Human error is not a root cause
“Human error” are the two most expensive words in a CAPA, and they are expensive precisely because they feel so cheap to write. They close the investigation. They point at a person instead of a process. They let everyone in the room go back to work. And they almost never survive contact with a serious question, which is exactly why an investigator loves to find them.
I have watched hundreds of investigations across a quarter century of audits, and I can tell you the pattern without knowing your product. When a CAPA blames human error and stops there, it is not a conclusion. It is the place a weak investigation gave up.
What the regulation is actually asking for
Corrective and preventive action has one job, and the regulation is unusually plain about it. The requirement, long carried in 21 CFR 820.100 and now brought forward under the QMSR through ISO 13485, is to investigate “the cause of nonconformities relating to product, processes, and the quality system,” then identify and verify the action needed to correct the problem and prevent its recurrence. Read those two ideas together. You have to find the cause, and the action you take has to prevent recurrence.
That second half is the trap “human error” walks into. To prevent recurrence, an action has to change something that will still be there next week. You cannot prevent the recurrence of a person having a bad moment. There is no corrective action for “be more careful.” So the moment you write human error and close the file, you have written a CAPA that, by the regulation’s own logic, cannot do what a CAPA is for. It is not just weak. It is non-conforming on its face.
The four questions you owe before those two words
Human error is a symptom. The cause sits underneath it, and it is almost always in the system that put a person in a position to fail. Before human error goes anywhere near a CAPA, I ask four questions, and I do not let the investigation stop until each has a documented answer.
Was the person trained? Not “did they attend training.” Was the training effective, current, and matched to the task they actually performed. A training record with a signature is not evidence of competence. It is evidence of a signature.
Were they qualified? Training and qualification are not the same thing. Qualification asks whether this specific person was the right one to perform this specific task at this specific moment, given experience, certification, and the demands of the job.
Did they have the procedure, and was it right? A person who follows a wrong, unclear, or out-of-date instruction perfectly has not made an error. The document has. More failures than anyone likes to admit trace back to a procedure that did not match reality on the floor.
Did they have the time? A step that needs ninety seconds, in a schedule that allows sixty, will fail on a predictable fraction of units, and no amount of care fixes a rate that was designed in. That is not human error. That is a production plan writing checks the process cannot cash.
Answer those four honestly and “human error” usually dissolves into something real: a training gap, a qualification gap, a document defect, a capacity problem. Every one of those is correctable. Every one is preventable. Every one is a cause you can actually close.
How one training gap travels
Here is what makes this more than a paperwork argument. A weak cause does not just produce a weak CAPA. It leaves the real cause in place, and real causes move.
Picture a single training gap. An incoming-inspection procedure gets revised, but the retraining reaches only the day shift. The night shift keeps working to the old method. A batch of nonconforming components clears inspection and moves into the line. Production builds them into subassemblies before the mismatch is caught. Now the problem is in three departments: incoming quality, where it started; manufacturing, which built on top of it; and CAPA, which is about to investigate the escape. If that investigation writes “inspector error” and retrains one person, the training system that missed the night shift is untouched. The next revision will escape the same way. The cause did not go away. It went looking for its next department.
I have watched a single unasked training question surface months later as a complaint trend, a scrap spike, and a supplier dispute, three symptoms in three functions, all tracing back to the same root nobody named the first time. That is the real cost of the cheap words. Not the one event they closed, but the recurrence they guaranteed.
Automation raises the bar, it does not remove it
It is tempting to think software solves this, that enough automation takes the fallible human out of the equation. It does not. It relocates the four questions. Was the model or the tool qualified for this use. Was the person operating it trained to know when it is wrong. Did the procedure define who owns the decision. When you automate, you are obligated to keep a human in command of the judgment that matters, and “the system did it” is no more a root cause than “the operator did it.” Under the QMSR the personnel competence and training expectations come in from the incorporated standard with no less force than before. The bar did not move. The number of ways to trip over it went up.
The takeaway
The next time “human error” appears in a draft investigation, treat it as an unfinished sentence, not a conclusion. Make the investigator finish it: the person erred because the system let them, and here is the part of the system we are going to change. If you cannot name that part, you have not found the cause yet, and you certainly cannot prevent it from happening again.
Blaming the person is the fastest way to close a CAPA and the surest way to keep it open forever, one recurrence at a time. Ask the four questions. Fix the system that answered them badly. That is the whole discipline, and it is the difference between a quality system that learns and one that just files.