A one-page note arrives via messaging a few weeks after an event occurred on the other side of the world, on an aircraft type you operate. It draws no conclusions: it indicates that a component was found in an unexpected state, that testing is ongoing, and it recommends an inspection before return to service. No airworthiness directive accompanies it, and no regulation mandates it. The question lands on the operations desk the same day: what do you do with information that its author states is not yet consolidated?
What comes out of an investigation before the final report
Safety investigations pursue a single objective: understand to prevent. It carries a constraint that the hurried reader forgets: it takes time. Sampling, laboratory tests, flight data analysis, coordination with a manufacturer, and collaboration with one or more states: between the event and the final report, more than a year typically elapses. Throughout this period, the affected fleet continues to fly in the configuration that led to the event, and the investigator is aware of this.
The process therefore includes interim releases of varying nature. The first factual updates publicly document what is observed, dated, and measured: no analysis, no root cause, no contributing factors. Their purpose is as much informative as defensive—a void is always filled, and it is often filled poorly. The progress report, published when the final report cannot be issued within the year, states where the investigation stands: what has been confirmed, what remains under test, and especially what has been ruled out. The safety recommendation, finally, can be issued at any stage, as soon as the investigator deems that action cannot wait for the investigation’s conclusion.
Les enquêtes de sécurité prennent du temps, mais elles publient des informations partielles en attendant. Certaines sont juste des faits, d’autres indiquent où en est l’enquête, et parfois elles demandent des actions. La plus importante est la recommandation urgente, qui peut arriver à n’importe quel moment.
Après un incident sur un Airbus A320, une première note signale une anomalie dans le système de carburant. Un mois plus tard, une recommandation urgente demande d’inspecter les vannes de carburant avant chaque vol. L’enquête finale, publiée 18 mois plus tard, confirme que le problème venait d’une mauvaise installation sur certains avions, mais pas sur tous.
The operational distinction is simple: the first two inform, the third demands. Confusing the three is the most common error in the operations control room—reacting to a progress report as if it were an order, or filing an urgent recommendation as mere situational awareness.
| Document | Time of issue | What it establishes | What it expects from the operator |
|---|---|---|---|
| Preliminary factual information | Shortly after the event, within a short deadline set by the international convention | Facts established: sequence, configuration, conditions. No cause | Reading and checking applicability to one’s own fleet |
| État d’avancement | When the final report cannot be published within the year, then at regular intervals | Status of the work: lines of inquiry open, lines ruled out, tests under way | Update of the internal analysis; review of measures already taken |
| Urgent safety recommendation | At any point in the investigation, with no predefined deadline | Nothing on causes: it states an action deemed necessary without delay | A written response from the designated addressee, including a reasoned refusal |
| Rapport final | On completion of the work | Findings, causes and contributing factors, measures taken during the investigation | Incorporation into documentation and lessons learned |
Why the investigator speaks before concluding
Two variables move in opposite directions. Certainty grows over time, in stages: a test result comes in, a hypothesis collapses, another is confirmed. The still-avoidable exposure, however, decreases mechanically—not because the risk diminishes, but because the affected flights have already occurred. Perfectly established information that appears only after the investigation is complete has no preventive value for what happened in the meantime.
The dilemma is real, and it has costs in both directions. Publishing an unconfirmed hypothesis too early sends an entire profession down the wrong path, triggers costly inspections, and, most importantly, erodes the credibility of the issuer: an alert refuted twice in a row breeds indifference the third time. Waiting means accepting that the same sequence could repeat during the investigation, with another crew.
The decision criterion is not the probability of being right, but the product of that probability by the cost of error in each direction. When the suspected mechanism can be verified by visual inspection or neutralized by a procedural restriction, a modest level of certainty is sufficient: making a mistake only costs an unnecessary inspection. When the proposed measure would ground a fleet, the threshold rises, and the investigator waits for a test result rather than a presumption. It is the same reasoning you will apply later, with your own costs.
The trap. A progress report often highlights a leading hypothesis, because it is the one consuming test resources. It is then treated internally as an established cause, then formalized in a service note, then cited in a briefing. If the test disproves it six months later, the measure remains in place, and no one remembers its original basis. An investigation track is not a contributing factor; it only becomes one in the final report’s conclusions.
What an urgent recommendation commits, and what it does not
A safety recommendation is addressed to a named recipient: an authority, an agency, a manufacturer, an operator, or a service provider. What it creates is an obligation to respond—acknowledge receipt, then indicate the follow-up actions taken within the deadlines set by regulations, with a motivated refusal being an acceptable response. What it does not create is an obligation to comply. It does not modify a minimum equipment list, a flight manual, or a maintenance program, and it does not ground any aircraft.
The path to enforceability follows a different process. The certification authority evaluates the content and may issue an airworthiness directive, which is binding; it may also choose a safety information bulletin, which is not. On the operational side, the competent authority has the power to impose an immediate measure on its operators. Each of these steps requires technical review, and thus time.
This leads to a practical consequence that many organizations discover too late: between the issuance of the recommendation and the possible publication of a binding text lies a window where nothing obliges you, but everything invites action. A second, more insidious consequence: a recommendation addressed to a manufacturer or an authority may directly concern you, and no one is tasked with forwarding it. Monitoring recommendations targeting you is insufficient; it is monitoring your aircraft types and operational areas that yields useful information.
Deciding with unconsolidated information
The question is not whether you believe the investigator. It breaks down into four steps, in this order. First, applicability: the described mechanism assumes a configuration, a modification status, a procedure, sometimes an environment. This is found in the technical description, never in the summary; a modification number or an engine variant is enough to include or exclude you.
Then, the presumed severity if the mechanism fully unfolds in your case, not that of the original event, which depended on specific circumstances. Next, the cost and reversibility of the measure: a visual inspection at the station can be rescinded with a stroke of the pen, an empor limit can also be rescinded but incurs daily costs, a material modification cannot be rescinded. Finally, the measure’s status regarding your documentation: a restriction reducing the approved envelope is generally the operator’s prerogative, while anything affecting the approved content follows the standard amendment process. Resolving this point before drafting the note avoids issuing a directive that should have been approved.
There remains a factor that matrices overlook: a conservative measure consumes a resource. An additional inspection at the station takes time from the turnaround, and thus from something else—the preparation of the next flight, rest, or an unprescribed but routinely performed check. Adding a barrier without specifying what it replaces often just shifts another one. The measure is never free, and its true cost is seen on the ground, not in the note.
The right reflex. Any measure taken on unconsolidated information must include four mandatory fields: the exact reference of the triggering document, the review date, the name of the person who will re-examine it, and the exit condition—what result, what publication, or what absence of occurrence allows lifting it. A measure without an exit condition is not conservative: it is a permanent modification in disguise.
Writing uncertainty without distorting it
An internal note beginning with « we suspect » produces one of two undesirable effects: overreaction, where everyone devises their own countermeasure, or filing without follow-up. The formulation that stands separates four blocks and never mixes them: what is established and by whom, what is suspected and at what stage, what is asked of the crew in observable terms, and what would change the directive. The last block is the one often omitted, and it is the one that makes the note credible: it shows that the decision is reversible and will be re-examined.
The loop must then close. A conservative measure taken on a presumption is lifted only with evidence, and that evidence, no one but your crews will produce. Therefore, request a targeted report, described in observable symptoms rather than in diagnosis: what is seen, heard, smelled, or read, under what conditions, and not « system anomaly. » Without this feedback, the review date arrives, and there is nothing to decide except the absence of news—which is not data.
The opposite temptation is to toughen the tone to ensure compliance. The gain is immediate, the loss deferred—the next note will not be read the same way, and reports will dry up, because no one voluntarily signals what seems already decided.
The conservative measure that never comes down
This is the most common and costly flaw. A restriction is imposed in an emergency, the investigation proceeds, the final report is published and rules out the suspected mechanism, or confirms it by pointing to another barrier. No one revisits the restriction, because no date ever brought it back to the agenda. Years later, a crew finds in their documentation a constraint whose origin is no longer traceable, is surprised by it, applies it poorly, then bypasses it—and also bypasses it the day it would have been useful.
The stacking produces a second effect: the measures end up contradicting each other. A directive issued in the meantime becomes incompatible with a later service bulletin or a revision of the manufacturer’s manual. Without a single registry, no one detects the contradiction before a crew finds themselves facing two irreconcilable texts, at a station, with a departure to make.
The trap. The registry of conservative measures almost always exists; what is missing is the review. An effective review does not read the measures one by one: it starts from the list of expired dates and requires, for each, one of the three only admissible outcomes—lifted, extended with a new date and justification, or integrated as a permanent rule through the amendment process. Silence is not an outcome, and yet it is the most frequent one.
What to check in your organization
Three checks can be done without a project or budget. The first: where does an urgent recommendation concerning you physically land? Often in a generic inbox checked during business hours by a single function. Test the path on a Friday evening with the on-call person, and see how long it takes for someone to decide. The second: does your monitoring cover recommendations addressed to others but concerning your aircraft types, engines, or providers? The third: pull the list of active conservative measures and count those whose review date has passed. This number is an indicator of the state of your management system, more revealing than a report rate.

Two gray areas remain. The actual follow-up of recommendations is rarely visible in detail: you know a response was given, rarely what it produced. And many instructive events never reach the threshold that triggers an independent investigation: for these, no early alert or report, and the only source remains your own report stream.
For those preparing for a qualification or a selection interview for a leadership position, knowing how to distinguish a recommendation from a directive, explaining why a conservative measure carries a review date, and describing what is done with incomplete information immediately elevates the discussion. These mechanisms are covered in the courses dedicated to flight safety, and are discussed case by case on the professional forums when an alert has been issued.
This article describes a mechanism; it creates no obligation: the applicable texts, the documents published by the relevant investigation authority, and the operator’s approved documentation prevail in all circumstances over what is written here.

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