Residual risk is not something an organisation discovers. It is something a named person decides to live with, on a stated date, for a stated period. Everything that makes that decision defensible — the level of the signature, the evidence behind it, the review date attached to it — is procedural, cheap, and routinely absent. What an audit finds is almost never a badly judged risk. It is an acceptance nobody can attribute, justify, or date.
Acceptance is a decision, not an outcome
Safety risk management, in the framework that ICAO Annex 19 obliges states to require and that the European management system rules impose on operators and approved organisations, runs in a fixed order: identify the hazard, describe the consequence, assess the risk as it stands, apply mitigation, reassess, then decide what to do with what is left.
That last step is the one that gets skipped. In a great many risk registers the reassessment is the final column, and the file closes when the residual rating turns from red to amber. But an amber cell is not a decision. It is a measurement. The decision is a person saying: this organisation will operate with this level of exposure, knowingly, until a stated date, and I am the one who says so.
The distinction matters most when the risk materialises. An investigator, a regulator, or a court examining an accident will find the hazard in the register either way. What separates a considered organisational decision from an omission is whether somebody at the right level made a recorded choice, on evidence, with a review point — or whether the entry simply sat there going amber for four years.
Residual risk isn’t just a number in a spreadsheet. It’s a formal decision by a leader to accept a certain level of danger for a specific time. If no one signs off, it’s not a real decision—just a forgotten problem.
An airline assesses runway excursion risk on a short, wet runway. After adding a new braking procedure, the residual risk is rated “amber.” The flight operations manager signs off, stating the acceptance is valid until the next winter season, when runway conditions will be reassessed.
What is being accepted, exactly
Three different quantities travel under the same word, and mixing them corrupts the whole exercise.
The initial risk is the assessment made before the mitigation being proposed is in place — but with the defences that already exist. That last clause is where most registers go wrong: teams assess the raw hazard as if the organisation had no procedures, no training and no equipment at all, which produces a catastrophic initial rating and a flattering apparent improvement.
The residual risk is the assessment once the new mitigation is implemented and effective. Not designed. Not planned. Implemented, and demonstrated to work — which means the acceptance cannot honestly be signed on the day the action is written, only on the day the action is verified.
The substitute risk is the one nobody assesses: the new exposure created by the mitigation itself. An extra checklist item lengthens the taxi flow. A procedural restriction on one runway concentrates traffic on another. A new alerting function adds one more thing to inhibit. Doc 9859 is explicit that mitigations must be checked for the risks they introduce, and it is routinely the weakest paragraph in any risk assessment file.
The trap. An acceptance signed against a planned mitigation is an acceptance of the initial risk with optimism attached. If the action has a target date in the future, what is being accepted today is the exposure that exists today. Say so in the record, and set the review for the action’s due date rather than a year out.
You can’t accept a risk based on a future fix. The risk today is what matters, and the acceptance must reflect that—with a clear plan to check if the fix actually works.
A cargo airline plans to install new ground handling equipment to reduce ramp accidents. The safety manager drafts an acceptance, but the accountable manager insists it can’t be signed until the equipment is installed and tested—proving it actually reduces the risk.
The matrix nobody actually mandated
The five-by-five grid — severity from catastrophic to negligible, likelihood from frequent to extremely improbable, cells coloured red, amber and green — is so widespread that crews and managers alike treat it as a regulatory artefact. It is not. ICAO presents it as an example, and states it plainly: each organisation defines its own severity and likelihood scales, its own tolerability boundaries, and its own escalation thresholds, appropriate to its size and complexity.
Two consequences follow, and both are practical.
First, a matrix copied from another operator without recalibration will misclassify. A likelihood band expressed as « once per year » means something entirely different to a regional operator flying twenty thousand sectors a year and to a business aviation department flying four hundred. If the bands are not expressed in the organisation’s own exposure units — per sector, per flight hour, per movement, per maintenance event — the cells are decorative.
Second, the boundary between « acceptable with mitigation » and « unacceptable » is a policy choice the organisation must be able to explain. That boundary is where the signature ladder is anchored. An organisation that cannot say why its amber band ends where it does cannot justify who signs what.
Some regimes wrap this in the language of reducing risk as low as reasonably practicable — a concept from British safety law, adopted in aviation by the United Kingdom authority and by others, which requires the organisation to show that further reduction would be grossly disproportionate to the benefit. The European rules do not use that formula, and the FAA’s safety management requirements in 14 CFR Part 5 do not either; both require a defined risk acceptance process without prescribing a test. If an organisation invokes the reasonably-practicable argument, it has to be able to run it, which means documenting the options considered and rejected — not just the one adopted.
The red-amber-green matrix isn’t a rule—it’s just a tool. Each company must define what “acceptable” means for them, based on their own operations. If you can’t explain why a risk is “amber,” you can’t justify who should sign off on it.
A regional airline flying 10,000 flights per year defines “frequent” as 10 incidents per year, while a business jet operator with 500 flights per year defines it as 2. Their risk assessments must reflect their own exposure, not someone else’s rules.
The signature has to match the severity
The principle is simple and almost universally stated in some form: the higher the residual risk, the higher in the organisation the acceptance must be taken. The reasoning is not ceremonial. Accepting a significant risk usually means declining to spend money, or declining to reduce production, or both. Only somebody who controls those resources can genuinely make that choice; anyone below them is accepting a constraint, not a risk.
That is exactly why the accountable executive — the accountable manager, in the European vocabulary — is defined by their authority over financial and human resources rather than by their job title. A person who cannot authorise the expenditure that would remove a risk cannot meaningfully accept its retention.
| Residual risk after effective mitigation | Who may accept | Where it is recorded | Review |
|---|---|---|---|
| Intolerable | Nobody. The activity is stopped, restricted or redesigned | Decision and the restriction imposed | Not applicable — there is no acceptance to review |
| Top of the tolerable band | Accountable manager, in a safety review board | Minuted, named, dated, with the evidence referenced | Short and explicit; typically the shortest interval the organisation uses |
| Middle of the tolerable band | The responsible post holder for the domain — flight operations, ground operations, continuing airworthiness, training | Risk register entry carrying the post holder’s name | Periodic, at a stated interval |
| Broadly acceptable | Line management, or the process itself under a standing delegation | Register entry | On trigger only |
Two refinements are worth building into the ladder from the start. A delegation should be by post, not by person, so that it survives a departure — and the operations manual should say what happens when the post is vacant. And the ladder should be bidirectional: a residual risk that a post holder judges to sit above their delegation must go up, and the mechanism for pushing it up has to be as clearly written as the mechanism for signing it off.
Only the person with the authority to spend money or change operations can truly accept a high risk. If a manager signs off on a risk they can’t fix, it’s not a real acceptance—just a rubber stamp.
A maintenance organization assesses a high-risk tooling failure. The head of maintenance can’t sign off alone—the accountable manager must approve, because fixing the issue requires budget approval for new equipment.
The safety manager does not own the risk
This is the most common structural error in small and medium organisations, and it is usually invisible until an audit or an occurrence exposes it.
The safety manager runs the process. They collect reports, facilitate the assessment, maintain the register, prepare the material for the safety review board, monitor whether mitigations are actually implemented, and measure whether safety performance indicators move. They do not own the operational risk, and they do not command the resources to remove it. No rule spells out a prohibition on their signing in so many words; it follows from the architecture, which makes the safety manager the focal point for the system’s operation and leaves the exposure with the line. An acceptance signed by the process owner leaves nobody who both carried the risk and chose to keep it.
The board structure exists to enforce that separation. A strategic body, chaired by the accountable manager and attended by the post holders, sets policy, allocates resources and holds the acceptances at the top of the ladder. Tactical groups, one per operational domain, do the work: they build the assessments, implement the mitigations and report upward. The safety manager sits in both, as the person who makes the process run, and signs neither as owner.
The right reflex. Read your own risk register and look only at the name column. If the same name appears against most acceptances, and that name belongs to the safety department, the organisation has an administrative record rather than a decision trail — whatever the ratings say.
The safety manager coordinates the process, but they don’t decide what risks are acceptable—that’s the job of the leaders who run the operations. If only the safety team signs off, the company has no real risk acceptance system.
A small airline’s safety manager signs off on most risks in the register. During an audit, regulators note that none of the operational managers have approved anything—meaning the company has no real risk ownership.
The date without which an acceptance becomes policy
An acceptance with no review date does not expire. It hardens. Within a few rostering cycles the workaround is the way things are done, new joiners learn it as normal practice, and the conditions that justified it are no longer anybody’s memory. The risk has not been managed; it has been naturalised.
The fix is not sophisticated. Every acceptance carries two kinds of expiry, and needs both.
The first is a calendar date, chosen to mean something. A review « in twelve months » set by default is the same as no review, because the date carries no reasoning. A review set at the due date of the outstanding mitigation, at the end of the season whose traffic justified the acceptance, or before the next base opens, is a date somebody will defend.
The second is a set of triggers that reopen the file regardless of the calendar. Change management is already a required process in both the European management system rules and the ICAO framework; the point here is to connect it to the register, so that a change automatically surfaces the acceptances that assumed the old state of affairs.
| Trigger | Why the earlier acceptance may no longer hold |
|---|---|
| A recurrence of the occurrence the mitigation was meant to prevent | The mitigation is either not effective or not being applied |
| Fleet, type or configuration change | The consequence severity was assessed on a different aeroplane |
| New route, base, aerodrome or ground handling provider | Exposure and the defences available have both moved |
| Significant change in traffic volume | Likelihood bands are expressed per unit of exposure |
| Change of the post holder who accepted it | The delegation was to a post; the new holder has to own or escalate it |
| Regulatory change or a new airworthiness directive touching the hazard | The assumption that no further mitigation was required may have lapsed |
| Subcontracting or outsourcing of an activity in scope | Defences assessed as internal are now contractual |
A risk acceptance without a review date becomes permanent—and soon, nobody remembers why it was accepted in the first place. Every acceptance must have a clear end date and triggers to revisit it if conditions change.
An airline accepts a slightly higher risk of runway excursions during winter. The acceptance is valid until the end of the season, but if a new runway friction reporting method or a new de-icing product is introduced mid-season, the risk must be reassessed immediately.
What the record has to contain
An acceptance record is written for two readers who do not yet exist: the auditor who will read it in three years, and the investigator who may read it after an accident. Both will ask the same questions, and neither will accept a rating without a reason attached.
| Field | What it has to answer |
|---|---|
| Hazard and consequence scenario | What can go wrong, and what the worst credible outcome is — not a topic heading |
| Existing defences | What was already in place when the initial rating was assigned |
| Assessment before and after mitigation | Both ratings, with the scale and exposure unit used |
| Reasoning | Why the residual rating is what it is, and what was considered and rejected |
| Substitute risks | What the mitigation introduces |
| Evidence that the mitigation is effective | Verification, not a completed action item |
| Name and post of the accepter | That the level matches the organisation’s own escalation rule |
| Date of acceptance and review date | Both, with the reasoning behind the review interval |
| Monitoring indicator | What would tell the organisation the acceptance has gone stale |
The last field is the one most often missing and the most useful. An acceptance without an indicator can only be reviewed by opinion. An acceptance tied to a measurable — a report rate, an exceedance count, an on-time completion figure — can be reviewed by evidence, and its review can be automated into the safety performance report the organisation already produces.
A good risk acceptance record answers the key questions: What’s the risk? Why is it acceptable? Who approved it? And when will we check it again? Without these details, the record is just a piece of paper.
A maintenance organization accepts a slightly higher risk of delayed aircraft due to a new inspection procedure. The acceptance includes a monitoring indicator: if more than 2% of inspections exceed the scheduled time, the risk must be reassessed.
Where this meets reporting and just culture
Two systems intersect here and are frequently conflated. Occurrence reporting protects the person who reports; in the European scheme the reporting regulation establishes both the obligation and the protection regime. Risk acceptance carries no equivalent shield. It is a management decision, and it can be read by an auditor, an investigator or a court.

The practical consequence is that acceptance records must be written to be read by someone hostile, without becoming defensive prose. What survives scrutiny is specificity: the scenario stated in concrete terms, the options examined, the reason the chosen level of exposure was judged tolerable, and the date it was to be looked at again. What does not survive is a register full of green cells and empty name columns.
None of this requires a larger safety department. It requires four columns that most registers already have and rarely fill: who, on what evidence, until when, and measured by what. Organisations that fill them find that the ratings matter far less than they feared — and that the arguments they used to have about colours turn into arguments about dates, which are the ones worth having. It is half a day of work on an existing register, and it is the part of safety management training that transfers straight to the file on the desk, a theme running through the management system material on the FL330 training platform.

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