A CAPA closes when it can show three things: what the problem was, why it happened, and proof it has stopped. Write a problem statement with numbers, contain the affected product, find and verify the root cause, assign each action an owner and a date, and decide the effectiveness check before you implement anything. Most CAPAs stall because the problem is vague, the cause is a guess, or nobody defined what “effective” means. The steps below fix each of those, and the Example shows a completed CAPA with its closure math.
By Coplain Team
Last updated
Step by step
Write a measurable problem statement — state what is wrong, where, when, how many and against which requirement. “Bad welds” cannot close. “14 of 320 brackets from weld cell 2 failed the pull test (requirement ≥ 4.5 kN) between 3 and 7 March” can, because you can measure whether it stops.
Contain and correct — protect the customer first. Quarantine suspect stock, sort or test it, and record the disposition on an NCR. Note the date containment started and the evidence that it works. Containment is not the corrective action; it buys time to find the cause.
Find the root cause and prove it — use a method that fits the problem: 5 Whys for a single chain, a fishbone diagram when causes branch. Then verify it: can you turn the problem on and off by changing that cause? Also ask why the problem escaped detection, which is a second root cause.
Plan actions with owners and dates — write one action per line, each aimed at a verified cause, with a named owner and a due date. Prefer actions that change the process (error-proofing, fixture, program, parameter limits) over reminders and retraining. Include document updates: work instruction, PFMEA, control plan.
Define the effectiveness check up front — before implementing, write down the measure, the target and the window. For example: zero pull-test failures across the next 30 lots, or a defect rate below a set level for three months. Base the window on how often the problem used to occur.
Implement and verify each action — attach evidence to each action as it is completed: the new revision number, the photo of the fixture, the training record, the program change. Verification asks “was it done?”. It is not yet effectiveness.
Check effectiveness and close — when the window ends, compare the data with the target. If it passes, close the CAPA and look sideways for similar processes that need the same fix. If it fails, reopen the root cause step rather than adding more actions. Log the closure in the corrective action log.
Why CAPAs stay open
Open CAPAs are one of the most common audit findings. The usual reasons are predictable: the problem statement is too vague to measure, the root cause is a symptom, the actions depend on people remembering, or the effectiveness check was never defined, so nobody knows when to close it. Each step above closes one of those gaps. The root cause analysis and CAPA guide, the pillar for this cluster, covers root cause methods in depth, and the CAPA process guide covers running the CAPA system as a whole.
The problem is stated with data, and the scope of affected product is clear.
Containment happened promptly and covered stock at every location, including in transit and at the customer.
The root cause is supported by evidence, not opinion.
Actions address the cause and the escape, and the documents they changed are at the new revision on the floor.
Effectiveness was measured against a target set in advance.
Trend data (a Pareto chart of causes, for example) shows the CAPA system is learning, not repeating.
Writing the record so a stranger can follow it
A CAPA is read months later by people who were not there: an auditor, a customer, a new quality engineer. Write each field so it stands alone. Name parts, lots, machines and documents by number and revision. Attach the evidence instead of describing it: the data that sized the problem, the test that verified the cause, the photos of the change, and the effectiveness data. Date each entry, so the timeline from detection to containment to closure is visible without reconstruction. Plain sentences beat jargon; if an action says “improve process”, it is not finished being written.
Worked examples
Example: A completed CAPA for fictional weld pull-test failures
Problem: 14 of 320 brackets from weld cell 2 failed the pull test (≥ 4.5 kN) from 3 to 7 March. Failure rate: 14 ÷ 320 = 4.4%.
Containment: 1,150 brackets in stock and in transit were held; extra pull tests on each held lot found 9 more failures, and the 2 lots they came from were scrapped.
Root cause (occurrence): the wire feed roller was worn, so feed speed dropped about 15% under load. Verified by measuring feed rate: 7.6 m/min against a set point of 9.0 m/min, (9.0 − 7.6) ÷ 9.0 = 15.6% low; with a new roller, 40 test welds all passed.
Root cause (escape): pull tests were sampled 1 per 100 parts, so a drifting process could run for hours before detection.
Actions: replace roller (maintenance, 10 March); add feed-rate check to the daily start-up sheet (quality, 14 March); add roller to the PM schedule every 500 hours (maintenance, 21 March); increase pull test to 1 per 25 until effectiveness is shown (quality, 10 March).
Effectiveness check: target 0 pull-test failures in the next 30 lots. Result: 30 lots, 9,600 parts, 384 tests (9,600 ÷ 25), 0 failures.
Result: Example result: the CAPA closes 41 days after opening with zero failures in 384 tests, and sampling returns to 1 per 100. A preventive action extends the feed-rate check to weld cells 1 and 3. All numbers are example values.
Common mistakes
Vague problem statements. Include what, where, when, how many and the requirement. If you cannot measure it stopping, rewrite it.
Root cause is “operator error”. Ask why the process allowed the error, and change that.
Retraining as the only action. Retraining fades. Pair it with a process change, an error-proofing device or a clearer instruction.
Ignoring the escape cause. Ask why detection failed and fix the inspection or sampling too.
Closing on completion of actions. Close on evidence of effectiveness, against a target set before implementation.
Adding actions when the effectiveness check fails. Go back to root cause. A failed check usually means the cause was wrong.
Frequently asked questions
How long should a CAPA take to close?
There is no single rule. Your procedure should set target times for containment, root cause and implementation, with the effectiveness window based on how often the problem occurred. Track ageing so overdue CAPAs get attention.
What is a CAPA effectiveness check?
A measurement, defined in advance, that shows whether the problem has stopped recurring after the actions were implemented. It is different from verifying that the actions were done.
Does every nonconformance need a CAPA?
No. Use criteria in your procedure, such as severity, recurrence, customer impact and risk, to decide which NCRs escalate to a CAPA.
What is the difference between a CAPA and an 8D?
A CAPA is the corrective action process in your quality system. An 8D is one structured format for running it, common for customer complaints. [8D vs A3](/blog/8d-vs-a3) compares formats.
Who should own a CAPA?
One person who can make the changes happen, usually from the area where the problem occurred, supported by quality. Each action can have its own owner.
Sources
ISO 9001:2015 Quality management systems — Requirements (clause 10.2), ISO
ISO 13485:2016 Medical devices — Quality management systems (clauses 8.5.2 and 8.5.3), ISO
Effective Problem Solving Practitioners Guide (CQI-20), AIAG