The case is banal. An audit flags a handful of rotations operated by a captain whose proficiency check had lapsed several days earlier. The inspector classifies the finding as Level 1, issues the notification on a Friday evening, and waits for a plan. On Monday, the operator sends back four lines: flight crew awareness briefing, a reminder of the procedure, an update to the manual at the next revision. The plan is rejected without discussion. The finding itself could have been closed within a few weeks.
The level does not qualify the file, it qualifies the margin
The classification says nothing about how thick the file is or how many pieces are missing. It says where the safety margin stands. A Level 2 finding designates a deviation that could lower the level of safety; a Level 1 finding designates a deviation that does lower it and seriously endangers the safety of the flight. Everything else follows mechanically.
For a Level 2 finding, the authority grants a correction period proportionate to the nature of the deviation — European practice commonly sets an initial deadline of around three months — and an extension is only granted against a plan it considers satisfactory. For a Level 1 finding, the order is reversed: the authority acts first, through limitation, suspension or revocation, total or partial, of the privilege concerned, and only then looks at what you propose.
This reversal explains a good share of the rejections. On a Level 2 finding, the plan is the document that earns a deadline; on a Level 1 finding, it is the document that earns the lifting of a restriction. Two clocks then run in parallel: the containment action, without which activity does not resume, and the closure of the finding, which requires a treated cause and verified effectiveness. A plan that presents the first as the second fails on both counts.
A Level 1 finding means the safety margin is already gone, so the authority acts first to stop the risk. The plan is only accepted if it shows how to fix the problem and prove it works. If the plan doesn’t do that, the restriction stays in place.
An airline receives a Level 1 finding for a crew operating without a valid proficiency check. The authority suspends the crew immediately. The airline submits a plan with a crew awareness briefing and a manual update, but the authority rejects it. The crew remains grounded for another month while the airline revises the training process and implements a new verification system. The delay costs the airline $500,000 in lost revenue.
The five blocks of an acceptable plan
A plan that passes always reads in the same order, and each block answers a question the inspector will ask anyway. Skipping them does not save time: it just moves the question into a request for further information a few days later.
- The containment action. What was done immediately so the deviation stops producing effects: file withdrawn, aircraft taken off the programme, crew pulled from the line. It is dated and verifiable, it does not need to be elegant, and it addresses no cause.
- Scoping the boundary. How many other occurrences, across what population, over what period, and why you stopped there.
- Root cause analysis. Why the deviation could occur, and why it was not detected — two separate branches.
- The root-fix action. What changes in the process, the documentation or the barrier, so that the cause can no longer produce the deviation.
- Effectiveness verification. The observation made afterwards, on what sample, against what criterion, by what date, by which function.
The wording is not decorative: confusing correction with corrective action is one of the most frequent grounds for rejection, because it is the symptom of an analysis that never took place.
| Type | What it addresses | Effect | Expected evidence | If the plan stops there |
|---|---|---|---|---|
| Correction (containment action) | The occurrence found and its ongoing effects | Immediate, before the plan is submitted | Date, decision, corrected record or activity stopped | The finding comes back at the next audit, as a recurrence |
| Corrective action | The cause that made the occurrence possible | When the modified process enters service | Reference of the revised document, tool setting, targeted training | The finding closes with no guarantee against repetition |
| Preventive action | The same cause where it has not yet produced a deviation | After extension to similar areas | Areas examined and a reasoned decision for each | The mechanism resurfaces elsewhere, under a different approval |
Scoping the boundary: the question that brings down otherwise correct plans
After the containment action, an inspector almost always comes back to the same question: how many others? A plan that does not anticipate it is sent back even when its analysis is good, for a simple reason: an unscoped finding cannot be closed, because the authority does not know what it is accepting.
Scoping means producing four elements: the population examined — all files of a given type, all crews holding a qualification, all airframes in a fleet —, the period covered, the search criterion, and the number of occurrences found and then treated. If you have not traced back to the origin of the process, state where you stopped and on what fact: a documentation overhaul, a change of information system, a resumption of operations. A reasoned boundary is accepted; an implicit boundary reads as an omission.
The sample calls for the same rigour. Twenty files picked at random only have value if the draw is described and the result interpreted: zero occurrences out of twenty does not demonstrate the absence of a deviation, it bounds a frequency.
The trap: letting the containment action stand in for the root-fix action. A hundred-percent double-check put in place on Monday reassures everyone and holds for a few months; then staffing tightens, workload returns, and the check disappears without a formal decision. The finding resurfaces at the next audit, this time as a recurrence — a deviation already found and declared treated carries far more weight, because it documents a failure of the management system rather than a process incident.
Root cause analysis stops one step too high
The finding is a symptom. The analysis consists of going down to a cause that an action can grip, and the difficulty is not going down: it is not stopping at the first comfortable step. It always carries the same name — operator error, lack of rigour, oversight — and it is comfortable because it is true. It simply is not actionable.
A root cause is recognised by four tests. It is under the operator’s control. Removing it makes the deviation improbable without depending on someone’s vigilance. It explains why the existing barrier did not hold, not just why the error was made. And finally it is stated as a property of the system — instruction missing from the process, a check with no hold point, a form modified without a review of associated documents — and not of a person.
The third test doubles the work, which is why it is skipped. Every occurrence has two histories: how it was produced and how it went undetected. An overdue proficiency check that goes unnoticed for several days raises two independent questions — why was the deadline not honoured, and why did the alert system say nothing? Treating only the first leaves a dead barrier in the system.
The right instinct: write the effectiveness verification first. Before drafting the action, write the sentence you will write in six months — which record will be produced, on what sample, against what criterion. If that sentence cannot be written, the proposed action is not one: you cannot measure the effect of an awareness briefing, but you can measure the rate of files passing a hold point.
Verifying effectiveness is not proving completion
The two pieces of evidence are both required and do not substitute for one another. Proof of completion shows the action was done: the reference and date of the document revision, the record of the training session, a screenshot of the modified setting. It is produced on the day of implementation. Proof of effectiveness shows the deviation no longer recurs: it is produced later, on new cases, against a criterion fixed in advance.
Three conditions separate a verification from a box ticked. The criterion is decided before the observation, otherwise the result is always interpreted favourably. The sample covers cases after the action, never the ones used for scoping. And the timeframe exceeds the natural cycle of the monitored process: if the deviation can only appear on a six-monthly deadline, a verification after three weeks proves only that nothing was yet due.
What remains most often forgotten: the plan must state what happens if the criterion is not met. A verification with only one possible outcome is not one. The expected answer is simple — reopening the analysis, reverting to the containment action, informing the authority — and its presence signals a mature setup.
The wording that triggers a rejection
Some phrasings are rejected before the analysis is even read. They are not badly written, they are empty: none of them changes the state of the system, none of them can be verified.
| What is written | What is read | What passes |
|---|---|---|
| Awareness briefing for the staff concerned | No barrier added, effect not measurable | A blocking hold point in the process, with its crossing rule |
| Reminder of the existing procedure | The procedure was known: the cause lies elsewhere and was not sought | Modification of the wording at the point that invited misreading, with the revision reference |
| The staff member was retrained | Cause attributed to a person, recurrence guaranteed with the next one | Revision of the material and the competency criterion for the function, not the individual |
| Increased vigilance | Nothing has changed | Cross-check or automatism that makes the deviation visible without an effort of attention |
| Will be taken into account at the next revision | Deadline unknown, action not engaged | Filing date, date of application, containment action covering the interval |
| The file has been completed | That is the correction, not the corrective action | The correction, then separately the cause and the action that addresses it |
Then there is the form, which is not neutral. One action per line, an action verb, a responsible function — never a person’s name, which would make the plan obsolete at the first staff move —, a commitment date, an expected record. Actions depending on a third party carry that dependency in plain sight: the authority accepts a reasoned conditional deadline, not a firm deadline that later slips.
The plan is finally signed by the function that commits to it, the one holding the means to keep it. Signed only by the person responsible for compliance monitoring when it mobilises production resources, it reads as a plan the operator never arbitrated.
What to go check in your own documentation
The Operations Manual, or the organisation’s exposition, describes how you handle findings. Three points are often weak there. Is the handling of internal findings the same as external ones? If it is lighter, your compliance monitoring produces findings that cure nothing, and the inspector will see it in your own records. Who pronounces closure? If it is the function that proposed the action, there is no control. Is the effectiveness verification planned from the moment the finding is opened, with its own deadline, or does it depend on someone remembering?

Also check the interface with the SMS. A Level 1 finding almost always deserves a safety-side analysis in addition to the compliance analysis: one asks why the requirement was not met, the other why the intended barriers did not hold. Merging them to save time produces a document that satisfies neither.
There remains a grey area that no one settles for you: the boundary between a root-fix action you implement on your own and one that modifies approved documentation. In the latter case, the date of application is no longer yours to set, and the plan must say so — filing on such a date, application on acceptance, containment action maintained until then. Plans that promise a firm date on a document subject to approval end up in default of deadline without any fault having been committed. For anyone preparing to take on a compliance monitoring role, this is the most predictable interview topic: you will be asked to walk a plan through from start to finish, and this dependency separates the candidate who has read the procedure from the one who has practised it. The FL330 courses and the FL330 forum regularly work through real, anonymised plans.
This article describes a mechanism and monitoring practices; it does not replace anything. The applicable regulation, your authority’s decisions, and the operator’s approved documentation always prevail over what is written here.
Questions fréquentes
How long do you have to respond to a Level 1 finding?
The answer plays out in two stages. The containment action must be in place and demonstrable immediately, since the authority has already limited or suspended the privilege concerned and will not restore it without it. The full plan, with root cause analysis and effectiveness verification, follows the deadline set in the notification, and that deadline is short by nature. A Level 2 finding follows a different logic, with an initial correction deadline commonly around three months, extendable only against a plan judged satisfactory.
Can the same action be proposed for both the correction and the corrective action?
No, and it is one of the most frequent grounds for rejection. The correction addresses the occurrence found: it completes the file, pulls the crew from the line, stops the activity concerned. The corrective action addresses the cause that made the occurrence possible, so it changes a process, a piece of documentation, a tool or a barrier. A plan that repeats the correction in the corrective-action box proves that no root cause analysis ever took place.
How do you justify the scope boundary when you cannot examine everything?
By describing the reasoning rather than staying silent about it. Name the population, the period covered, the search criterion, and the sampling method if you used one. Anchor the boundary on a verifiable fact, for instance a documentation overhaul, a change of information system, or a resumption of operations. State explicitly that zero occurrences on a sample bounds a frequency without demonstrating an absence: a reasoned boundary is accepted, an implicit boundary reads as an omission.
Can a training action count as a valid corrective action?
Yes, but rarely on its own and never in the form of an awareness briefing. Training counts as a corrective action when the root cause is genuinely a competency gap attached to a function, not a person: the material is then revised, the competency criterion is updated, and every holder of the function is concerned. If the cause is a process that lets the deviation through, training only shifts the burden onto people’s vigilance, and its effect erodes over time.
When should the effectiveness verification be scheduled?
After a delay that exceeds the natural cycle of the monitored process, otherwise the observation can show nothing. If the deviation can only appear at a six-monthly deadline, a verification after three weeks only establishes that nothing was yet due. The criterion and the sample are fixed before the observation, on cases that postdate the action’s implementation, and the plan states what happens if the criterion is not met.
Why does a recurrence weigh more heavily than the initial finding?
Because it no longer documents a process incident but a failure of the management system. A deviation already found, declared treated, and then reappearing proves that the root cause analysis was wrong, or that the action did not hold, or that the effectiveness verification saw nothing. The authority draws a conclusion about the organisation’s ability to correct itself, which touches the credibility of compliance monitoring as a whole.

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