5WHY Root Cause Analysis Guide

Problem

Many issue investigations quickly drift toward explanations like “operator error” or “normal wear,” without a concrete control or root-cause test. 5WHY is often treated as asking “why” five times, but the useful discipline is to validate each answer and check whether a proposed action can prevent the same failure from recurring.

How It Works

• The user provides an initial issue, such as a machine alarm, surface scratch, or output drop.
• The agent starts a structured follow-up: “Why did XX occur?”
• Each response is checked in three layers: conciseness, requiring one cause at a time; invalid-answer filtering, removing subjective judgments, uncontrollable factors, and normal operating states; and ambiguity confirmation, asking when the answer cannot be reliably classified.
• Once a response passes validation, the agent proposes one specific countermeasure and checks whether that countermeasure can prevent the same issue from happening again.
• If it cannot, the answer is treated as an intermediate cause and the next question is “Why did the previous answer occur?” If it can, the agent outputs the root cause and summarizes the full analysis path.

Boundaries

This skill fits fault diagnosis, quality defects, and production anomalies where the analysis can be closed through cause-and-effect reasoning. It is less suitable for one-off advice, vague complaints, or problems that require external data lookups. Each turn should stay focused, ask one question at a time, and avoid mechanical five-round counting; the stopping condition is whether the proposed action validates the root cause.