Follow-up analysis of an occurrence report: what the authority expects after the acknowledgement of receipt

Tirages photographiques d'une pièce endommagée étalés sur un bureau avec loupe et pied à coulisse

An occurrence report goes out the same evening, the acknowledgement of receipt arrives within the minute, and the case looks settled. It is not. That receipt certifies that a record has entered a database, nothing more: neither that the classification is correct, nor that the narrative is usable, nor that anyone has opened the file. What happens next — the depth of the analysis, the nature of the actions, the way one writes that no conclusion could be reached — decides whether the case closes or comes back as a written question, and then as a finding.

The acknowledgement of receipt certifies a filing, not any processing

In most national systems, the receipt is issued by the exchange gateway before any analyst has read anything at all. It confirms three things: the file is readable, the mandatory fields are present, a unique identifier has been assigned. That identifier is the only useful piece of information in the message: it will serve as the reference for every follow-up query, and an operator who does not match it to its own internal number loses traceability from the very first exchange.

Two clocks run in parallel, and they are often confused. The first runs on the transmission of the report itself: the operator has a few days from the moment it became aware of the occurrence. The second runs on the analysis: the European regulation on occurrence reporting expects preliminary results within the order of a month, then final results within the order of a quarter. Both deadlines are counted from the date of notification, not from the date of the occurrence — a nuance that costs dearly: the time lost internally is not given back to the analyst.

The two clocks of an occurrence reportA timeline carrying four successive markers: notification of the occurrence on day zero, transmission to the authority after a few days, preliminary results of the analysis within the order of a month, final results within the order of a quarter. A light band covers the whole period during which the case remains open at the operator.notificationday zerotransmissiona few dayspreliminaryorder of a monthfinalorder of a quartertimecase open at the operator
The acknowledgement of receipt falls at the second marker, not the last. Both analysis deadlines are counted from the date of notification: a report that surfaces three weeks late gains nothing from it.

The receipt just means the report was received, not that anyone has read it or started working on it. The real work begins after the receipt, and how well you do that work determines whether the case stays closed or comes back to you.

A crew reports a hard landing on an Airbus A320 at 3PM. The authority’s receipt arrives at 3:01PM. The operator has 7 days to send the report, but the analysis must be complete within 30 days of the notification (not the occurrence). If they wait 21 days to send the report, they still only have 9 days left for analysis.

Between the filing and the final results, the case lives at the operator. It is precisely this period that the authority assesses when it draws a sample from it during an inspection: it does not re-read the reports to check that they were sent, it re-reads the analyses to check that the safety management system works. The difference in posture is total: the first exercise is declarative, the second is demonstrative.

The five blocks of a follow-up analysis that stands up to reading

An acceptable analysis rarely fits on one page and never on ten. It comprises five blocks, and the absence of any one of them is spotted in thirty seconds. The factual reconstruction first: phase of flight or of the task, configuration, mass and balance where relevant, external conditions, state of the onboard documentation, history of the part or of the deferred defect. Then the coding, in the European taxonomy, which determines what the occurrence will become once aggregated with the reports of the whole sector. Then the risk assessment, the causal analysis, and finally the action plan together with its effectiveness verification.

The factual reconstruction is the block that is most often rushed, even though it conditions everything else. The downloaded flight parameters, the journey log, the technical log entry raised at the station, the history of deferrals on the same system: these elements already exist and perish quickly. An operator who waits for the crew’s answer before going to fetch the data runs the risk of no longer finding it.

What the authority’s analyst cross-checks first is the consistency between the risk declared and the action retained. An occurrence classified at high potential severity whose only action is a reminder at briefing constitutes an internal contradiction: either the classification is too high, or the action is too weak, and in both cases the case comes back.

What each stage must establish, and what triggers a follow-up query
Stage Question it answers Expected depth Most frequent reason for a follow-up query
Initial report What happened, where, when, under what conditions? Verifiable facts, without interpretation Narrative too short to be coded in the taxonomy
Preliminary results What risk, what barriers, what interim measures? Reasoned classification, first causal hypotheses Classification based on the actual damage instead of the potential
Final results Why did the system allow this, and what is being changed? Cause traced up to organisational level, dated action plan Cause that stops at a person; action with no owner and no deadline
Effectiveness verification Did the action produce the expected effect? Indicator measured afterwards, on a date fixed in advance Block missing: the case is closed on the day the action is decided

How far down to go in the search for causes

The expected depth is described simply, in three levels. Level 1 establishes what occurred: the selection was made the wrong way round, the screw was not lockwired, the altitude was left without clearance. Level 2 identifies the barrier that should have intercepted the deviation and explains why it did not: the cross-check existed but was not placed at the right point in the sequence, the alert was inhibited in that configuration, the task card listed the item but with no pass criterion. Level 3 names the organisational decision that left that barrier in that state: a workload trade-off, a choice made in drafting a procedure, a delay in incorporating a manufacturer’s instruction.

Three levels of depth in the search for causesThree blocks arranged as a staircase descending to the right: the fact observed, the barrier that gave way, the organisational decision that left that barrier in that state. A horizontal dashed line, placed just below the first block, marks the level at which most analyses stop.1 · what happened2 · the barrier that gave way3 · what left it that wayfrequent stop
A cause that names a person is a level 1 in disguise. Until the analysis has named the organisational decision that made the error possible, it has produced nothing that can be corrected.

If you stop at “the pilot made a mistake,” you haven’t really found the cause. The real cause is why the system allowed that mistake to happen. Did the procedure make it easy to make? Was training lacking? That’s what you need to find.

A Boeing 737 crew lands long at an airport with a 2,500m runway. The analysis finds the crew didn’t follow the stabilized approach procedure. But instead of stopping there, the operator digs deeper and finds the procedure had conflicting information about stabilisation criteria, which led to confusion. They then revise the procedure to make it clearer.

The control test is well known and remains the most effective: mentally replace the people involved with others, of equivalent competence, on the same day, in the same conditions. If the occurrence recurs, the cause written down is not the right one. “Failure to follow the procedure”, “lack of vigilance”, “crew error” are not causes: they are restatements of the fact, and they close the investigation at the very moment it was becoming useful. They cost twice over, because they signal to the authority that occurrence processing serves to attribute rather than to understand — which calls into question the just culture policy set out in the manual.

The trap. A cause that names an individual extinguishes voluntary reporting for months. Staff read the feedback, see a name in it, and stop reporting whatever is not mandatory; the volume drops, the operator believes it is an improvement, and the first inspection that compares the voluntary reporting rate with the sector’s opens a very unpleasant discussion.

An acceptable action plan: owner, deadline, evidence

An action is worth only its robustness and the evidence one plans to draw from it. The hierarchy is stable from one sector to another: eliminate the hazard, replace it with a device that makes the error impossible, change the procedure, train, inform. The further down one goes, the more the effect depends on people’s memory, and therefore the faster it wears out. An analysis that proposes only the last rung must explain why the others were set aside, failing which it will be read as a surrender.

Residual effect of two types of corrective actionTwo curves starting from the same high point. The solid line, corresponding to a barrier built into the hardware or the documentation, stays almost horizontal. The dashed line, corresponding to an awareness action alone, declines steadily to the bottom of the frame.built-in barrierawareness onlyresidual effecttime elapsed
The briefing wears out, the lock does not. That is why an awareness action on its own almost always calls for a review date, whereas a change to the documentation is verified once.
Comparative robustness of the types of action and the evidence of effectiveness expected
Type of action What it changes Durability over time Evidence expected at verification
Elimination of the hazard The situation can no longer arise Permanent Single observation: the configuration no longer exists
Technical interlock The error becomes impossible or is detected As long as the modification is in place Embodiment evidence, fleet configuration tracking
Procedure or documentation The right action becomes the easiest one High if the amendment is approved and distributed Referenced amendment, read-and-sign acknowledgement
Training or practice The competence is acquired and assessed Medium, tied to the recurrence of the sessions Assessment results, crew coverage rate
Information, briefing Risk awareness, for a while Low, declining Occurrence indicator reviewed on a fixed date

Every line of the plan carries four attributes, and they are not negotiable: an owner named by function, a deadline, an indicator that will say whether the effect is obtained, and a review date later than the deadline. The fourth attribute is the one that is almost always missing. Without it, the operator closes the case on the day it decides the action, not on the day it observes its effect.

The right reflex. Date the effectiveness verification before dating the action, and enter it in the same tracking table as the other deadlines of the management system. An action closed without an indicator stays open for the inspector; an action closed with a measured indicator raises no question.

Classify the potential risk, not the damage observed

The classification determines the priority, therefore the resources, therefore the depth: getting it wrong there distorts everything else. The European principle of risk classification reasons on what could have happened and on what remained to prevent it. Two dimensions, then: the most plausible severity had the sequence continued, and the number of barriers still available at the moment it stopped. An occurrence with no material consequence at all, but whose last defence gave way and where only luck did the rest, is dealt with before a visible incident where two barriers worked normally.

Potential severity crossed with the remaining barriersA matrix of three columns and three rows. In columns, the barriers still available after the occurrence: several, one only, none. In rows, the potential severity: limited, serious, major. The cells run from simple recording at the bottom left to immediate treatment at the top right.majorseriouslimitedplanact fastimmediatemonitorplanact fastrecordmonitorplanseveralone onlynonepotential severityremaining barriers
It is not the damage sustained that classifies, it is what remained between the occurrence and the accident. A case with no consequence whose last barrier gave way comes before a visible incident that is still protected.

The trap. Classifying by the damage sustained is the costliest reflex in occurrence processing. It produces cases marked “without consequence”, therefore barely investigated, whose recurrence then appears in the authority’s trend analysis before it appears in yours. The question to ask is not “what happened?” but “what stopped it going any further?” — and if the answer is luck, the potential severity is maximal.

Closing for want of information means writing a finding about oneself

The wording recurs regularly in case files: “closed, the crew did not answer our requests”. It is understood by the reader as information about the operator, not about the crew. It says that the feedback loop does not work, that the request went out from a generic mailbox with no name and no deadline, that it arrived three weeks after the event, or that the staff have good reasons not to answer — in which case it is the just culture policy that is at stake, and the subject becomes distinctly heavier.

Decision sequence before the closure of a caseA decision diagram. The follow-up analysis passes through two successive questions: does the cause reach up to the level of the organisation, and then is the action dated and verifiable. Two positive answers lead to closure; a negative answer to either one leads back to a follow-up query or a finding.follow-upanalysiscause traced upto the system?action datedand verifiable?closurequery, then findingyesyesnono
Closure is not a date, it is the passing of two gates. A case that fails one of the two comes back, and it comes back with the record of the first answer.

If you close a case just because you didn’t get information, it looks like you didn’t try hard enough. You need to show what you did to get the information, what you couldn’t find out, and what you did based on the worst possible scenario.

A crew reports an engine vibration on a Fokker 100. The operator can’t find the maintenance records because the subcontractor didn’t respond. Instead of just saying “no answer, closed,” they write: “We requested maintenance records from the subcontractor but didn’t receive them. We assume the worst case (engine defect) and have added extra checks to the next maintenance visit.”

There is a correct way to conclude without having established everything, and it consists of four elements: what was looked for and by what means, what could not be established, the most unfavourable hypothesis retained by default, and the action taken on that hypothesis. A case that says “the origin of the disconnection could not be established; the hypothesis of an under-torqued fastening is retained; the check has been added to the daily inspection until further notice” is a closed case. A case that says “no answer, no further action” is an open case, even if it carries the mention “closed”.

The right reflex. Call the reporter back while the memory is fresh — a few days, not a few weeks —, through an identified person, with a precise question rather than a form; and send them the result of the analysis when it is closed, even briefly. The response rate to requests for further information follows exactly the feedback rate: a department that never gives anything back stops obtaining answers, and the discussions between practitioners confirm it sector by sector.

What to go and check in your own documentation

Three points in the safety management manual deserve a careful re-reading: they are rarely aligned with practice. The first is the closure criterion: many manuals describe who closes, never under what conditions. The second is the level of delegation: an occurrence classified at high potential severity should not be closeable by the person who conducted the analysis. The third is the interface with subcontractors and ground handling providers: does the contract provide for access to the facts, within what time, and what is done when the answer does not come?

A man sits at a desk in a dimly lit room, working on a laptop and writing in a notebook.
Thorough follow-up analysis requires diligent documentation and research.

A grey area remains, and it is better to know it than to discover it. When an occurrence is also the subject of a safety investigation conducted by the national investigation body, the operator is not exempted from its own analysis: the two processes coexist, with different purposes and with access to the evidence that is not the same. Waiting for the investigation report before acting is a classic mistake. In the same way, an occurrence on an aircraft operated under wet lease or within a multinational group presupposes knowing, before the occurrence, which entity reports, to which authority, and who keeps the audit trail.

For anyone preparing to take up a post in a flight safety department or a nominated person qualification, these elements are the very substance of the interview with the authority: one is rarely asked to recite a text, often to walk through a real case, from notification to effectiveness verification. Practising on three closed cases from one’s own organisation, asking which of them would withstand the question “and what if the crew had been another?”, teaches more than a re-reading of the safety management training programme.

This text describes a mechanism and orders of magnitude; it replaces nothing. The applicable regulation, the decisions of the Agency and the operator’s approved documentation — safety management manual, occurrence processing procedure, operations specifications — prevail over everything above.

Questions fréquentes

Does the acknowledgement of receipt of an occurrence report amount to closing the case?

No. The receipt only certifies that the file is readable, that it carries the mandatory fields and that an identifier has been assigned to it. It is generally issued automatically by the exchange gateway, before any human reading. Closure presupposes a follow-up analysis transmitted, a cause traced up to the level of the organisation and an action plan whose effectiveness has been verified.

From what date do the analysis deadlines run?

From the date of notification of the occurrence, not from the date of the occurrence itself. The European regulation on occurrence reporting expects preliminary results within the order of a month, then final results within the order of a quarter. A report that surfaces late internally does not shift these deadlines: the time lost before notification is taken out of the analysis time.

Is “failure to follow the procedure” an acceptable cause?

No, it is a restatement of the fact. An acceptable cause explains why the intended barrier did not intercept the deviation, and then which organisational decision left that barrier in that state. The test consists of mentally replacing the people involved with others, in the same conditions: if the occurrence recurs, the cause written down is not the right one.

Can a case be closed when the crew has not answered?

Yes, but not by writing it that way. The wording “no further action, no answer” designates the operator’s feedback loop, not the reporter. An acceptable closure states what was looked for and by what means, what could not be established, the most unfavourable hypothesis retained by default and the action taken on that hypothesis.

Can an occurrence with no consequence be classified as high risk?

Yes, and it is frequent. The European classification reasons on the most plausible severity had the sequence continued and on the number of barriers still available at the moment it stopped. An occurrence with no damage whose last defence gave way, only luck doing the rest, calls for priority treatment; classifying by the damage sustained produces barely investigated cases that reappear as a trend.

Is an awareness action enough as an action plan?

Rarely, and never without justification. Awareness-raising is the weakest rung in the hierarchy of actions, and its effect declines over time. If the higher rungs — elimination of the hazard, technical interlock, procedure change, training — have been set aside, the analysis must say why, and the action must come with an indicator reviewed on a date fixed in advance.

Commentaires

Leave a Reply

Découvrez l’écosystème FL330

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