Assessing a trainee without breaking them: building a defensible grade

A flight instructor and trainee in a cabin trainer during a training exercise, with the instructor pointing to a seat and the trainee hesitating under focused lighting.

A trainee opens their door without looking outside, the exercise stops there, and the session ends on a failure. Six weeks later the file resurfaces, and the only written record amounts to three words: “lacks assertiveness”. The decision was probably correct; it has become indefensible, because it now rests on nothing but the memory of the person who made it. A grading almost never falls on the mark itself. It falls on what was not written down the same day, on a criterion nobody had announced, or on the trainee next door treated differently for the same deviation.

What gets challenged is almost never the mark

When an assessment is challenged — at a progress review, before a training manager, later before a personnel department or an inspector — nobody asks whether the trainee was good. Three things are asked, in this order. The fact: what exactly happened, at which point in the exercise, under what conditions? The procedure: was the criterion known to the trainee before starting, was the assessor qualified to conduct that check, did the session run as the approved programme provides? Equal treatment: did anyone else commit the same deviation in the same month, and what was decided for them?

An instructor who answers those three questions by re-explaining their impression has already lost. The file must speak without them, weeks later, before someone who was not in the cabin trainer. That is the only useful test to put your own grading form through: read it again assuming you remember nothing.

This requirement is not a legal precaution in disguise. A grade that holds is first of all a grade that teaches. The trainee who reads “order given once, two people in the rear row remain seated, order not repeated” knows what to do differently the very next day. The one who reads “lacks assertiveness” learns nothing and has nothing to refute — which is more or less the definition of an unfair assessment.

A good grade explains exactly what happened and why it matters. A bad grade just says someone failed without showing how or why. The difference is whether the trainee can learn from it or not.

Two ways of recording the same momentAt the top, an observation in a single block: an overall judgement, with nothing to support it. At the bottom, the same observation broken into five elements — the moment, the conditions, the behaviour, its effect and the standard aimed at — which remains readable long after the session.A single impression“lacks assertiveness”nothing to counter six weeks laterFive dated elementsmomentconditionsbehavioureffectstandardcan be re-read, checked, discussed
An observation that holds can be broken down. The block at the top can neither be verified nor refuted. The one at the bottom gives the trainee what to do differently, and the assessor something to show.

The camera test

A criterion is observable if a camera fixed to the ceiling could have recorded it. That rule removes, at a stroke, half the entries that clutter grading forms. “Stressed” cannot be recorded; a faster delivery, a sequence performed out of order and a silent self-correction can. “Not engaged” cannot be recorded; two days without a single question can, but prove nothing on their own and are therefore not graded.

A usable observation has four elements: a behaviour, expressed by an action verb; its object; the condition under which it appeared; its effect. Remove one and the sentence becomes an opinion again. Written into the file, it carries two more: the point in the exercise where it appeared, and the standard it is referred to. “Goes to the wrong station” has no effect; “looks for the extinguisher at the neighbouring station, retraces their steps, resumes the sequence with a visible delay” does, and that delay is what counts on the line.

The distinction to hold firmly is the one that separates the competency from the indicator. A competency is graded — applying a procedure, communicating under pressure, coping with the workload, taking charge of a cabin — but it is never seen. Only indicators are seen: precise behaviours which, put end to end, warrant a conclusion. The grading form serves to work up from seconds to the competency, never the other way round — and a single isolated indicator never carries the weight of the whole competency.

Three columns, one test: the third could be filmed
Wording that does not hold What it means Wording that holds
“lacks assertiveness” the evacuation order was not followed order given once, at normal voice level; two role-players remain seated; the order is not repeated
“stressed” sequence performed out of order starts moving the handle before the outside check; corrects themselves unprompted, without announcing it
“insufficient knowledge” one piece of equipment wrongly located looks for the extinguisher at the neighbouring station, points to the right location once told
“not engaged” nothing observable none: either a behaviour is cited, or the item stays empty
“very good candidate” halo effect across the whole form announces each step of the sequence; does not check the pressure before closing the stowage

Anchoring the levels rather than grading an impression

A numerical mark without a descriptor is an opinion converted into arithmetic. Two assessors who return marks several points apart are usually unable to say what separates them, and the trainee even less. An anchored scale does the opposite: each level is described by what the trainee does, not by what they are worth. Performs the sequence unaided and self-corrects; performs it after one prompt; performs it with help and does not detect the deviation; does not perform it, or compromises safety. Four lines of that kind are enough, provided they are written before the session and known to the trainee.

The threshold is not the group average, it is the operational standard: what will have to be done on the line, at night, with an unfamiliar crew. A weak class does not lower it, a brilliant class does not raise it. It is the only tenable position, and it is also the one that protects the assessor, since they are not comparing people with one another but each one with a text.

The stage of training, on the other hand, changes everything. The same behaviour is not graded the same way on day three and at the final check. A prompt on day three is teaching and is recorded as such; the same prompt at the check turns the item into an item not achieved, since the help was part of the performance. Writing that rule into the grading form defuses the most frequent argument at progress reviews: “I was helped, and you are counting it against me”.

Good grading uses clear levels that everyone agrees on. If you can’t explain why a trainee got a 3 instead of a 4, your system is broken. The standard should be what the job actually requires, not what the class can do.

On an A320 cabin crew check, a trainee fails to properly secure an overhead bin. At the initial training, this would be recorded as “needs practice” and retrained. At the final check, it would be marked as “not achieved” because the standard is what’s needed for real operations. The difference is the purpose of the session, not the trainee’s ability.

Two assessors, one group, one thresholdThe same group graded by two assessors. The solid curve, the more severe one, places a substantial part of the group below the pass threshold. The dashed curve, the more lenient one, leaves almost no one there. The trainees’ actual level is nevertheless identical: it is the scale that shifts.pass thresholdmore severemore lenientnumbermark awarded
Without standardisation, the mark mainly says who gave it. The threshold does not move, but each assessor’s scale slides: at equal actual level, the share of the group declared to have failed varies several-fold depending on the assessor.

There remains this spread between assessors, a structural weakness of any human assessment system. It is not corrected by goodwill, but by joint work: standardisation sessions where the same case is graded and the differences in marking are then discussed, double marking on high-stakes checks, and rotation of assessors so that a trainee never depends on a single judgement across their whole course.

Evidence perishes faster than memory

What can still be recalled after the sessionTwo curves against elapsed time. Factual detail, as a solid line, fades quickly. The overall impression, as a dashed line, holds up far longer: it is what remains available when writing late, and it proves nothing.record herefactual detailoverall impressionend of sessionone weekone monthrecalledelapsed time
The detail goes, the impression stays. Writing late does not produce a weakened observation: it produces a different material, made of judgement rather than facts. The useful window closes before the end of the day.

What disappears first is the detail: the exact word, the order of the actions, the point in the exercise where the deviation appeared. What remains is the overall impression, and it is faithful to itself, not to the facts. Hence a rule with no exception: the grading form is filled in during the exercise, the report is written before leaving the site.

What goes into it takes few lines. A short verbatim quote, when something the trainee said is significant. The moment, located by the stage of the exercise rather than by the clock. The conditions: scenario, role held, lighting, presence of smoke, number of role-players, equipment used. What was said to the trainee during the session, and their reaction. Finally the standard the deviation is referred to, quoted, not summarised.

Record what goes well too, and above all the deviations corrected unaided. A file containing only negatives reads like a case for the prosecution, even if it is accurate, and it loses its force at the moment it is needed. A file that shows progress, with its corrections and its achievements, makes the final finding credible because it proves that the assessor was seeing something other than mistakes.

One last point rarely escapes an attentive reader: a report written after the decision carries the decision, not the observation. It is recognisable by its too-perfect coherence, where every line converges on the conclusion. The chronology of the file — time-stamped grading form, debriefing signed the same day, report afterwards — is worth more than any wording.

The right reflex. Write the sentence you would read aloud to the trainee, in front of them, looking at them. If it does not pass that test — because it judges the person, because it cites no fact, because it would have to be commented on to be understood — it will not pass a re-reading six weeks later either. And do read it out: a disagreement is better handled on the spot, over a fact, than in writing over an impression.

The biases that bring a grade down

Four biases are enough to explain most indefensible gradings. The halo effect: a good overall impression colours every column, and the form comes out uniform when the performance was not. The recency effect: the end of the exercise weighs more than everything that preceded it, especially if it was spectacular. The contrast effect: the trainee is graded against the one who has just gone, and not against the standard. Contamination by the file, finally: you go into the session knowing that “this one is borderline”, and you find what you are looking for.

The countermeasures are mechanical, which is good news. Filling in the form item by item during the exercise, rather than from memory at the end, cuts the halo and recency effects. Not re-reading earlier assessments before a check cuts contamination; where the operator’s manual allows it, having checks conducted by someone who did not deliver the training cuts it better still. Putting everyone through in the same order, on the same scenario, with the same briefing, finally deprives any challenge of its most effective argument: the conditions were not the same.

The trap. The mark rounded up so as not to break a trainee who is otherwise likeable moves the risk, it does not remove it. The item not achieved goes out on the line with them; the next assessor inherits a file that says “achieved” and will therefore have to contradict a predecessor alone, which they will rarely do. And if the deviation reappears one day in an occurrence report, the first document actually describing that trainee’s behaviour will be the one from the incident. Leniency protects only the person who grants it, and only until the next flight.

The same person does not do the same job depending on the session
What changes Training session Checking session
Purpose to build competence to establish what has been acquired
Deviation observed it is used: it is named, corrected, done again it is graded; the correction comes afterwards, in another setting
Help given normal, and it is recorded as a step it is part of the performance: the item becomes not achieved
Record produced progress: what has been seen, what remains to be seen dated result, conditions, decision and follow-up
End of session what we work on tomorrow what is validated, what is not, and when it is redone

Declaring a failure without breaking the person

Not every deviation is a failure, and confusing the two wears out trainees as much as grading forms. An isolated deviation that does not affect safety is stated at the debriefing and has no business appearing anywhere else. The same deviation repeated after being named changes in nature: it no longer tells you about an action, but about the ability to take in a correction, and that is recorded. A deviation that affects safety is corrected immediately and re-checked under the same conditions — without that re-check, there is no way of knowing whether it has been corrected or merely understood.

Not every mistake is a failure. Some are just learning moments. Only repeated mistakes that affect safety should lead to failing someone. The key is to give clear feedback and chances to improve.

On a Boeing 737 cabin crew check, a trainee forgets to check the pressure before closing an overhead bin. At the initial training, this would be corrected and practiced again. At the final check, if it happens again after being pointed out, it would be marked as “not achieved” because it affects safety. But if it was just a one-time mistake, it would just be noted for practice.

A trainee in a flight simulator cockpit is being assessed by an instructor holding a clipboard, with the trainee looking concerned.
Constructive feedback must balance assessment with trainee confidence.
Which follow-up for which deviationA two-way matrix: on the horizontal axis the deviation is isolated or repeated after being named, on the vertical axis it affects service comfort or safety. Three cells are settled within the session; the fourth, a safety deviation that recurs, is the only one leading to an item not validated.corrected and re-checkeditem not validatedstated at the debriefingrecorded, to be revisitedscopesafetycomfortisolatedrepeatedrecurrence
Only one cell leads to failing an item. Two questions are enough to decide: does the deviation affect safety, and does it reappear after being named? The other three cells are dealt with within the session and have no business weighing on a file.

The announcement is prepared, because it decides what the trainee will do next. One order works: the fact first, as it was observed; then the standard, as it is written; the gap between the two; then what follows, with its date. No preamble, and above all no finding slipped between two compliments: the device is spotted by everyone, and it discredits the compliments as much as the finding. No suspense either. A session never ends on “we’ll see”.

What protects the person is the separation between the item and them. You do not declare that someone is not up to standard; you record that on this date, under these conditions, this item was not achieved. The nuance is not diplomacy, it is accurate, verifiable, and it leaves open the only useful path, that of the retake. Ask the trainee first to recount what they did: if they name the deviation themselves, which often happens, any disagreement will be about what follows and not about the fact. That is an entirely different conversation.

The record is closed the same day. The report is handed over, the trainee signs that they have received a copy and not that they agree with it, and a comments box remains open to them. A file that contains their disagreement is stronger than a file that does not, because it proves they were heard. A refusal to sign is recorded as a fact, without comment, and prevents nothing.

The path from a deviation to an item not validatedA deviation observed is first named to the trainee and corrected within the session, then re-checked under the same conditions. If the standard is met, training continues and the record stays in the file. Otherwise the item is declared not validated, notified in writing the same day, and the follow-up is decided outside the session.what plays out within the sessiondeviationnamed, correctedre-checkedvalidatedmetnot metitem not validatedwritten notification the same day, follow-up decided outside the session
Failure is an end point, not a starting point. Between the deviation and the item not validated, two steps must have taken place and been written down: the deviation named to the trainee, then the re-check under the same conditions.

What to have checked before you need it

Open the training part of your operations manual and look there for five precise answers, before the day they are needed. Who declares an item not validated: the instructor alone, or after the training manager’s opinion? Within what time must the trainee be informed in writing? How many attempts are provided for, and in what form? Must a second check be conducted by a different assessor? How long are files kept, and what may the trainee obtain from them? If one of those answers is missing, that is where the file will give way, not on your grading form.

A grey area remains, better known than settled in a box: the boundary between a competence deviation and what belongs elsewhere — fatigue, health, command of the working language, personal circumstances. A session grading form is not the tool for that. You record there what was observed, refraining from any hypothesis, and in parallel you open the channel provided by the operator. Mixing the two loses both: the assessment becomes challengeable and the real difficulty goes untreated.

For anyone preparing an instructor or examiner qualification, that is the main shift to accept. You will not be judged on the severity of your decisions nor on your ease in front of a group, but on standardisation — grading as the others grade, on the same anchors — and on traceability: showing, on an old case, what you observed, when you wrote it down, what the trainee knew of it and at what point. Assessor competence is a writing competence as much as a field competence, and it is the one that is least spontaneously developed.

On each of these points, the applicable regulation, the approved programme and the operator’s documentation prevail over this article: it describes a mechanism of assessment, it does not replace your reference framework or your company’s procedures.

Questions fréquentes

Can a trainee be failed on a single exercise?

Yes, when the item affects safety and the deviation persists after being named and then corrected within the same session. What makes that decision tenable is not its seriousness but its traceability: the behaviour observed, the conditions, the correction and the re-check must appear in writing. An item not validated is not, moreover, an end to training: it opens the retake procedure where the operator’s manual provides for one.

How long after the session does an observation remain usable?

In practice, most of its value is lost within the day. Factual detail — the exact word, the order of the actions, the point in the exercise — fades quickly, often within hours, whereas the overall impression holds up far longer and takes its place without anyone noticing. The safe rule is to fill in the grading form during the exercise and to write the report before leaving the site.

Must the trainee sign the assessment report?

Their signature attests that they have received the document and read it, not that they agree with it: say so explicitly when handing it over. Provide a comments box where they can write down their disagreement, which strengthens the file rather than weakening it. A refusal to sign is recorded as a fact, without comment, and blocks nothing that follows.

Should the criteria be announced before the exercise?

Yes, and it is the first point where a contested grading gives way. Not revealing the scenario in no way obliges you to hide the criteria: the trainee must know which behaviours they will be graded on and what threshold is expected at that stage of training. A criterion discovered after the event is almost always dropped at the progress review.

Can a trainee’s attitude be graded?

What is graded are the behaviours that show it, never the attitude itself. “Not engaged” is not a finding but a conclusion, and a conclusion cannot be defended; “does not open their manual between exercises, asks no question over the two sessions, does not go back over the sequence suggested” are facts. If no precise behaviour can be cited, the item stays empty.

Who can go back on a grading once given?

The framework is the operator’s manual: depending on the organisation, a progress review, a training manager or a second assessor may re-examine the decision, most often by having the item taken again rather than by correcting the mark. The oversight authority, as a general rule, does not re-judge a trainee’s performance: it checks instead that the announced process was followed and that it is traceable.

Commentaires

Leave a Reply

Découvrez l’écosystème FL330

FL330, c’est toute l’aviation professionnelle — explorez nos autres espaces.