Root cause analysis and corrective action

The corrective action that most often fails is the one written in the first five minutes after the audit: “Retrain operator. Remind staff of procedure.” It closes the record and leaves the cause in place, so the same finding comes back at the next audit, often with a heavier grade. This guide walks through what ISO 9001 clause 10.2 actually asks for, two practical root cause methods and how to choose an action that will hold.

What clause 10.2 asks you to do

ISO 9001:2015 clause 10.2 sets out a sequence for any nonconformity, whether it comes from an audit, a complaint or a process failure. In paraphrase:

1. React to it: take action to control and correct it, and deal with the consequences.
2. Evaluate whether action is needed to remove the cause, by reviewing the nonconformity, determining its causes and checking whether similar nonconformities exist or could occur.
3. Implement the action needed.
4. Review the effectiveness of the action taken.
5. Update risks and opportunities, and change the management system if necessary.
6. Keep documented information as evidence of the nature of the nonconformity, the actions taken and the results.

The equivalent clauses in ISO 14001, ISO/IEC 27001 and ISO 45001 follow the same pattern; ISO 45001 adds requirements on involving workers and on incident investigation. Step 4 is covered in its own guide, checking a corrective action worked.

Correction is not corrective action

These two words are used interchangeably on many remediation forms, and the confusion is behind many weak responses.

• A correction fixes the instance. The out-of-date drawing is replaced. The overdue gauge is quarantined and calibrated. The suspect batch is held and re-inspected.
• A corrective action removes the cause. The drawing release process is changed so old revisions cannot reach the shop floor. Calibration due dates are visible to the people using the gauges.

Most findings need both, in that order: contain first, then investigate. If you only correct, the finding will recur. If you only take corrective action, the nonconforming product, record or access already out there is still a problem.

Then ask the extent question that clause 10.2 builds in: does the same problem exist elsewhere? If one gauge on line 2 was overdue, check line 1 and the goods-in inspection bench before you close anything.

A worked 5 whys

The five whys method asks why a problem happened, then why that happened, until you reach a cause you can act on. It works well for single-thread problems and is quick enough to use on every finding.

Take a finding from an internal audit at Kestrel Metals, a fictional fabrication business: bore gauge G-207 was used on 2 September to accept a batch, four weeks after its calibration due date.

1. Why was an overdue gauge used? The inspector did not notice the calibration label had expired.
2. Why did they not notice? The label is on the underside of the gauge case and is not visible in normal use, and the inspector was not expecting to check it.
3. Why was the gauge not recalled before its due date? The calibration technician, A. Mensah, works from a recall list, and G-207 was not on the August list.
4. Why was it not on the list? The list is generated from the calibration register, and G-207 had been transferred from line 1 to line 2 in June. The transfer created a new entry with no due date.
5. Why did a transfer create an entry with no due date? The register has no rule for equipment transfers; the transfer form does not ask for the calibration date to be carried over.

The answer at step 1 would have produced “retrain inspector”. The answer at step 5 is a process gap that would affect every transferred gauge, and a check of the register found two more.

Good practice when running five whys:

• Do it with the people who do the work, not just their manager.
• Test each answer against evidence: look at the register, the transfer form, the label.
• Stop when you reach a cause inside your control whose removal would stop recurrence. “Human error” is never a stopping point; ask why the error was possible.
• If an answer branches, follow both branches. Here, step 2 is a second cause in its own right, since the label position contributes even when the recall works.

Fishbone when causes branch

When a problem has several contributing causes, a fishbone (Ishikawa) diagram helps the team look wider before narrowing down. Write the problem at the head, then brainstorm causes under standard headings. For manufacturing, a common set is:

• Method: procedures, work instructions, transfer and handover steps
• Machine: equipment, software, tooling
• Material: inputs, suppliers, labels
• Measurement: gauges, inspection criteria, data
• People: competence, workload, shift patterns
• Environment: layout, lighting, noise, time pressure

For an IT or service team, swap the headings for things like policy, process, tools, people, supplier and data.

The fishbone is a brainstorm, not an answer. Use it to collect candidate causes, then test the likely ones with evidence and run five whys on those that hold up. Keep the diagram with the record; it shows an auditor that the team looked beyond the first explanation.

Common traps in root cause work

A few patterns turn up again and again in weak investigations:

• Stopping at a person. “Operator did not follow the procedure” is a description, not a cause. Ask why it was possible, easy or tempting not to follow it.
• Working back to a pre-chosen answer. If the team already wants a new system or more staff, every chain of whys will lead there. Test each step against evidence.
• Investigating alone. A quality manager writing the analysis at their desk will miss what the people on the line know.
• One cause for every problem. If “communication” or “resources” appears in most of your analyses, the method is not being applied, or the real issue belongs at management review.
• Rushing to meet a deadline. A finding that must be answered within 30 days needs a correction and a plan in that time, not necessarily a finished root cause. Say so, and set a date for the full analysis.

Choosing actions that hold

Once the cause is clear, choose actions that act on it. A useful rule is to prefer actions further up this list, because they depend less on people remembering to do the right thing:

1. Remove the possibility of error. The register will not save a transfer without a due date.
2. Make the error obvious. Calibration labels moved to the top of the case and colour-coded by quarter.
3. Change the process or instruction. The transfer form adds a field for the calibration due date, checked by the receiving supervisor.
4. Train or brief people. Inspectors briefed on checking calibration status before each shift.

Training on its own sits at the bottom because it is the easiest action to write and the weakest at preventing recurrence. It is the right action when the analysis shows a real knowledge gap; for example, a new welder who had never been shown the weld map. It is the wrong action when the procedure was unclear, the system allowed the error or the workload made shortcuts inevitable.

Other checks on a proposed action:

• Does it address the cause found, not a different one?
• Is it proportionate to the risk? A critical finding may justify a system change; a low one may not.
• Does it cover the extent, every place the cause exists?
• Could it create a new risk? Clause 10.2 asks you to update risks and opportunities where needed.
• Is it specific, with an owner and a date?

Writing up the action plan

A complete record, the thing a certification auditor will sample, usually contains:

Element	Kestrel Metals example
Nonconformity	Bore gauge G-207 used to accept product after calibration due date
Correction	G-207 quarantined and calibrated; batch 26-1184 re-inspected with a calibrated gauge and passed
Extent	Calibration register checked: two more transferred gauges had no due date, both now recalled
Root cause	Equipment transfer process does not carry over calibration due dates; label not visible in use
Corrective actions	Register made to require a due date; transfer form updated; labels repositioned
Owner and dates	K. Lewis, register change by 30 September; A. Mensah, labels by 15 October
Effectiveness check	Audit of all transfers and recall lists in January

Corrective actions in Orisynx sit on the finding, are accepted or disputed by the person responsible, and break down into tasks with an assignee, priority and due date. However you record them, keep the cause, the action and the evidence together so the story reads from finding to closure. ISO 9001:2015 dropped the separate preventive action clause, so do not invent a separate preventive record for this; the extent check and the update to risks cover the same ground.