Description
Trained observers record threats, errors and undesired states during normal operations under strict non-jeopardy conditions, then analyse how they were managed.
LOSA observes routine work, not incidents. Peer observers ride along on normal flights and code, with a standard form, the threats (external complications such as weather or ATC), errors made by the crew, whether each was detected and managed, and any undesired aircraft states that followed. Data are confidential and de-identified, which is what makes honest observation possible. The TEM framework then reveals which threats are most frequent, which errors go unmanaged and which countermeasures work. For UX this is an exceptional model for observing expert users of complex tools in their real work: it counts errors that never become incidents, measures recovery, and reveals the gap between procedures and practice.
Attributes
- Discipline
- Human Factors & Error
- Loss type
- Unmigrated
- Stage
- Post-launch
- Data
- Mixed
- Effort
- Months
Steps
- Agree with the organisation the ten LOSA operating characteristics, especially anonymity, voluntary participation and non-jeopardy.
- Train observers who know the domain to use a TEM coding form for threats, errors, undesired states and their management.
- Observe a representative sample of normal work sessions (aviation LOSAs typically use 50-100+ observations).
- Code each threat and error with type, who detected it, response (managed, mismanaged, unmanaged) and outcome.
- Aggregate to find prevalence and mismanagement rates, and identify the threats and interface features most associated with unmanaged errors.
- Feed back findings to frontline users and designers, and repeat the audit after changes to measure improvement.
Sources
- ICAO, Line Operations Safety Audit (LOSA), Doc 9803 (2002)
- Helmreich, Klinect & Wilhelm, Models of threat, error and CRM in flight operations (1999)
Also known as
- Line Operations Safety Audit
- Threat and Error Management
- TEM
Categories
- Accessibility
- Human Factors & Error