The visit rarely begins with an open-ended question. The auditor selects a name from the previous month’s schedule, requests the corresponding training file, and asks three questions: the date of the last check, the next deadline, and proof that the approved program was indeed carried out between the two. The tone for the rest of the day depends on how quickly these three answers arrive — and especially on their consistency with each other, because it’s this consistency, not the quality of the training, that the visit measures.
What the audit actually verifies
A training file is not a logbook: it’s a demonstration. It doesn’t recount what happened; it proves that what was supposed to happen did happen, on the scheduled date, by the qualified person, according to the version of the program in effect that day. The auditor isn’t looking to assess whether the crew is competent — they don’t have the time or authority to measure that. They’re looking to determine whether the operator can demonstrate that their training system is functioning, which is a different and, in practice, more demanding question.
This demonstration follows a five-link chain: a regulatory or internal requirement, an approved program that covers it, a session that implements the program, a record that attests to the session, and a deadline that follows from the record. The operator builds this chain from left to right over the course of the year. The auditor, however, reads it in reverse: they start with a displayed deadline and demand the record that supports it, then the session that produced the record, then the program from which the session originated, and finally the requirement that the program covers. Each link must clearly reference the previous one — a record that doesn’t reference its session, or a session that doesn’t cite the program version, are floating links.
L’auditeur vérifie surtout si les documents s’enchaînent correctement, comme une chaîne. Chaque pièce doit prouver la suivante, sans trous. Ce n’est pas la qualité de la formation qui compte, mais la preuve écrite que tout a été fait comme prévu.
Pour un contrôle de navigateur sur un Airbus A320, l’auditeur vérifie : 1) la date du dernier contrôle dans le fichier individuel, 2) cette date correspond à la session dans le registre de l’instructeur, 3) la session fait référence au bon programme de formation, 4) le programme couvre bien l’exigence réglementaire. Si un seul maillon manque, c’est un problème.
A chain rarely breaks at the session level itself: training sessions occur, instructors work, and crews pass their checks. It breaks at the most administrative level — a poorly dated record, a signature from the wrong person, or a deadline calculated from the wrong starting point. This is why a conscientious training manager may leave a visit with findings: the visit wasn’t judging their work; it was judging the trace of their work.
The mechanics of deadlines, the first source of findings
A periodic validity is not an anniversary. In most regimes, it runs until the end of a month, and its renewal follows a non-intuitive anchoring rule: a check passed within the revalidation window — the period immediately preceding the deadline — extends the validity from the original deadline, as if the check had occurred on the last day. A check passed outside the window, too early or after expiration, resets the counter from the date of the check itself. A navigator who anticipates their check too much gains nothing: they delay their deadline.
This anchoring rule is the first thing the auditor tests, because it can be verified in thirty seconds and immediately reveals the quality of the tracking tool. They take a recent check, note its date, note the displayed deadline, and recalculate. If the tool mechanically adds the validity period to the check date, each regular extension “offers” a few weeks of unwarranted validity, and the error propagates: each cycle starts from the wrong date, the gap accumulates, and after a few years, the displayed calendar no longer matches the regulatory calendar. Conversely, a check passed outside the window but treated as an extension creates a period where the navigator flew under a validity that didn’t exist — and then we’re no longer talking about a tool error.
The trap. The spreadsheet that calculates the next validity by adding the period to the check date — instead of anchoring it to the original deadline when the check falls within the window — creates deadlines shifted by several weeks. The discrepancy is invisible until the tool is compared to the instructor’s register and the written rule: it’s exactly the comparison the auditor makes, and it turns a software defect into a non-compliance of the tracking system.
The inconsistencies that tip the visit
An overdue deadline, detected by the operator, recorded, and addressed by removing the navigator from duty and rescheduling: it’s an unpleasant event, but it’s proof of a functioning system. What tips a visit into a formal finding is of a different nature: two records that cannot both be true. A ground session signed on a day the schedule places the trainee on a rotation. An instructor whose qualification, in their own matrix, was issued — or had expired — on the date of the session they signed. A report referencing a program version released after the session. The same check dated differently in the individual file and the tracking tool. None of these pairs require interpretation: the two dates are there, side by side, and one of them is false.

The severity of the response then follows two dimensions, and the practitioner benefits from knowing them because they structure the entire closing discussion: does the discrepancy affect validity — did a navigator operate without a valid qualification? — and is the discrepancy isolated or systemic — a data entry error, or a wrong rule applied across the fleet? The intersection of these two questions determines whether the visit concludes with an observation, a finding with an action plan, or a demand for immediate action.
| Response | Typical trigger | What the authority expects | Horizon |
|---|---|---|---|
| Observation | Isolated discrepancy with no effect on validity: ambiguous caption, missing version reference, corrected entry | An improvement, with no formal obligation for a plan | As things go, reviewed at the next audit |
| Level 2 finding | Demonstrated non-compliance with no immediate impact on safety: wrong calculation rule, instructor qualification not recorded, repeated incomplete evidence | A root cause analysis and an accepted corrective action plan, then closed | Deadline agreed with the authority, met and recorded |
| Level 1 finding | Discrepancy that directly lowers the level of safety, or falsification: crew member operating out of validity, signed session that never took place | Immediate action, which may extend to the limitation or suspension of all or part of the operation | Immediate |
A precision that changes the reading of the table: falsification short-circuits the matrix. A signed session that didn’t occur isn’t a traceability error; it’s a breach of the trust on which all supervision is based — and it’s treated as such, regardless of the operational importance of the session in question. A minor item backdated weighs more than a major item honestly declared late.
Cross-sampled auditing: three systems, one truth
The auditor doesn’t review two hundred files; they trace two or three in depth, and the technique is always the same: take a session and follow it through every system that should know about it. The navigator’s individual file. The day’s schedule and assignments. The instructor’s register and their qualification matrix. The session report, with its program reference. The deadline tracking tool. If applicable, the simulation center’s or subcontractor’s records. These systems are maintained by different hands, at different rhythms, for different purposes — precisely what makes the cross-checking effective: an honest error stands out immediately, and a falsification consistent across all systems is improbable.
| Auditor’s question | Document queried | The triggering discrepancy |
|---|---|---|
| When did this crew member take their last check? | Individual file, expiry tracking tool | Two different dates for the same check |
| Where were they that day? | Crew rostering, assignments | A flight scheduled on the day of the signed session |
| Who ran the session? | Instructor register, qualification matrix | A qualification issued after the session, or expired that day |
| Under which programme? | Operations manual, document management | A programme version issued after the session that refers to it |
| And the next validity? | Tracking tool, written rule in the manual | An applied calculation rule that is written nowhere |
This approach has a direct consequence for the operator: the quality of a file isn’t determined by the file itself. It’s determined by the consistency between the file and everything around it. A training service that meticulously maintains its binders but never reconciles its dates with the schedule is, unwittingly, preparing the raw material for the next finding.
Living file, reconstructed file
Experienced auditors recognize a reconstructed file before even verifying a date, from formal signals: signatures clustered in the few days before the visit, uniform ink and handwriting across months of records, round dates systematically falling at the start of the month, forms of a single version during a period when the manual changed, and — most telling — the total absence of failure, retake, or rescheduling over years of activity. A real training system produces dated disorder: a rescheduled session with its reason, a retaken item, an instructor’s note, a retaken check. This disorder is the signature of a living file; its complete disappearance is a signal, not proof of quality.
One point must be clear: teams sometimes discover too late that reconstructing a missing record by giving it the appearance of a contemporaneous record is falsification, even when the session actually occurred. The honest response to a lost record exists, and it’s accepted: a dated statement from the day it’s established, signed by someone who can attest, explaining what it reconstructs and why. It’s worth less than an original record, but it doesn’t destroy trust — whereas backdating does.
The right reflex. Audit your own sample each month, in reverse, exactly as the auditor will: start with a displayed deadline, demand from your own system the record, then the session, then the program version, and recalculate the anchoring manually. Three files per month are enough — the calculation error or floating link is then discovered internally, where it’s worth a correction, and not during the visit, where it’s worth a finding.
Before the visit: what to verify in your own documentation
Four checks cover the essentials, and none takes more than half a day. The deadline calculation rule must be written in black and white in the documentation — not just encoded in the tracking tool: an auditor who asks “where is this rule written?” and is told “the spreadsheet does it” already has their finding. The instructor and examiner qualification matrix must be maintained with the same rigor as the navigators’ files, including deadlines: it’s the link that operators forget, because those who check others are rarely checked themselves. The reports and waivers must each have their complete trace — who authorized, on what documented basis, and until what date. And the files of departed navigators must remain retrievable for the required retention period: the navigator’s departure doesn’t extinguish the operator’s obligation.
Remains the gray area of the moment: digitization. A fully digital file is perfectly defensible — often more so than a binder — provided the system guarantees what paper guaranteed by nature: a timestamp that can’t be rewritten, nominative write permissions, and a record of all modifications. Two traps concentrate the difficulties. The conversion of paper archives, first: an import that timestamps records to the digitization date erases the only information the auditor wanted — the original date of the session. The electronic signature, next: a simple name typed into a field commits no one, and a system where the administrator can modify a validated record without leaving a trace is worth less in an audit than a paper document with corrections and countersignatures. For those preparing for or overseeing preparation, the practical conclusion is the same: maintain in parallel their own personal log of training and checks, dated as they occur — it’s the reflex that FL330 preparation courses work on, and it’s also the only independent verification of the operator’s system.
The mechanisms described here — validity anchoring, finding classification, record retention requirements — have variations depending on the regimes and authorities: the regulation applicable to the operator, the procedures of their authority, and their approved documentation, with the operations manual at the top, prevail in all circumstances over this article.

Leave a Reply
You must be logged in to post a comment.