
Incident Investigation: Root Cause Analysis That Actually Prevents Recurrence
Incident investigation establishes what happened, why the event was possible and what needs to change to reduce the chance of recurrence. Root cause analysis is the part that examines correctable weaknesses beneath the immediate event. A useful investigation connects its conclusions to evidence and its actions to those conclusions.
“The worker was careless” is not enough. Neither is an impressive diagram with no supporting records. The aim is a defensible explanation that helps the organisation make a better decision about equipment, work arrangements, competence or management controls.
This guide gives HSE teams and supervisors a general investigation framework, with a fictional worked example. It is not an emergency-response procedure or a technical investigation method for every major event. Complex failures, serious harm, specialist equipment or contested evidence require suitably qualified investigators and any applicable regulatory process.
Start with the right scope and team
Emergency response, protecting people and preventing further harm take priority over collecting material for a report. Only enter an area or inspect equipment when authorised and safe to do so. Follow applicable instructions about preserving the scene and evidence; do not disturb it merely to reconstruct a photograph.
Once the immediate situation is managed, record the investigation's purpose, responsible lead, participants and boundaries. Consider what could reasonably have happened as well as the actual outcome. A near miss can expose an important control failure even when nobody was injured.
OSHA's incident-investigation guidance recommends looking beyond immediate causes and involving management and workers. Their different knowledge helps distinguish the written process from actual working conditions. Keep the learning process focused on evidence rather than an early judgement about fault.
For Pakistan, verify incident-notification, recordkeeping and evidence-preservation obligations against the relevant province, sector and event. The UK and US sources linked here provide investigation principles; their reporting thresholds and deadlines must not be assumed to apply locally.
Gather information before fixing the explanation
Set up a simple evidence register. Give each item an identifier, source, collection date, custodian and relevant limitation. Preserve originals where appropriate and control access to personal or sensitive information. Distinguish a photograph taken after the event from evidence of the conditions at the time.
Physical information: authorised photographs, layout information and specialist examination findings where needed.
Work records: relevant instructions, assessments, work orders, inspection findings and change records.
People's accounts: what each person directly observed, what they were told and what remains uncertain.
Timing information: logs, messages and other records that help establish sequence.
Earlier signals: similar reports or unresolved actions that may explain why the problem persisted.
Interview people respectfully and avoid leading questions. “What information did you have at that point?” is more useful than “Why did you ignore the instruction?” Record conflicting accounts without forcing immediate agreement. A missing record is a limitation, not automatic proof that an activity never happened.
The HSE's HSG245 investigation workbook separates information gathering, analysis, control selection and action planning. It is an established older resource; its historical statistics and legal references should not be presented as current requirements.
Build a timeline and label uncertainty
Construct a sequence from the available information. Each entry should say what happened, which evidence supports it and whether the time is exact or estimated. Keep analysis separate so a later reader can see where evidence ends and inference begins.
For example, a maintenance system may show when a defect was entered, while a witness estimates when it was first noticed. These answer different questions. Do not collapse them into a single supposedly precise timestamp.
Include relevant changes before the event: a new work sequence, replacement equipment, altered staffing or a reporting-system change. Check whether these were causal, incidental or still unresolved. A timeline organises facts; it does not prove causation by itself.
Separate immediate causes from underlying weaknesses
An immediate cause describes a condition or action close to the event. An underlying cause explains why that condition existed. A root cause identifies a system weakness the organisation can address. These categories help organise thinking, but a real event can have several interacting explanations.
OSHA and EPA's root cause analysis factsheet emphasises examining system failures and choosing analytical tools suited to the event. Its US regulatory discussion is not Pakistani law. Use its distinction between a surface correction and a systemic change without importing unrelated legal duties.
Ask whether the proposed explanation accounts for the evidence. If the conclusion is “inadequate training”, identify the specific knowledge gap, supporting evidence and relationship to the event. If people understood the task but could not carry it out with the equipment provided, training alone addresses the wrong problem.
Worked example: a repeated trip hazard
Hypothetical teaching example: a worker trips over a temporary cable across an office passage. The example below assumes the stated records were obtained; it is not an actual incident, legal finding or electrical-work instruction.
Analysis layer | Example finding | Evidence needed |
|---|---|---|
Event | A person tripped while using the passage during a temporary room rearrangement. | Accounts, authorised scene records and location information. |
Immediate condition | A temporary cable crossed the walking route at the relevant time. | Contemporaneous photographs or other reliable evidence of the arrangement. |
Underlying arrangement | The layout change was made without an agreed assessment of pedestrian access and equipment placement. | Change request, instructions and interviews with those organising the move. |
System weakness | The change process had no responsible person or acceptance check for temporary office layouts. | Current process, role assignments and comparable previous changes. |
This analysis is provisional until the evidence supports it. Perhaps the cable was moved after the event, or an assessment existed but was not followed. Those alternatives change the findings and must be tested. Do not select the explanation that most conveniently supports a preferred action.
Use analytical tools without forcing an answer
Five Whys can help examine a relatively simple chain, while a fishbone diagram can prompt questions across equipment, people, methods and organisational conditions. More complex events may require other techniques and specialist expertise. The tool is a way to structure inquiry, not evidence that the inquiry was correct.
There is no need to stop after exactly five questions, force one root cause or make every branch end at “training”. Record unanswered questions. Where evidence cannot distinguish between explanations, state that limitation and identify how uncertainty affects the action decision.
For human factors, examine work design and conditions as well as individual actions. HSE's guidance on human failure distinguishes unintended errors from deliberate deviations and explains why factors such as workload or design can influence performance.
Write actions that address the supported causes
An action should identify the intended change, accountable owner, due date, interim arrangements where necessary and evidence required for acceptance. “Be more careful” and “conduct awareness” do not describe a verifiable system improvement.
For the fictional office example, the following action structure is an illustration for a facilities team to adapt with competent input.
Action | Accountability | Effectiveness check |
|---|---|---|
Have the temporary layout and service arrangements assessed and corrected by the relevant responsible people. | Facilities manager; technical work remains with authorised competent personnel. | Verify the accepted arrangement in use, not just the completed work request. |
Add a named coordinator and acceptance stage for future temporary layout changes. | Facilities process owner. | Sample subsequent changes and confirm the checks occurred before occupancy. |
Communicate the revised reporting and approval route to affected teams. | Department managers. | Ask users to explain the route and examine a later request to see whether it worked. |
Dates and review intervals should match the risk and practical change. Do not invent a universal seven-day closure rule. If a necessary change cannot be delivered promptly, the responsible authority must decide what interim restrictions or controls are needed; an overdue date is not an acceptable control.
Verify closure and share useful learning
Administrative closure asks whether the assigned work was completed. Effectiveness asks whether the underlying problem was reduced under relevant conditions. Keep those decisions separate, especially when a purchase or procedure revision precedes actual implementation.
Communicate the finding, the changed arrangement and what other teams should check. Remove unnecessary personal identifiers. Avoid circulating a dramatic story without explaining what action recipients can take. Link repeated findings to the organisation's health and safety performance review so unresolved causes do not disappear after the investigation meeting.
When developing investigators, compare the ICTQual Level 2 Award in Incident / Accident Investigation with the tasks learners need to perform. Relevant development should include assessing evidence and writing actionable findings, with additional specialist support for events outside the investigator's competence.
Research checked 16 November 2026. Prepared with AI assistance using the linked primary guidance. The example and action plan are hypothetical; independent human technical review is not recorded. Seek specialist assistance where the event's complexity or consequences require it.
Frequently asked questions
These answers cover common decisions about investigation scope and closure.
Yes, proportionately to their potential and circumstances. A fortunate outcome does not establish that controls were effective. Decide the investigation level from the credible risk and any applicable requirements.
It usually describes a starting point for further inquiry. Establish what the person intended, what information was available and what work conditions shaped the action before deciding on corrective measures.
No. Several interacting system weaknesses may contribute. Explain which conclusions are supported and how the proposed actions address them.
No. Each causal link needs support. A discussion can generate hypotheses, but repeating why does not make an unsupported explanation true.
Complete the findings and action assignment through the organisation's process, then track implementation and effectiveness separately. A final report does not prove that its recommendations have worked.
