The event
A run-of-mine silo on a platinum concentrator sustained a serious structural failure — the kind of event that stops feed to the plant, carries real safety exposure, and pulls senior management straight into the room. A formal Root Cause Failure Analysis was commissioned, and for the first time on this site the investigation was run as a structured method rather than a debate: define the problem precisely, build the why-tree, separate fact from opinion, and list the hypotheses that still had to be verified on the asset.
The technical work was sound. The team reached the point every good RCFA reaches: a short list of candidate causes and a set of actions to go and check, measure and test before declaring the root cause.
Where it broke
The analysis did not fail on the engineering. It failed at the verification step — for a purely organisational reason. Leadership had expected to sit in a boardroom and be handed “problem number one and its solution.” When the method instead asked them to fund and wait for a go-look-see phase — inspect the structure, pull the records, test the loading and operating assumptions — they read it as the analyst not knowing the answer. Confidence drained, the verification work was never authorised, and the investigation quietly stopped one step short of proof.
The two root causes
This is the insight that makes the case worth teaching: the event had two root causes, and only one of them was about the silo.
| Layer | Root cause | Owned by |
|---|---|---|
| Technical | A physical / structural failure mode, reachable through the why-tree and confirmable by inspection and records. | The asset |
| Organisational | A reactive, “silver-bullet” expectation of RCFA — verification seen as delay rather than rigour; no agreement up front that the method requires a go-and-verify phase. | The culture |
The technical root cause is the one everyone looks for. The organisational root cause is the one that actually stopped the job — and it is invisible unless you name it.
Why this happens (and why it isn’t the analyst’s fault)
RCFA is a verification discipline, not a boardroom oracle. Its credibility comes precisely from the step that low-maturity organisations find hardest to sit through: we will not declare a root cause until we have gone and proven it. In a reactive culture, that honesty is mistaken for weakness. The analyst is punished for being rigorous; the room rewards whoever sounds most certain fastest. Naming a cause confidently in the boardroom is easy — proving it on the asset is the job.
How to stop it stopping — the fix
The failure is preventable, but the fix happens before the analysis, not during it.
- Contract the process up front. Before the first session, get explicit leadership agreement that the RCFA has a verification phase, that “we don’t know yet” is the correct answer until then, and that the deliverable is a proven cause, not a fast one.
- Set the expectation that there is no single silver bullet. Frame the output as ranked, evidence-backed causes with a verification plan — and say so on day one.
- Protect the go-and-verify step with a small, fast win. Deliver one quick, visible confirmation early to hold faith while the fuller verification runs.
- Make leadership the sponsor of the method, not the audience for the answer. Their role is to authorise verification, not to receive a magic number.
Where it fits in the framework
This case sits across two pillars at once, which is exactly why it is valuable teaching material: Pillar V — Defect Elimination (the technical RCFA method itself: why-tree, fact vs. opinion, hypothesis verification) and Pillar VIII — Integration, People & Sustainment (the change-management reality that determines whether the method is allowed to finish). Anyone can name a problem in a boardroom. The discipline is going out to prove it — and the organisations that can sit through that step are the ones that get reliable.
A reliability case study. Names, sites and identifying details removed. Based on a real first-of-kind RCFA on a Southern African platinum operation.