Root cause failure analysis, step by step
Root cause failure analysis is a disciplined way to explain a specific failure and to recommend actions that can be checked. It is not a workshop to confirm a cause someone already chose in the hallway. The sequence below is a working order, not a branded method with a trademark.
- Write a problem statement that is observable: what failed, when, under what conditions, and what the effect was. Avoid causes in this sentence.
- Preserve evidence. Parts, photographs, settings, process data, and the work-order history. If the part is already in the scrap bin, say what you no longer know.
- Collect physical, documentary, and human evidence before you lock a mechanism. Talk to the people who were there.
- Build a causal chain from the failed state back to the conditions that allowed it. Separate what is demonstrated from what is still a hypothesis.
- Recommend actions that address those causes, and say how you will verify that the action held — a follow-up inspection, a recurrence check, a process measure.
- Write it down in a form someone else can read in six months.
When the hardware itself has to be examined, bench failure analysis can sit inside the same investigation. For a facilitated RCFA, see root cause failure analysis services. If you are still choosing which failure to investigate, start with bad actors.
Worked example: a hot bearing is an observation
Original fictional exercise. A pump bearing runs hotter after a lubrication task. “Too much grease” is a possible explanation, not a demonstrated root cause. Preserve the as-found condition and compare these alternatives before choosing corrective work.
- Lubrication hypothesis: compare the recorded lubricant, quantity, interval and application method with the approved procedure and bearing arrangement. A recent task establishes sequence, not causation.
- Loading or alignment hypothesis: compare operating duty, maintenance history and available alignment/condition records. A temperature increase alone does not distinguish mechanical loading from lubrication effects.
- Measurement or cooling hypothesis: check measurement location/method, comparable operating conditions and cooling-path condition. A changed measurement method can create an apparent change.
Keep each explanation marked open, supported or rejected, with its evidence and counter-evidence. If records are missing, state what cannot yet be concluded. Any inspection or test needs an agreed safe method and appropriate isolation; this example is not a live-equipment test instruction.
Connect the action to an effectiveness check
If evidence eventually supports a lubrication cause, an action might revise the controlled quantity/method and verify the revised task was carried out. That checks implementation. A separate effectiveness check compares bearing condition and recurrence under comparable duty for an agreed observation period. Completion of training or a work order alone does not prove the failure mechanism is controlled. Assign an owner, due date, measure and review point before closing the action.
See the fictional gearbox investigation report for an example evidence register and action plan. Its outcomes are illustrative, not achieved customer results.
Source and review notes
Public source reviewed November 14, 2026: U.S. Department of Energy, Contractor Investigation, Causal Analysis, and Corrective Actions, EGS 05-01. This archived nuclear-enforcement guidance supports evidence, cause/action linkage and effectiveness verification; it is not presented as a current general-industry requirement. The worked example and wording above are original Apex explanations.
Related reading
- Bad actors and the jack-knife
- RCA / RCFA service
- Bench failure analysis