Out-of-Date Hazard Log: Diagnosis, Removal and Reactivation

A professional reviews a cluttered hazard register in a dimly lit aviation safety office, surrounded by stacks of reports and a faded risk assessment chart.

Most hazard registers do not fail noisily. They inflate, line by line, until they contain six hundred entries that no one reads, each marked with a color that no one questions, owned by a safety office that cannot act on any of them. The register continues to exist, is always presented to the auditor, and long ago quietly ceased to inform a single decision. Putting it back into service is less a matter of better software than of a distinction that is lost early: a hazard is a condition, and the occurrence that revealed it is only evidence.

What the register is intended to decide

Safety risk management, within the framework established by ICAO Annex 19 and developed by Doc 9859, exists to allocate limited attention and resources. Hazard identification feeds risk assessment; risk assessment feeds a decision on controls; safety then ensures whether these controls do what was claimed. The register is the working document that moves a hazard from the first of these stages to the last.

There is one decisive test for any line in it. If the assessment of this line changed — if it went from tolerable to intolerable from one day to the next — would anyone do anything different? If the honest answer is no, the line does not do the work of a hazard entry. It may remain as a useful record of something, but it belongs elsewhere.

Applied to an inflated register, this test disqualifies most of the contents in an afternoon. It is not a failure of the register; it is the diagnosis.

A hazard register should only list things that, if they change, would make someone act differently. If a line doesn’t trigger action, it doesn’t belong there.

The confusion at the root

Four different objects end up in the same table, and once they are mixed, the register cannot be reviewed, because each type requires a different question to be asked of it.

Four objects that are not equal, and where each one belongs
Object Definition in use Example Belongs to
Hazard A condition or an object with the potential to cause harm, present whether or not something has occurred Ground handling by a subcontractor whose staff has high turnover, at a station operated only part of the year The hazard register
Occurrence A specific event, on a date, on an aircraft, that has occurred A ground damage event at a cargo door on February 3 The occurrence database, linked to the hazard as evidence
Consequence A result that the hazard could produce Undetected structural damage released in service A field of the hazard entry, not a line in itself
Control failure An absent, weak, or unverified barrier Absence of independent damage inspection after handling by the subcontractor The tracking of corrective actions, referenced from the hazard entry

Two symptoms indicate that the confusion is present. The first is a register in which the lines begin with a past tense verb — it is a register of events. The second is a register in which the lines name results: runway excursion, loss of separation, in-flight fire. These are consequences, and a register of consequences has the particular property of being identical for every operator in the world, which is a good sign that it says nothing about theirs.

The trap of one line at a time. Linking the reporting system directly to the register seems rigorous and produces a register. Twenty reports on de-icing at the same station are a hazard with twenty pieces of evidence, not twenty hazards. Once the register grows faster than the operation changes, it has become a database of the past rather than a description of the present.

An airline with 500 hazard entries finds that 400 are past events (e.g., “Cargo door damaged on Feb 3”). After cleanup, only 100 remain as actual hazards (e.g., “High subcontractor turnover at seasonal stations”). The register is now actionable.

How a register dies

Degradation is rarely a single failure. It is an accumulation of small failures, each individually reasonable, and they are recognizable.

Symptoms, cause, and the repair
What you see What is really happening Repair
The register simply keeps growing Absence of exit criteria: the condition under which a line is closed has never been defined Giving each line a closing condition when it is opened
Almost everything is amber Uncalibrated scales, and a culture of assessment that reads green as complacency and red as an admission Defining severity by a described worst credible outcome and probability by an explicit exposure base
The owners are all in the safety office The department that facilitates risk management was made to own the operational risk that it cannot control Assigning each line to the line manager with the budget and authority to change the condition
The controls read « training » and « procedure » Paper barriers, credited at full strength, never verified Recording the evidence that each control works and its date
Only the initial risk is recorded The register was built as a one-time assessment, not as a loop Carrying the initial risk, the controls, the residual risk, and the review date in each line
New operations do not produce new lines Change management is disconnected from hazard identification Making every change assessment an input that opens register entries by default
The review dates are all in the past The pace was uniform and was not achievable Setting the pace according to the level of risk, plus event-driven triggers

Behind several of these is a single organizational fact: the register is kept for the audit rather than for the meeting. A document produced to be shown is optimized for completeness. A document produced to decide is optimized to be short enough to be read.

Why everything ends up in amber

The risk matrix deserves its own paragraph, because it is both the most used and the most misused tool in the kit. Two structural problems are repeated.

The first is the arithmetic that is not arithmetic. The severity and probability scales are ordinal — the categories are ranked, not measured — and combining two ordinal ranges into a product or score implies a spacing between categories that no one has defined. The matrix is a communication and classification device. Treating its cell as a calculated quantity, then comparing scores between unrelated hazards, gives a false authority to the classification that it does not have.

The second is the uncalibrated language. « Occasional, » « remote, » « unlikely » mean what the evaluator’s experience makes them mean, and the same hazard rated by two departments will land in different cells. The solution is not a finer matrix, but a defined denominator: probability expressed by sector, by movement, by component-year, by season — something countable, even if the count is an estimate. The same applies to severity: describing the worst credible outcome in a phrase, then placing it, instead of choosing an adjective directly.

The slide to the center of the matrix then follows human incentives. Green seems negligent, red demands action and money, amber demands neither and harms no one. A register where four-fifths of the lines are amber is not a portrait of the operation; it is a portrait of the rating process.

The cleanup: separation of three piles

Rehabilitation begins with a copy. Freeze the existing register, archive it with a date, and work on a duplicate — partly to lose nothing, largely to have people ready to delete.

Then classify each existing line into one of the three piles, and do it quickly, without reassessing anything for now.

  • Events. Anything that describes something that has occurred. These go to the occurrence database, where they should already be, and will return later as evidence attached to a hazard.
  • Control failures and findings. Anything drafted as an absence — no procedure, no training, no verification. These go to corrective action tracking with an owner and a date. These are real tasks; they are not hazards.
  • Conditions. Anything that describes a state of the operation that exists independently of any individual event. This pile is the future register, and it will be much smaller than expected.

Then merge into the third pile. Group the surviving conditions into families that share the same underlying state, and keep one line per family with the variants recorded within. A register that goes from several hundred lines to a few dozen has not lost information; it has stopped storing the same information multiple times.

There is no regulatory figure for how large a live register should be. The practical test is different and more demanding: the responsible manager should be able to keep the list of significant hazards in mind. Practitioners describe manageable registers in tens of entries rather than hundreds, and a register that cannot be read in a single session will not be read at all.

Rewriting the hazard statement

Most of the value of the reconstruction comes from a discipline in language, because a poorly drafted hazard cannot be assessed, assigned, or closed.

A usable statement has three parts: the condition, the operational context that makes it important, and the mechanism by which it produces harm. « Fatigue » is not a hazard statement; it is a topic. « Accumulated fatigue among short-haul crews on early start schedules during summer time, leading to degraded vigilance during the approach phase » is a statement that can be assigned, measured, and eventually closed, because each clause in it is something the operation can act on.

Three tests are applied to each rewritten line. Is it a state rather than an event? Would it remain true if no report had been submitted? Does it name a context specific enough that a different operator would write it differently? A statement that fails the third test is generic, and generic entries are exactly those that never generate action.

The reflex worth adopting. When a report arrives, ask what condition made the event possible rather than what went wrong. The event gives you a date and an aircraft; the condition gives you something to manage. Archive the event as evidence and check whether the condition is already a line before opening a new one.

Ownership, residual risk, and the review that closes

An entry that no line manager owns will not move, because the safety office can analyze a hazard and cannot allocate resources for its treatment. Ownership belongs where the budget and authority are: flight operations, engineering, ground operations, the training department. The safety function facilitates the assessment, maintains the register, and challenges the evidence — a distinction that ICAO safety management material draws clearly and that many organizations blur in practice.

A man in a high-visibility vest reviews a risk register on a clipboard in an aircraft hangar.
Risk registers must be regularly updated to remain effective.

Each live line then needs to carry a small and fixed set of fields, and the discipline is that a line with an empty field is not a line.

What a live entry must carry
Field What good looks like What kills the entry
Hazard statement A condition, in a declared context, with a mechanism A topic word, or an event in the past
Considered consequences The worst credible outcome, described A list of all conceivable outcomes
Existing controls Named barriers, with the evidence that each one works and the date of this evidence « Training and procedures »
Initial and residual risk The two recorded on a declared basis, with the difference explained by the controls A score, without a date
Owner A named position outside the safety office A department, or the safety manager by default
Review trigger and date A pace set by the level of risk, plus events that force an early review An annual date shared by each line
Closing condition The declared circumstance under which this line exits the register Nothing — therefore, it never goes

Residual risk is the field that most frequently is missing and the one that makes the register a management tool. Conventions differ in what the first score means: some organizations do not credit any control, others assess the condition with the existing barriers in place and before any new mitigation. A register that never declares which convention it uses cannot be compared with itself year over year. What is not optional is the second figure. Residual risk describes the operation as it is, and it is the only one against which an honest decision to accept, mitigate, or stop can be taken.

Reconnecting the entries

A register fed solely by voluntary reports sees what people choose to tell it, which is a fraction of the picture and a biased picture. A rehabilitated register should have several inputs, each with a defined path.

  • Occurrence and voluntary reports — the traditional source, now feeding evidence rather than lines.
  • Flight data monitoring — exceedance trends that identify conditions that no one has reported, and, equally useful, evidence that a control works.
  • Audit and inspection findings, internal and external, including those of the authority — findings are often control failures, and a group of them often points to an unlisted hazard.
  • Change management — a new destination, a new type, a fleet transfer, a subcontractor change, a schedule change. It is the only input that identifies hazards before they produce evidence.
  • Subcontractor and supplier data — ground handling, maintenance, training providers. Hazards do not stop at the organizational boundary, and the register is where subcontracted risk becomes visible.
  • Line observation and crew feedback — the source that captures conditions that people have normalized and no longer report.

The output counts as much as the input. A register that never feeds back to the line produces fewer reports each year, because people stop seeing the point. A periodic short summary of what has changed, what has been closed, and what has been funded is the safety promotion that does its job, and it is the cheapest intervention available for the reporting rate. The discussion of how operators actually structure this feedback is a recurring thread in the FL330 forums.

What an inspector really looks at

Organizations that prepare for oversight tend to polish the wrong thing. The size of a register impresses no one; the evidence that a process works is what is examined.

Expect the questions to focus on movement. Show me a hazard identified this year and how it got here. Show me a hazard that has been closed, and why. Who is the owner of this line, and what do they do? What has changed since the last safety review meeting? How does a subcontractor change become an entry? What tells you that this control still works? A register with fifty live entries, dated, with an owner and in motion, answers these questions. A register with six hundred static entries cannot answer any of them, and the volume itself becomes the finding.

Therefore, rehabilitation is not a documentation exercise. It restores the only property that makes the register worth maintaining: someone, somewhere, does something different as a result of what is written in it.

Commentaires

Leave a Reply

Découvrez l’écosystème FL330

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