for AS9100 auditors?

AS9100 auditors usually expect CAPA effectiveness to be demonstrated with objective evidence that the root cause was addressed and the problem did not recur under defined conditions. The strongest metrics are not generic CAPA counts. They are measures tied to the specific nonconformance, failure mode, product family, process step, supplier, or control that the corrective action was intended to fix.

A dashboard can support the discussion, but it does not prove effectiveness by itself. Auditors will typically look for the link between the original issue, root cause analysis, action plan, implementation evidence, effectiveness check, and follow-up results. The metric must be traceable to records, not just reported as a summary number.

Metrics that commonly support CAPA effectiveness

The most useful metrics are those that show whether the same or related problem came back after the corrective action was implemented. Common examples include:

• Recurrence rate: repeat nonconformances for the same cause, part family, operation, work center, supplier, or requirement after CAPA implementation.
• Repeat escape rate: customer escapes, internal escapes, or downstream escapes linked to the same failure mode or missed control.
• Effectiveness verification pass rate: percentage of CAPAs that pass the planned effectiveness check without reopening or additional action.
• Audit finding recurrence: repeated internal, customer, or third-party audit findings against the same process weakness or requirement.
• NCR trend after implementation: change in relevant nonconformance volume, defect code frequency, rejection rate, or rework for the affected process.
• Process performance measures: first-pass yield, scrap, rework, inspection acceptance, process capability, or test failure rates where those measures are directly related to the root cause.
• Containment effectiveness: number of additional suspect parts found after containment, missed containment events, or later escapes from the contained population.
• Action aging and overdue rate: open, overdue, or repeatedly extended CAPAs, especially for high-risk or customer-impacting issues.
• Supplier corrective action recurrence: repeat supplier nonconformances or escapes after supplier CAPA closure.

These metrics are more persuasive when they include a baseline, an implementation date, a defined observation period, and a clear population. For example, “no repeat defect code X on line Y for three lots or 90 days after fixture revision and operator retraining” is usually stronger than “CAPA closed.”

What auditors usually challenge

Auditors often challenge metrics that measure activity rather than effectiveness. CAPA closure rate, average days to close, and training completion can be useful management indicators, but they do not prove that a root cause was eliminated or controlled. They show that work was completed, not that it worked.

Training records are a common example. If the corrective action was retraining, an auditor may ask why training was an adequate corrective action, how competence was verified, whether the work instruction or process control changed, and whether the error recurred. Attendance alone is weak evidence.

Another weak pattern is using broad site-level quality metrics to justify a specific CAPA. A lower overall scrap rate does not necessarily show that one machining setup issue, documentation error, or supplier process weakness was corrected. The metric needs to be close enough to the cause to be meaningful.

Evidence matters as much as the metric

For AS9100 audit purposes, the metric should be supported by controlled records. Depending on the process, those records may come from the QMS, MES, ERP, PLM, inspection system, maintenance system, supplier portal, or customer quality portal.

In brownfield environments, this is rarely seamless. CAPA data may live in a QMS, defect data in MES, disposition records in an NCR workflow, revisions in PLM, and shipment or return data in ERP. If those systems are not well integrated, the organization may need documented manual reconciliation, defined data owners, and review evidence. Auditors may accept manual controls when they are controlled, repeatable, and traceable, but uncontrolled spreadsheets and undocumented data manipulation create obvious risk.

If electronic reports are used as audit evidence, the organization should be able to explain data source, report logic, access control, change control, and audit trail expectations. In regulated aerospace environments, replacing legacy systems just to improve CAPA reporting is often unrealistic because of validation cost, qualification burden, downtime risk, integration complexity, and long equipment lifecycles. A more practical approach is usually to strengthen traceability and controls around the existing systems.

Typical failure modes

• Closing CAPAs based on action completion rather than verified effectiveness.
• Using a verification window that is too short for the production rate or failure mode.
• Grouping unrelated defects under broad cause codes that hide recurrence.
• Failing to connect customer escapes, internal NCRs, supplier issues, and audit findings to the same root cause pattern.
• Changing procedures without verifying implementation at the point of use.
• Relying on dashboards that cannot be traced back to controlled records.
• Extending due dates repeatedly without risk-based justification or management review.

Practical expectation

A credible CAPA effectiveness package usually includes the original nonconformance, root cause analysis, implemented actions, changed documents or controls, training or qualification records if applicable, objective post-action data, and a documented effectiveness decision. The exact metric depends on the risk, process, production volume, customer requirements, and how quickly recurrence would realistically appear.

No metric guarantees an AS9100 audit outcome. The defensible position is to show that the organization selected relevant measures, controlled the evidence, allowed enough time or production exposure to test the fix, and reacted when the data showed the action was not effective.