What is a CAPA effectiveness check?

A planned, dated review of objective evidence that the corrective action eliminated the problem and did not create a new one. The pass criterion (a metric, a threshold, a monitoring window, and a data source) should be defined when the CAPA opens, and the window should be long enough for the defect to have had a fair chance to recur.

How long should you wait before checking CAPA effectiveness?

Long enough for the defect to have realistically recurred. Scale the window to the old defect rate: a few weeks for a daily defect, 8-12 weeks for a weekly one, and three to six months for a monthly or quarterly one. For rare defects, watch a faster proxy such as a process parameter staying in range on a control chart.

What is the difference between verification and validation of a CAPA?

Verification confirms the action was carried out and removed the cause; validation confirms a changed process reliably produces conforming output across its operating range. FDA uses the phrase 'verify or validate' because a documentation change is verified, while a process change that could affect the product may need validating before you trust it at volume.

What counts as objective evidence for CAPA effectiveness?

Data a stranger could read and reach the same conclusion: defect and scrap trends across the window, process readings showing the driving parameter back in range, complaint codes going quiet, or inspection results at the escape point. FDA does not accept verbal confirmation that the problem 'went away.'

What happens if a CAPA fails its effectiveness check?

You reopen it. A failed check means the root cause analysis missed something and stopped at a symptom, so the defect returned by the same path. Go back to the cause with what you learned, push the action from changing people toward changing the process, and keep a containment action in place until the new fix is verified.

Why do most CAPAs close without being verified?

Because it is easy to close on implementation evidence ('operator retrained, signature on file') and hard to close on outcome evidence that cannot exist until the monitoring window passes. The CAPA ages, turns red on the metrics board, and gets closed to clear the board. Defining the effectiveness criterion on day zero fixes the incentive.

How to Verify CAPA Effectiveness

A CAPA effectiveness check is a planned, dated review of objective evidence, gathered far enough after implementation for the problem to fairly recur, that confirms the corrective action removed the root cause and created no new problem. It is the step that separates "we closed the CAPA" from "we fixed it."

Most corrective and preventive action systems do the hard parts and skip the last one. The team finds the root cause, changes the process, updates the work instruction, and closes the record the same week. Six weeks later the same defect is back, and nobody notices because the CAPA already reads "closed, effective." Effectiveness verification is the discipline that stops that loop. It is also the part auditors read most closely, because a plant that verifies its fixes is a plant whose whole quality system can be trusted.

Why do most CAPAs close without ever being verified?

Because closing on implementation evidence is easy and closing on outcome evidence is not. "Operator retrained on 12 March, signature on file" is a real record you can attach today. "Zero recurrences over 60 days" cannot exist until 60 days have passed, so the CAPA sits open, ages, turns red on the metrics board, and someone closes it to make the board green. The incentive points the wrong way.

The fix is to design the check in at the start, not bolt it on at the end. When you open the CAPA you already know what "fixed" will look like in data. Writing that down on day zero turns the day-90 review into a five-minute reading of a chart instead of an argument about whether the problem is really gone. It also forces an honest root cause: if you cannot describe the evidence that would prove the fix worked, you probably have not understood the failure yet.

What is the difference between verifying and validating a corrective action?

Verification confirms the action was carried out and removed the cause; validation, where it applies, confirms the changed process reliably produces conforming output over its full operating range. FDA's device rule uses the phrase "verify or validate" for exactly this reason: a paperwork change (revising a work instruction) is verified, while a process change that could affect the product (a new fixture, a new set point) may need validating before you trust it at volume.

For most plant-floor CAPAs, effectiveness verification is what you owe: evidence the defect stopped and no new defect started. Reserve validation for the cases where the action itself changes how product is made and a bad outcome would be expensive to discover late. Either way the logic of the check is the same, measure the thing that would move if the fix worked, over a window long enough to believe the result.

A criterion needs all four parts. Miss one and the check becomes an opinion.

How do you write a pass criterion you can actually check?

A checkable criterion has four parts: a metric, a threshold, a monitoring window, and a named data source. Drop any one and the check turns into an opinion. "The problem is resolved" has none of them. "Zero recurrences of defect code 14 on line 3 across 60 days of production, read from the station quality log" has all four, and any auditor or any night-shift lead can settle it without a meeting.

Two habits keep criteria honest:

Measure the defect, not the activity. Retraining, new signage, and a revised setup sheet are things you did. They belong in the implementation record. The criterion has to measure what the customer would feel: scrap rate, escape rate, a dimension staying in tolerance, a complaint code going quiet.
Write it before you act. A criterion set after you already know the outcome is not a test, it is a rationalization. Lock it when the CAPA opens, next to the root cause and the planned action. Our guide to root cause analysis covers how to get to a cause specific enough that a clean criterion is even possible.

When should the effectiveness check happen?

Long enough after the fix that the problem had a fair chance to come back. The window is set by how often the defect used to occur, not by the calendar or the CAPA due date. A defect that showed up daily can be judged in a few weeks of clean data. A defect that showed up quarterly needs six months, or a faster proxy signal you can read sooner.

The proxy is the trick that keeps rare-defect CAPAs from staying open forever. If a burr appeared once a quarter and the cause was tool wear, you do not have to wait two quarters to believe the fix. Watch tool-wear measurements or the affected dimension on a control chart: if the parameter that drove the defect is back in its expected range and staying there, that is early evidence the fix is holding, even before enough time passes to count zero recurrences.

A fix for a weekly defect that stays clean for three days proves nothing. Scale the window to frequency.

The 7-step effectiveness-verification process

Run every effectiveness check the same way so the result means the same thing each time.

Define the pass criterion at open. Metric, threshold, window, and data source, written into the CAPA the day it starts, not the day you check.
Confirm the action is actually in place. Before you judge outcomes, verify the change was implemented as planned: the fixture installed, the work instruction revised and trained, the parameter changed and locked. A criterion cannot be met by a fix that never fully happened.
Set the review date and owner. Put the check on the calendar with a named owner the moment the window is defined, so it is not left to whoever notices the CAPA is old.
Collect the evidence over the window. Pull the defect data, scrap reasons, complaint codes, or process readings for the full monitoring period. Partial windows do not count.
Check for side effects. Confirm the fix did not create a new problem elsewhere: a slower line, a new defect code, a bottleneck at the added inspection. Both good sense and ISO 13485 require verifying the action did not introduce new issues.
Compare to the criterion and decide. Met the threshold across the full window with no new problem? Effective. Anything short of that is not effective, whatever the effort behind it.
Close and standardize, or reopen. If effective, close with the evidence attached and extend the fix to sister lines and similar products. If not, reopen to root cause, the first analysis missed something.

Verification is a gate, not a formality. The only two exits are close-and-standardize or reopen.

What counts as objective evidence?

Data a stranger could read and reach the same conclusion. FDA is explicit that verbal confirmation the problem "went away" is not evidence; specific metrics and dated records are. In practice, objective evidence for an effectiveness check is one of a short list:

Defect and scrap trends by code, line, and shift across the window, ideally on a Pareto chart or run chart so recurrence is visible at a glance.
Process data showing the parameter that drove the defect staying inside its expected range, the proxy signal for slower defects.
Complaint and return codes going quiet for the affected failure mode over the monitoring period.
Audit or inspection results at the point where the defect used to escape, showing it no longer does.

The common failure here is structural: the evidence lives on paper. When nonconformances, holds, and scrap reasons are on clipboards in a binder, nobody can pull 60 days of defect-code-14 history on line 3 without an afternoon of counting, so the check gets waved through. Trend visibility is the raw material of an honest effectiveness check, which is why it depends on the same defect-tracking data a good QMS already captures.

The numbers behind the scrutiny

FDA's CAPA subsystem guidance for investigators directs them to confirm that "corrective and preventive actions for product and quality problems were implemented and documented" and that their effectiveness was verified or validated (FDA, Corrective and Preventive Actions (CAPA) subsystem).
21 CFR Part 820 requires manufacturers to verify or validate corrective and preventive action to ensure it is effective and does not adversely affect the finished device; since 2 February 2026 the requirement flows through FDA's Quality Management System Regulation incorporating ISO 13485:2016 (FDA QMSR).
ASQ reports quality costs commonly run 15–20% of sales revenue, the recurring-defect money a verified CAPA exists to stop (ASQ, Cost of Quality). See our breakdown of the cost of quality.

How do you handle a CAPA that fails its effectiveness check?

You reopen it, and you treat the failure as information, not embarrassment. A failed check means the root cause analysis missed something, usually it stopped at a symptom. The action removed a contributing factor but not the real driver, so the defect came back by the same path it always used.

Go back to the cause with what you now know: the fix you tried and the fact that it did not hold are both clues. If the first pass used 5 whys and landed on "operator error," the reopened investigation should ask why the error was possible and design that possibility out. Actions that change the process outlast actions that change people, and a CAPA that failed once is telling you the first action was too weak. This is the same containment-then-fix discipline that runs through 8D and CAPA: protect the customer with a containment action while the reopened analysis runs, so the reopening does not put anyone at risk.

Get effectiveness verification right and CAPA stops being a folder of optimistic closures. It becomes the plant's proof that serious problems are investigated once, fixed once, and stay fixed, which is exactly the trend visibility Harmony's quality and downtime intelligence is built to give a floor team, without ripping out the systems you already run. See how one plant put that to work in our CLS case study.