Start from the event and ask why? until you hit a system cause (usually 3–5 levels). Add parallel chains for distinct causal branches.
💡 Stop rule: you've gone far enough when the answer is a management-system weakness (no standard, no training, no check…) — not a person. "Operator inattentive" is never a root cause.
Add causes to each of the 6M branches — the diagram draws itself. Click a cause to remove it.
ILCI loss-causation model: tick what applies — immediate causes (what you saw) then basic/root causes (why it existed).
Each action should target a root cause, at the highest feasible level of the hierarchy of controls.
| Action | Type | Root cause addressed | Owner | Due |
|---|
📈 Frequency & severity rates
🔺 Bird pyramid — your reporting health
Bird (1969) found ≈ 1 serious : 10 minor : 30 damage : 600 near-misses. If your near-miss layer is thin, you're not seeing the base of your iceberg — strengthen no-blame reporting.