A taxonomy for organising findings

HFACS was developed in aviation to structure human factors analysis in accidents. Its classification supports grouping findings and comparing patterns across events, provided categories are applied consistently and supported by sufficient evidence from each case.

Its connection with human and organisational factors is direct: the framework aims to extend attention beyond the final action. However, assigning a label in a report does not itself demonstrate that a causal relationship existed.

Four levels of analysis

The framework distinguishes unsafe acts, preconditions for those acts, unsafe supervision and organisational influences. These levels examine matters such as decisions, operator conditions, coordination, planning, resources and organisational processes relevant to the event.

The terminology describes analytical categories within the method. It should not turn a hypothesis into an accusation or attribute intentions without evidence. Unintentional error and deliberate departure from a rule require different evidence and explanations in the investigation.

From individual cases to patterns

Using HFACS requires clear scope, sources and coding rules. Interviews, records and observations should justify each classification. Comparison between analysts helps reveal inconsistent interpretations before findings are consolidated into a shared database for further analysis.

For example, several incidents involving omitted checks might share planning or information-availability problems. Counting them only as individual failures would conceal that pattern, while automatically assigning every organisational category would not prove it either.

Adaptation and limitations

Applications outside aviation should establish whether the taxonomy adequately represents the context. Some adaptations add or modify categories, making version identification important. Results from different frameworks should not be combined as though they were directly equivalent.

HFACS helps organise and compare information but does not replace event reconstruction or assessment of controls. Subsequent actions should connect to supported findings and their effectiveness should be checked. Analysis quality depends on both the evidence and the consistency of interpretation. Reports should retain uncertainties rather than forcing incomplete information into a category merely to fill a field in the investigation database.