By Cedric Paillard and Captain Andy O’Shea, Amris Aviation
A competency-based system is only as good as the information moving through it. The instructor sits inside a learning system as both sensor and actor — the sensor that produces competency data, and the actor who responds to it. What happens in a simulator doesn’t stay there. Captured properly, it becomes evidence that shapes a cadet’s next detail, informs the design of the scenario itself, and, in aggregate, tells the training organisation and the operator whether their programme is working. This is how isolated judgements become a self-correcting whole, and it is where Evidence-Based Training earns its name: the “evidence” in EBT is a description of how the system is meant to run.
ICAO’s Manual of Evidence-based Training defines EBT as competency-based recurrent training and assessment “based on operational data”, aimed at the pilot’s overall capability rather than performance in isolated manoeuvres. That phrase carries weight: what is trained, how often, and to what standard should trace back to what actually threatens safe operations — and the traceability has to run both ways, from data into training design and from training back out as fresh data. Doc 9995 builds this data-driven basis into its treatment of the EBT development process and instructor systems. This applies that logic to the operator’s ab-initio and type-rating environment, where the same discipline governs how information flows between the ATO and its partner operator.
The Instructor as Sensor and as Actor
Both halves deserve stating plainly, because in practice they drift apart and both suffer for it. As a sensor, the instructor generates the raw signal the whole enterprise runs on: a structured, competency-linked reading of how a crew performed under defined conditions. This isn’t a casual impression. It is the output of ORCA — the observe–record–classify–assess discipline — completed after a facilitated debrief and expressed on a common scale. As an actor, that same instructor consumes signal: the anchors reviewed before duty, the cadet’s recent trend, the cohort patterns surfaced at moderation, and the operator’s line experience all shape how the next session is briefed, coached and graded. A programme that treats instructors as data-entry clerks feeding a dashboard nobody acts on has broken the loop as surely as one that lets them grade on instinct with no feedback at all — the failure the handbook calls “collecting data without feeding it back into design”. The quality of the signal matters as much as whether it exists. A grade with no narrative behind it is noise; a grade that records opinion rather than observed behaviour can’t be aggregated or audited. So the handbook requires every graded event to carry a three-part narrative — observed behaviour, operational effect, agreed corrective action. This isn’t bureaucratic tidiness; it’s the condition under which many instructors’ observations can later be compared and combined.