Corrective Action
From Defect to Verified Fix: The CAPA Workflow
Corrective action is fundamentally different from a repair or a rework disposition. Reworking a single out-of-tolerance WR-28 waveguide section restores that one part to specification, but it does nothing to explain why the part drifted out of tolerance or to stop the next part from failing the same way. Corrective action targets the process, not the part. It is triggered when a nonconformance is systemic, recurring, escaped to a customer, or involves a critical characteristic, and it does not close until the responsible team has shown, with data, that the failure mode is gone.
The most common framework in aerospace and defense RF shops is the eight-discipline (8D) method, which maps cleanly onto the CAPA records required by AS9100. The early disciplines contain the problem so no further suspect product ships; the middle disciplines drive to the true root cause using tools such as the five-whys technique, fishbone (Ishikawa) diagrams, and fault-tree analysis; the later disciplines implement the permanent corrective action, verify its effectiveness, and standardize the change across every affected process and product line. Each step generates objective evidence that becomes part of the auditable quality record.
Root cause is the discipline that separates a genuine corrective action from a band-aid. If a batch of frequency-converter housings fails a plating-adhesion pull test, replacing the parts is not corrective action; identifying that the cleaning bath pH had drifted out of range, restoring it, and adding monitoring to keep it in range is. The distinction matters because regulators and customers audit corrective actions specifically to confirm that root cause, not symptom, was addressed.
Effectiveness Verification and Recurrence
The hardest part of any corrective action is proving it worked. A completed action is not an effective action. Verification requires production data gathered after the change: a run of consecutive conforming lots, a process capability index that has recovered above 1.33, a measurable reduction in defects per million opportunities, and zero recurrence of the original failure code over a defined monitoring window. The sample size for verification is chosen so the probability of missing a recurring defect is acceptably small, which is why low-defect-rate processes demand thousands of conforming units before the CAPA can be signed off and closed.
Governing Equations and Metrics
Cpk = min[ (USL − μ) / (3σ) , (μ − LSL) / (3σ) ] ≥ 1.33
Defect Rate (Poisson yield model):
Y = e−D×A (first-pass yield), DPMO = (defects / opportunities) × 106
Verification Sample Size (zero-failure, rule of three):
n ≈ 3 / p (95% confidence the true defect rate is below p)
Where USL/LSL = upper/lower spec limits, μ = process mean, σ = process standard deviation, D = defect density, A = area or count of opportunities (so D×A is the expected defects per unit), p = target defect proportion. Example: to confirm a recurring defect rate has fallen below p = 0.001, the rule of three requires n ≈ 3000 conforming units with zero failures.
Disposition vs. Corrective vs. Preventive Action
| Activity | Trigger | Scope | Addresses Root Cause? | Closure Evidence | Typical Record |
|---|---|---|---|---|---|
| Rework / Repair | Single nonconforming unit | The affected part only | No | Part re-passes inspection | NCR disposition |
| Containment (interim) | Suspect lot identified | All at-risk inventory | No (temporary) | Suspect stock quarantined | 8D D3 |
| Corrective Action | Systemic / recurring defect | The process & all affected lines | Yes (eliminates cause) | Cpk ≥ 1.33, 3 to 5 clean lots | CAPA / 8D D5 to D7 |
| Preventive Action | Trend or risk, no defect yet | Potential failure modes | Yes (before occurrence) | Risk reduced, monitoring added | CAPA / FMEA update |
Frequently Asked Questions
What is the difference between corrective action and preventive action?
Corrective action responds to a nonconformance that has already occurred by eliminating its root cause; preventive action is proactive, addressing a potential failure mode found through trend analysis or risk assessment before any defect appears. The two are paired in the CAPA system. Example: replacing a contaminated gold-plating bath after WR-15 flanges fail an adhesion test is corrective, while adding daily bath-chemistry monitoring and an SPC chart to catch drift early is preventive.
How do you verify that a corrective action was effective?
Effectiveness verification needs objective evidence gathered after the fix is in place, not just proof the task was done. Typical evidence is 3 to 5 consecutive conforming lots or 30 to 90 days of production, a recovered Cpk above 1.33, a measurable drop in DPMO, and zero recurrence of the original failure mode. For a target defect rate near 0.1%, the zero-failure rule of three calls for about 3000 units to pass with zero failures to show the rate has fallen at 95% confidence. The CAPA stays open until that evidence is reviewed and signed.
When does a nonconformance require a formal corrective action versus just a repair?
A single unit that reworks to specification is handled through the nonconformance and material-review process, which fixes the part but not the cause. A formal corrective action is triggered when the defect is systemic or recurring, escaped to a customer, involves a critical characteristic, or dominates scrap cost in a Pareto analysis. Many systems set a numeric trigger, such as the same defect code appearing 3 or more times in a rolling 90-day window or a first-pass yield drop below an action limit, to decide when rework alone is insufficient.