# Five whys always stop in the same place. Just before the answer implicates an organisational decision. The chain then ends on "the operator did not follow the procedure", and the corrective action writes itself: a refresher course. **Six months later, the deviation is back.** "Human error" It is an observation: somebody did something other than intended. It is not the cause. A cause is what, once removed, prevents recurrence — and you do not remove a human from a workstation. The five whys chain stops before reaching the organisational level.W1W2W3W4W5StopWhat lies beyond: the organisational decision. Inadequate root cause analysis is among the most frequent regulatory findings in the pharmaceutical industry. The test is simple and unforgiving: **mentally replace the person with another one, equally trained, on a day of normal workload. Does the deviation happen again?** If the answer is yes, the cause is not in the person: it is in the workstation, the procedure, the tool, or the workload. In the vast majority of files we review, the answer is yes — and nobody had asked the question. "Our analysis method is validated, so our investigations are sound." The method is almost never the problem. Five whys, Ishikawa, fault tree analysis are honest tools. **What is missing is the implicit permission to go all the way.** When the fifth answer points to under-staffing, an unworkable schedule, or a procedure never adapted after an equipment change, the investigator knows they are opening a file above their pay grade. So they stop one box short. This is not a competence problem: it is a mandate problem. ## Three requirements the method alone does not cover. ### Proportionate effort ICH Q9 asks that the formality of the analysis be commensurate with risk. Applying the same template to every deviation guarantees two defects at once: too much effort on minor cases, and not enough on the ones that deserved it. Selection is a requirement, not a convenience. ### A cause that explains everything A valid root cause must explain **why it happened at that moment and not before**. If the cause you retained had already existed for two years without producing a deviation, it is incomplete: the trigger is missing. It is the quickest check to run on a report, and the most often skipped. ### A transversal view The pharmaceutical quality system asks you to use trends, not only cases. Three apparently unrelated deviations can share a single cause — an organisational change, a modified tool. Handled in isolation, they produce three ineffective CAPAs instead of one action that works. ## Effective QA looks at the product and at getting the action right, more than at the inspector and the paperwork. Going all the way to the real cause exposes the organisation: sometimes you will have to write that the schedule was unworkable, or that equipment changed without the procedure following. **That is uncomfortable, and it is precisely the job.** An investigation that protects the organisation from its own analysis protects neither the product, nor the patient, nor the site on inspection day. ## A harder analysis, a lighter backlog. Recurrence A cause genuinely addressed does not return. It is the only known way to bring deviation volume down for good without adding people. Standing A report that names an organisational cause and proposes action at that level establishes the quality function as a board-level counterpart. "Human error" establishes it as a clerk. The floor Operators stop reporting when they know the conclusion will land on them. Looking for the cause somewhere other than the person brings weak signals back up — an immediate, measurable effect. So the lever is not buying another method. It is **giving the investigator an explicit mandate to go beyond the workstation**, and deciding in advance who arbitrates when the cause points at a management decision. Without that mandate, the best method in the world will keep producing the same conclusion. ## Go further. [ ### The CAPA that holds Prove the effect, don't just close it. Read →](/en/blog/capa-effectiveness-corrective-action/)[ ### OOS and human error What lab onboarding hides. Read →](/en/blog/oos-human-error-onboarding-qc-lab/)[ ### The backlog is a symptom Why it rebuilds itself. Read →](/en/blog/deviation-backlog-symptom/) ## Root cause, plainly. Should five whys be abandoned? + No. It is an excellent unblocking tool on simple, single-cause deviations. It becomes misleading on multi-cause ones, because it imposes a linear chain where several factors combine. On those, a fault tree or a structured Ishikawa gives a more honest picture. Can you conclude on human error? + You can observe it, and that observation belongs in the file. But it calls for the next question: what made this error likely, here and now? Workstation design, ambiguity in the document, workload at that moment, interruption, an unforgiving tool. The answer to that question is the cause an action can be built on. How do you review an investigation report in five minutes? + Three questions are enough. Does the cause explain why it happened that day rather than another? Does the proposed action remove exactly what is written in the cause field? Is the effectiveness criterion observable and dated? A report that fails any of the three will not hold in front of an inspector. Train the investigators or change the process? + Both, in this order: mandate and arbitration first, competence second. Training investigators to go further without giving them permission to do so produces frustration, not causes. Coaching on real files, with the decision-maker in the loop, is the route that holds. ## How many of your investigations conclude on human error? The proportion is one of the most telling indicators of an investigation system. It is measured on your own history in two days.