01

What is changing or failing?

Refineries produce incident investigations, root-cause analyses, work-order histories and reliability recommendations. Yet the same equipment symptoms and failure patterns can return because the documents are treated as completed records rather than inputs to future decisions.

The investigation may identify a plausible cause but stop before the operating conditions are tested. A corrective action may be assigned without a measurable acceptance condition. Similar failures elsewhere may remain hidden because equipment names, taxonomies and document formats do not match.

The problem is therefore not simply that knowledge is difficult to search. It is that the chain from evidence to corrective action and verified effect is incomplete.

OSHA’s process-safety guidance makes the distinction explicit: corrective action includes prioritisation, responsibility, follow-up and documentation, while refinery inspection guidance looks for both resolutions and verification that deficiencies were corrected. A task marked complete is therefore evidence of activity, not proof of risk reduction.

02

Why does it matter commercially?

A repeat failure carries more than the cost of another repair. It can consume production opportunity, emergency labour, spares, contractor capacity and management attention. It also weakens confidence in the reliability process: teams learn that an investigation can close without changing the underlying exposure.

Buying another analysis tool will not solve this if the organisation cannot distinguish an observed symptom, a tested cause, an action and proof that the action worked.

03

What must management decide?

Management must decide which recurring failure family deserves a closed-loop review and who owns the result across operations, maintenance and technical authority.

The review should answer four questions:

  1. What evidence supports the stated cause?
  2. Which action was meant to change that condition?
  3. Was the action implemented as intended?
  4. What evidence shows the failure exposure has reduced?

Only then should the organisation decide whether better retrieval, changed maintenance strategy, equipment modification or a different operating control is required.

Separate three closure decisions

  • Investigation closure: the event, contributing factors and uncertainty are documented.
  • Action closure: the approved remedy has been implemented and affected documents, training and controls are updated.
  • Effectiveness closure: evidence from a defined operating period shows that the exposure has changed as intended.

Using one status for all three hides the point at which management is still carrying risk.

04

What evidence is required?

  • Failure notifications, work orders and operating context for the selected equipment family
  • The original investigation, assumptions and evidence trail
  • Corrective actions, owners, due dates and completion records
  • Repeat events, near misses and related symptoms across comparable assets
  • Changes to procedures, maintenance plans, spares or operating limits
  • A verification measure and review date after implementation
  • Evidence that related procedures, training, drawings and maintenance strategies were updated where required

AI-assisted search can help find related events and compare language across documents. It should surface patterns and conflicts, not declare technical causation.

05

What should happen next?

  1. Select one repeat failure family with a visible operating consequence.
  2. Reconstruct the evidence-to-action chain for the most recent cases.
  3. Separate unverified causes from tested causes and completed tasks from effective remedies.
  4. Assign one owner for closing the remaining decision gaps.
  5. Review the result after a defined operating period before applying the method more broadly.