chapter eight

8 Root cause thinking for developers

 

This chapter covers

  • Breaking the symptom-fix cycle that creates endless rework
  • Applying 5 Whys, Fishbone, Pareto, and Is/Is Not to expose root causes
  • Tracing from stack traces to systemic failures, not code-level guesses
  • Catching AI-generated patches before they become technical debt
  • Writing A3 reports that communicate problems without drama

Your payment service throws 100+ errors every few hours. The pattern looks random. You restart the service, the errors stop, and you move on. Two days later, the same alert fires at 3 a.m. You restart again. A week later, it happens a third time. On the third restart, you stop and actually trace the failure. The service crashes on large transactions because the amount_cents column is a 32-bit INTEGER. Any transaction over $21,474,836.47 overflows the column and throws a SQLException. Three restarts fixed nothing because the column type was the problem the entire time. Lean addresses this through jidoka, the principle of stopping when a defect occurs, correcting its cause, and preventing it from passing further through the system. In software development, that means looking beyond restoring service to understanding why the failure occurred and changing the conditions that allowed it. Restarting the payment service restored availability. Correcting the database schema prevents the same overflow from recurring.

8.1 Going beyond symptoms to the root cause

8.2 Lean problem-solving tools and techniques

8.2.1 The 5 whys

8.2.2 Fishbone diagrams

8.2.3 Pareto analysis

8.2.4 Is/Is not

8.2.5 Combining them: Root cause analysis

8.3 Debugging at the gemba

8.4 When the agent proposes a non-RCA fix

8.5 Countermeasures, not quick fixes

8.6 A3 reports: Structured, blameless problem solving

8.7 RCA beyond your value stream

8.8 Exercise: Write an A3 report for a production issue

8.9 Summary