Corrective vs Preventive Action: A Practical CAPA Guide
Correction, corrective action and preventive action explained with examples, the CAPA process step by step, how to write a CAPA that gets closed properly, how to verify effectiveness, and the mistakes that keep CAPAs open.
Find the root causes in your own feedback, free
Share an export of your tickets, reviews or survey comments. We send back the top root causes within 48 hours on business days. No sales call.
Quick answer
Corrective action removes the cause of a problem that has already happened, so it does not happen again. Preventive action removes the cause of a problem that has not happened yet, but could. A correction only fixes the immediate result, such as refunding one customer. Together, corrective and preventive action are called CAPA: identify the problem, contain it, find the root cause, act, verify that it worked and standardise the fix.
Every company fixes problems. Fewer fix them for good. The difference is usually not effort but structure: whether someone asks why the problem happened, removes that cause, checks a few weeks later that it is really gone, and makes sure the same weakness is not waiting somewhere else.
That structure is what CAPA, short for corrective and preventive action, describes. This guide explains the three terms people mix up most (correction, corrective action and preventive action), walks through the CAPA process, shows how to write a CAPA that gets closed properly, and covers how to verify effectiveness, which is the step most CAPAs skip.
What is CAPA?
CAPA stands for corrective and preventive action. It is a quality management process for dealing with problems systematically: find the real cause of a problem, act on it, and prove the action worked. The term is used across manufacturing, medical devices, pharmaceuticals, food, logistics and services, and a CAPA is often recorded in a form or a quality management system so it can be tracked and audited.
A CAPA can be triggered by many things: customer complaints, internal audit findings, defects, service failures, supplier problems, incidents, or a trend that is heading the wrong way. The common thread is that someone decided this problem deserves more than a quick fix.
Correction vs corrective action vs preventive action
These three are easy to confuse, because they often happen in the same week and in the same document. The question each one answers is different.
| Correction | Corrective action | Preventive action | |
|---|---|---|---|
| Question it answers | How do we fix what went wrong right now? | How do we stop this problem happening again? | How do we stop a similar problem happening here or elsewhere? |
| Acts on | The result (the defect, the affected customer) | The root cause of a problem that happened | The cause of a potential problem that has not happened yet |
| Timing | Immediately | After the root cause is verified | Before the problem occurs |
| Done alone, it means | The problem will come back | This problem is fixed, similar ones may not be | The problem that already happened is not fixed |
Example 1: a manufacturing defect
A batch of machined parts is found to be out of tolerance.
- Correction: sort the batch, rework what can be reworked and scrap the rest.
- Corrective action: the root cause is a worn clamping fixture; replace it and add a wear check to the machine's maintenance schedule.
- Preventive action: three other machines use the same fixture type and have not produced bad parts yet; put them on the same wear-check schedule and add the failure mode to the risk analysis.
Example 2: customers charged twice
Several customers complain that one payment was taken twice.
- Correction: refund the duplicate charges and apologise to each customer.
- Corrective action: the root cause is a payment retry that fires when the bank responds slowly, even if the first attempt succeeded; change the retry so it checks the first attempt's status before charging again.
- Preventive action: review other flows with the same retry pattern (subscription renewals, top-ups) before they fail the same way, and set up a daily check for duplicate charges.
Example 3: wrong information from the contact centre
Customers say agents told them they could return a product within 30 days, but the policy changed to 14 days.
- Correction: contact the affected customers and honour what they were told.
- Corrective action: the root cause is that the knowledge base article was not updated when the policy changed, because nobody owned it; update it and assign an owner.
- Preventive action: link every policy document to the knowledge base articles that quote it, so any policy change triggers a review of those articles.
A useful test: if the only thing that happened is that the affected customers or products were fixed, you have a correction. If the cause is gone, you have a corrective action. If you also closed the same gap where it has not bitten yet, you have preventive action.
The CAPA process, step by step
Companies use different names and numbers of steps, but most CAPA processes follow the same logic.
1. Identify the problem
Write down what happened, where, since when and how much, against a baseline. Keep causes and solutions out of the problem statement. "Complaints about X rose from 15 to 140 a month since June" is a problem statement. "Complaints rose because the new system is buggy" is a guess.
Also decide whether this problem needs a CAPA at all. Opening a CAPA for every small issue clogs the system; a good rule is to open one when a problem repeats, affects many customers, has a cost, safety or compliance impact, or when the cause is unclear.
2. Contain it
Protect customers while you investigate: hold suspect stock, add a temporary check, switch a feature off, call the customers already affected. Containment is not the fix, and it should have an end date. It belongs in the CAPA record so everyone knows what is temporary.
3. Investigate the root cause
Find the cause that, if removed, would stop the problem coming back, and verify it with evidence. Use 5 Whys to follow a chain of causes, a fishbone diagram to map candidate causes before you choose, or the root cause analysis template to document evidence. A verified root cause explains when the problem started and why it happens in some places and not others.
Also ask why your existing checks did not catch the problem earlier. In 8D language this is the escape point, and it often deserves its own action.
4. Take corrective and preventive action
Plan actions that remove the verified cause (corrective) and close the same gap in similar processes, products or regions (preventive). Each action needs an owner, a due date and a clear link to the cause it addresses. If an action does not address a cause in the investigation, ask why it is there.
5. Verify effectiveness
Check, with data, that the actions worked: the problem stopped and has not come back. Define the metric, target and time window before you act. This step is covered in detail below.
6. Standardise and close
Update procedures, work instructions, checklists, training material, system settings and risk analyses so the fix is part of how work is done. Share the lesson with teams that run similar processes. Then close the CAPA formally, with a named person signing off.
For serious customer complaints, many teams run these steps as an 8D. 8D adds team formation, a structured problem description (is/is-not) and recognition at the end, and the result is usually recorded as a CAPA.
How to write a CAPA that gets closed properly
A CAPA that drifts for months usually has a weak problem statement, a vague root cause or actions nobody can verify. Writing each field well from the start is what makes closure possible.
| Field | Weak | Strong |
|---|---|---|
| Problem | Customers unhappy with billing | Tickets tagged "charged after cancelling" rose from about 15 a month to 140 in September; baseline Jan–May |
| Containment | Monitor the situation | From 2 Oct, finance checks every renewal against the day's cancellations before charging; owner: billing lead; ends when the fix is verified |
| Root cause | System error | Annual renewals run in a separate billing job that ignores cancellations made in the mobile app (app cancellation launched in June); confirmed in 412 accounts |
| Corrective action | Fix the bug | Annual renewal job checks the cancellation flag from every channel; owner: billing engineering; due 20 Oct |
| Preventive action | Be more careful | Release checklist requires a test of every plan type and channel for cancellation changes; weekly report of charge-after-cancel tickets with an alert above 5; owner: QA lead; due 31 Oct |
| Effectiveness check | Action done | Zero annual accounts charged after cancelling in the monthly billing audit, and charge-after-cancel tickets at or below 15 a month, at 30, 60 and 90 days |
Illustrative example, not customer data.
A few rules that keep CAPAs closable:
- One problem per CAPA. If the investigation finds two unrelated causes, split them.
- Numbers in the problem statement, with a baseline, so you can tell later whether it improved.
- Evidence for the root cause: data, a test, a reproduction, not just agreement in a meeting.
- Every action linked to a cause, with an owner and a date.
- Success criteria written before acting, so nobody moves the goalposts afterwards.
- A realistic scope. A CAPA that tries to redesign the whole company never closes.
Next step: start from a filled-in example with our free corrective action plan template. It has a CAPA plan, an action log with overdue highlighting and an effectiveness check with 30/60/90-day results.
How to verify effectiveness
Effectiveness verification answers one question: did the actions solve the problem? Not "were the actions completed", which is a different and much weaker question. A training session can be completed and change nothing.
A practical approach:
- Choose the metric before you act. It should measure the problem itself: complaint rate, defect rate, repeat incidents, audit findings. For the example above: charge-after-cancel tickets per month and accounts charged after cancelling.
- Record the baseline, both before the problem started and at its peak.
- Set a target and a time window. For example: at or below the old baseline at 30, 60 and 90 days after the fix is live. Pick a window long enough to cover the normal cycle of the problem; a monthly billing issue needs at least a few billing cycles.
- Check that the cause is gone, not only the symptom. If the root cause was a billing job ignoring a flag, test that job directly as well as watching complaints.
- Look for side effects. Did the fix create a new problem somewhere else?
- Decide honestly. If a check fails, reopen the investigation. The root cause may have been wrong, or only one of several. Adding more actions on top of an unverified cause rarely helps.
Many teams record the result as PASS or FAIL for each check and close the CAPA only when every check passes.
Common CAPA mistakes
- Closing on "retrained staff". Training is sometimes part of the answer, but on its own it assumes the problem was people not knowing what to do. Ask what in the process allowed the mistake, and change that. If you do retrain, check whether the error rate actually fell.
- No effectiveness check, or an effectiveness check that only confirms the action was completed.
- Recording a correction as a corrective action. Refunding customers or reworking parts fixes this instance, not the cause.
- "Human error" as the root cause. People make mistakes in every process; the root cause is why the process let the mistake through.
- Actions not linked to the root cause. A long list of good ideas that do not address the verified cause.
- Containment that never ends, hiding the problem and costing money every day.
- Too many CAPAs. Opening one for every small issue buries the important ones. Use clear criteria.
- Working from a sample. If the evidence is a few complaints someone remembered, the root cause may be the loudest one, not the biggest.
- No preventive look-across. The same weakness often exists in similar products, sites or flows.
What about preventive action in ISO 9001 and FDA rules?
Standards treat preventive action differently, which is a common source of confusion. A short summary, not legal or compliance advice:
- ISO 9001. The 2015 edition dropped the separate preventive action requirement. Prevention moved into planning, as actions to address risks and opportunities (often called risk-based thinking), while nonconformity and corrective action stayed in the improvement section. A new edition, ISO 9001:2026, was published in September 2026 and keeps the same top-level clause structure (clauses 4 to 10); certified organisations have a transition period. Check exact wording against your own copy of the standard.
- ISO 13485 (medical devices). ISO 13485:2016 keeps separate clauses for corrective action and preventive action.
- US FDA medical device rules. CAPA used to be spelled out in 21 CFR 820.100. Since 2 February 2026, the FDA's Quality Management System Regulation (QMSR) incorporates ISO 13485:2016 by reference, so CAPA requirements for device makers now come through that standard, alongside FDA-specific requirements.
Whatever your standard says, the practical logic is the same: fix what happened, remove its cause, look for where else it could happen, and prove it worked. If you work in a regulated industry, your quality and regulatory team should confirm what applies to you.
Where customer feedback fits in CAPA
For many companies, the "identify" step starts with customers: complaints, support tickets, survey comments and reviews. That is also where it often goes wrong. A rising problem can sit unnoticed in thousands of comments, or be filed under the wrong category, and by the time it reaches a CAPA the investigation works from a handful of examples someone remembered.
This is the part Pivony helps with. Pivony brings tickets, NPS surveys, call centre transcripts, app reviews and other feedback into one place, groups what customers say by meaning, and overlays your segment and operational data, so a CAPA starts from which issue, which segment, which channel and since when, based on all of your feedback rather than a sample. Watching the same feedback after the fix also gives the effectiveness check a second, customer-side signal. The CAPA itself, with its owners, actions and sign-off, stays in your quality process.
Related reading
- 8D problem solving: the eight disciplines, with an example
- Root cause analysis in customer feedback: the complete guide
- 5 Whys root cause analysis: how it works, with CX examples
- Fishbone analysis: complete guide for CX teams
- Free root cause analysis template
- Free 8D report template
- Free FMEA template
Next step
Download the free corrective action plan template and write your next CAPA from the filled-in example. If the problem starts in customer complaints, request a free root cause audit on a sample of your tickets, reviews or survey comments, or book a demo.
Frequently asked questions
What is CAPA?
CAPA stands for corrective and preventive action. It is a quality management process for fixing problems for good: identify the problem, contain it, find and verify the root cause, take corrective and preventive actions, verify that they worked, and standardise the fix.
What is the difference between corrective action and preventive action?
Corrective action removes the root cause of a problem that has already happened, so it does not happen again. Preventive action removes the cause of a potential problem that has not happened yet, for example by closing the same weakness in similar products, processes or sites.
What is the difference between a correction and a corrective action?
A correction fixes the immediate result, such as reworking a defective batch or refunding a customer charged twice. A corrective action removes the cause, such as replacing the worn fixture or fixing the payment retry that caused the double charge. Without the corrective action, the problem comes back.
What are the steps of the CAPA process?
Most CAPA processes follow six steps: identify and describe the problem, contain it to protect customers, investigate and verify the root cause, take corrective and preventive actions, verify effectiveness with data, and standardise the fix before closing.
How do you verify the effectiveness of a corrective action?
Choose a metric that measures the problem itself before you act, record the baseline, set a target and a time window, for example 30, 60 and 90 days, and check that the cause is gone as well as the symptom. Close the CAPA only if every check passes; if one fails, reopen the investigation.
Is retraining a valid corrective action?
Rarely on its own. Retraining assumes the problem was that people did not know what to do. Usually the process allowed the mistake and did not catch it, so the corrective action should change the process. If retraining is part of the answer, measure whether the error rate actually fell.
Does ISO 9001 still require preventive action?
ISO 9001:2015 dropped the separate preventive action clause and handles prevention through actions to address risks and opportunities, often called risk-based thinking. ISO 9001:2026, published in September 2026, keeps the same top-level clause structure. Check the exact wording in your copy of the standard.
What is the difference between CAPA and 8D?
CAPA is the wider quality process for corrective and preventive action. 8D is a structured, team-based method for investigating and fixing one serious problem, from emergency response to recognising the team. Many companies run an 8D for serious complaints and record the result as a CAPA.
When should you open a CAPA?
Open a CAPA when a problem repeats, affects many customers, has a cost, safety or compliance impact, or has an unclear cause. Small one-off problems with an obvious cause usually need only a correction that is logged. Opening a CAPA for everything buries the important ones.
Compare it on your data
A short demo shows the difference. If you would rather skip the call, start with a free RCA.
Related articles
Free quick scan
See your top complaint themes in 24h