Root cause as a U-turn

Abstract Failure analysis is the process of identifying the causes and factors leading to undesired loss of functionality. Failure investigators use several kinds of notions to explain this loss. An important one is that of a root cause, but investigators still disagree about the exact meaning of this term. We maintain that two approaches to define root causes can be found in the literature. One originates in backward-looking causal analysis, which aims at determining the causes and factors accompanying a specific failure event; it is token-based and comprises mainly deterministic reasoning. The other is associated with forward-looking effects analysis, which is type-based, and sets out to find correctable factors and prevent recurrence by mainly probabilistic reasoning. Drawing on case studies from the engineering failure-analysis literature, we propose to combine the two approaches to form a new sensible notion of root cause as a U-turn.