Investigating how the loss became possible

CAST, Causal Analysis based on System Theory, applies the STAMP model to learning after accidents or incidents. It examines why existing controls failed to prevent the loss and which changes could strengthen the system’s safety controls.

It differs from root cause analysis reduced to a search for one explanation. Investigation considers interactions and decisions across levels, without assuming that the final visible action contains the complete explanation of the event.

Reconstruct responsibilities and context

The loss, system boundaries and necessary safety constraints are defined. Analysis then examines the control structure: responsibilities, actions, available information and feedback channels, including external organisations where they are relevant to the case being investigated.

The question is not only what an actor did, but why it could appear appropriate given available information. Pressures, objectives, resources and ambiguous signals help explain decisions without turning every explanation into an excuse or replacing separate legal responsibilities.

Avoid conclusions dictated by the outcome

Knowing how an accident ended makes earlier signals appear obvious. CAST seeks to reduce that bias and document what people could know at each moment, separating it from information discovered during the subsequent investigation.

For example, authorisation based on incomplete information calls for examination of how that information was assembled and verified. Simply concluding that someone should have checked more carefully fails to explain why decisions could rest on an incorrect picture of the operation.

Traceable recommendations

Recommendations should address specific control weaknesses, such as incompatible responsibilities, missing feedback or inadequate decision criteria. Follow-up should check implementation and whether changes actually improve the ability to maintain the necessary safety constraints.

CAST is neither a template that automatically produces answers nor a mechanism for assigning blame. It requires evidence, access across organisational levels and competent analysis. A useful report preserves unresolved questions and states its limitations rather than presenting untested hypotheses as established facts. The strength of each conclusion should remain proportionate to its supporting evidence.