Alternative Means of Compliance in Certification: Demonstrating an Equivalent Level of Safety

Éprouvette composite serrée dans un bâti d'essai hydraulique avec jauges de déformation câblées

A design organisation proposes to justify a strength requirement by another route: the published means assumes a test that the part’s architecture makes unworkable. The file goes out and comes back with a question that has nothing technical about it — what is the equivalence measured against? Until that answer is written down, review does not begin, and the clock keeps running.

What the published means gives you, and what it does not impose

A certification specification reads on two levels of unequal force. The requirement itself enters the product’s certification basis: it is binding, and it is what the demonstration must address. The acceptable means of compliance that accompanies it binds no one: it describes a path the authority has announced in advance that it will accept. Following it gives a presumption of compliance; departing from it remains lawful, provided it is demonstrated another way.

The comfort gap between the two situations explains most of the difficulty. With the published means, the discussion is about execution: representative rig, sufficient instrumentation, conditions respected. Without it, the discussion is first about the demonstration itself — its logic, its domain of validity, its coverage — and only afterwards about the results. The bar does not rise, the subject changes, and what used to be settled between specialists of the same test now moves up a level.

A mechanical effect follows from this, often discovered too late. The authority calibrates its involvement to the novelty and risk of what it is being offered: an unusual means raises that level on the item concerned — in documents requested, in hold points, sometimes in attendance at the test. The cost of an alternative means always includes the cost of the oversight it triggers.

The published means is a pre-approved path to compliance. If you take a different path, you must prove it achieves the same safety level. This takes more work and scrutiny than just following the published method.

Four gates, and the routing error that costs the most

Everyday language files under a single word situations that review treats very differently. Two questions are enough to tell them apart: does the specification have anything to say about the design, and is the letter of the requirement met?

Routing between the four compliance pathsFour questions asked from top to bottom, each with its own side exit: specification silent on the case, special condition; published means suitable, ordinary path; letter of the requirement met, alternative means; deviation compensated and demonstrated, equivalent level of safety. Otherwise the branch runs down to the deviation.Does the specification address the case?Does the published means fit?Is the letter of the requirement met?Is the deviation compensated and demonstrated?noyesyesyesnospecial conditionthe requirement is yet to be writtenordinary pathpresumption of compliancealternative meansdifferent path, same proofequivalent leveldeviation compensated and demonstrateddeviationdecided at another level
Two questions are enough to pick the path, and picking the wrong one costs more than the demonstration itself. A file built for an alternative means does not have the content an equivalent-level-of-safety finding requires: once reclassified, it starts over from its framework.

The most costly routing error is to present as a simple alternative means what is in fact a request for an equivalent level of safety. The frameworks differ: one establishes that another path reaches the same point, the other that a different point is worth as much as the first — which requires naming and measuring a compensation. Reclassifying is not asking for an addition: it sends the demonstration back to be rebuilt.

Four paths when the published means does not fit: what each one requires
Path Situation What must be established What remains on the product
Alternative means of compliance The requirement is met, the published path does not apply That another path establishes the same compliance and covers the same cases Nothing, beyond the scope of the accepted means
Equivalent level of safety The letter of the requirement is not met That identified compensating factors bring it back to the intended level of safety An entry on the data sheet, often a limitation or a task
Special condition The specification does not address the design That a criterion written for the occasion is appropriate, then that it is met The condition itself, added to the certification basis
Deviation The gap remains without demonstrated compensation That a rationale justifies it and that the residual risk is controlled An entry with conditions attached, decided at another level

An Airbus A350 wing design team proposes a new method to test fatigue resistance because the published means require a test rig that would damage the wing structure. The authority asks if this is an alternative means (same safety level, different method) or an equivalent level of safety (different safety level, but acceptable). The team must prove their method covers all critical cases the published test would have covered, plus any additional compensations needed for the differences.

Equivalence is measured against the intent, not against the text

The text of a requirement is the trace of an intent, and it is the intent that an equivalence demonstration must reach. It is reconstructed from three sources: the explanatory note published with the amendment that introduced the requirement, which states what was meant to be prevented; the associated acceptable means of compliance, which betrays the feared scenario by the very way it measures it; and the neighbouring requirements, whose content negatively delimits what this one alone covers.

The symmetrical trap is to copy out the requirement and present it as its own intent. « The intent is that the structure carries the load without failure » is not an intent, it is the text, and the demonstration becomes circular. A usable intent names the hazard, the conditions under which it arises, and the credit given — or refused — to the crew; it excludes things, that is how you recognise it.

One stubborn confusion remains. The published means is not the safety objective, and nothing requires matching it everywhere; but it does embody the level the authority has already judged acceptable, and so is the de facto benchmark. The demonstration almost always goes through case-by-case comparison with what the published path would have caught.

The right reflex. Before writing a single line of the file, produce two one-page documents: the intent of the requirement in one sentence, and the list of cases the published means actually covers — including those it catches without naming them, failure combinations or domain extremes. These two pages fix the scope, and every question from the reviewer will refer back to them.

Anatomy of an equivalence demonstration

An acceptable file is almost always organised the same way, whatever the subject. The following items, in this order.

A group of engineers in a lab examine a mechanical component on a test rig, with one using a handheld device to measure it.
Engineers verify safety compliance through rigorous testing and measurement.
  1. The requirement targeted and its intent, stated apart from the text.
  2. Why the published means does not fit: a verifiable fact — geometry outside the limits of the described test, material falling outside the assumed model, architecture with no accessible measurement point. Neither schedule nor budget is admissible.
  3. The proposed means, described in enough detail for a third party to redo it: rig, method, assumptions, uncertainties.
  4. The coverage map: case by case, what the published means would have covered and what the proposed means covers. This is the item that gets forgotten, and the one on which files get rejected.
  5. The compensating factors, each named and tied to an uncovered case. A factor that compensates nothing specific is just an argument.
  6. The acceptance criterion, stated before the results and agreed with the authority. A threshold written after the fact no longer demonstrates anything.
  7. The scope: exact configuration, domain covered, and the explicit list of changes that would reopen the finding.
  8. What flows down to the aircraft: a flight manual limitation, a maintenance task at a mandated interval, an installation constraint. A residual gap that appears nowhere has not been addressed, it has been lost.
Coverage compared between the published means and the proposed meansA frame represents the set of cases targeted by the requirement. Two disks overlap partially within it. The left disk, outlined with a solid line, is the coverage of the published means; the right one, outlined with a dashed line, is that of the proposed means. The left portion the second disk does not cover represents the cases left aside: these are what must be explicitly compensated. The right portion that extends beyond represents excess coverage, with no exchange value. In the centre, the shared zone is covered both ways.cases targeted by the requirementpublished meansproposed meansnot coveredto compensateexcesscoveragecovered both waysexcess on one sidedoes not buythe shortfall on the other
Equivalence is judged on the crescent the proposed means leaves uncovered. What it covers beyond that elsewhere does not enter the balance: safety objectives are not traded against each other. This is the case-by-case map the reviewer asks for.

The coverage map harbours the most common reasoning error. A proposed means almost always covers cases the published means ignored: a numerical model explores combinations no test plan would have funded. This abundance does not trade against the missing case. Covering the everyday situation ten times better does not compensate for the rare case — precisely the one the requirement was written for.

Just because your new method covers more cases than the old one doesn’t mean it’s acceptable. You still need to specifically address the critical cases that the original test would have caught, even if your method covers many more things.

The compensating factor: what counts as one, and what does not

A compensating factor is admissible only under three simultaneous conditions: it is under the applicant’s control, it is made mandatory by a document that subsequently binds the operator, and it is not already credited elsewhere.

The first rules out anything that relies on a third party’s behaviour. The second restricts the vehicles to a short list: a design feature, which cannot be removed without an approved modification; an approved flight manual limitation; an airworthiness limitation or a task mandated by the instructions for continued airworthiness; required equipment. A habit, a training practice, a procedure the type certificate holder cannot impose: none of that holds up.

The third is the most discreet and the most fatal. A redundancy already credited in the safety assessment for another failure condition cannot be presented again here: the margin has already been spent. A reviewer who knows the product makes that inventory, and early; a file that counts the same device twice loses its credibility on everything else.

The trap. The compensating factor borrowed from someone else. An operating procedure, a training instruction, a check assumed to be done on the ramp: none of that is enforceable in a design demonstration, for lack of any means to impose its maintenance. If the compensation must live with the operator, it only holds through an approved limitation — and the aircraft then carries that constraint for its entire operational life.

What a reviewer looks for and does not find

Rejections cluster around a small number of grounds, all spottable before submission with a methodical re-read.

Common grounds for rejecting an alternative-means file, and what makes it recoverable
What gets rejected What the reviewer was looking for What makes the file recoverable
Incomplete coverage The list of cases from the published means, checked off one by one Address the missing case, or assign it a named compensating factor
Intent copied from the text A statement that excludes something Reconstruct the intent from the requirement’s supporting documents
Compensation outside the applicant’s control The document that makes the factor mandatory Turn the procedure into an approved limitation, or change the compensation
Margin already counted elsewhere The inventory of credits already taken on the same device Find a compensation independent of those already used
Criterion stated after the fact A criterion set before the results, and agreed Redo the demonstration against a criterion agreed in advance
Scope missing The configuration and domain explicitly bounded Set bounds, even if it narrows the scope of the finding obtained
Unvalidated method The domain where the model has met test data Add the calibration tests, or narrow the domain claimed

One last ground appears in no checklist, and yet it weighs heavily: the order of arrival. Proposed after the published means has failed, an alternative means arrives carrying an awkward piece of information — the design does not pass the ordinary path. It no longer reads as a technical choice but as a catch-up.

Timing decides as much as content

The certification programme fixes, requirement by requirement, the means selected, and it is agreed before the demonstration begins. That is where an alternative means belongs, still one option among others, and where the authority sets its level of involvement. Accepted at that stage, it commits both parties; proposed later, it commits no one and gets negotiated with no schedule reserve left.

Two schedules depending on when the alternative means is proposedTwo horizontal bars on the same time axis, without a numeric scale. The top one, solid line, corresponds to the alternative means carried in the programme from the start: scoping, demonstration, finding, and an early finish. The bottom one, dashed line, corresponds to the proposal made after the published means fails: path abandoned, late scoping, re-demonstration, finding, and a finish much further out. The lengthening does not come from the review but from the demonstration redone.carried in the programme from the startproposed after the published means failsscopingdemonstrationfindingpath abandonedlate scopingre-demonstrationfindingwhat costs is not the review,it is the demonstration redonetime
Proposed late, an alternative means does not cost a review delay: it costs an entire demonstration. The published path already committed to is lost, scoping is redone under pressure, and the authority reviews knowing the design has not passed the ordinary path. Qualitative pace, not to scale.

Hence a counter-intuitive consequence: the right time to propose an alternative means comes before it is certain one is needed. Spotting early the requirements where the published means fits the architecture poorly costs a few meetings; discovering it on a broken specimen costs the specimen, the campaign, and the trust built up on the rest of the file.

A Boeing 787 team identifies early that the published means for testing lightning strike protection on composite materials won’t work with their new design. They propose an alternative method in the certification plan. The authority agrees to the approach, and the team completes the testing on schedule. If they had waited until the published test failed, they would have had to redo the entire test campaign with the new method, causing significant delays.

What the finding leaves behind

An equivalent-level-of-safety finding is not an event, it is a permanent item. It enters the product’s certification basis and shows up in plain text on the Type Certificate Data Sheet — a public document read by foreign authorities, buyers and maintenance organisations. What was negotiated once gets reread for decades.

Three consequences follow. The finding bounds future changes: any evolution touching the compensating factor reopens it, however minor otherwise, and this is often how a design office discovers it exists. It does not travel: obtained by a different applicant or before a different authority, it confers no acquired right, and the cross-validation provided for under bilateral agreements bears precisely on the points where judgement can differ. It can finally expire: when a specification amendment absorbs the case, the gap disappears for subsequent applications — not for the product already certified, whose basis stays frozen at its date.

What to go check, and what remains debated

On a type one operates or maintains, the list of special conditions, equivalent levels of safety and deviations carried on the Type Certificate Data Sheet maps out the places where the demonstration was not standard. Almost every one of these lines has a counterpart elsewhere: a flight manual limitation, a task at a mandated interval, equipment required from the outset. Cross-checking the two lists explains a surprising number of constraints reputed to be arbitrary.

For anyone preparing a type rating or a certifying function, the exercise pays off directly. Knowing that a limitation traces back to an equivalence finding changes how it is treated: it is not a comfort margin left by the manufacturer, it carries a gap, and it is not up for discussion in flight. This is what separates reading a manual from reading a file, as worked through in the FL330 training programmes.

Two areas remain grey. The first: when the published means is so detailed that it effectively stands in for the requirement, departing from it is treated as an equivalence even though the letter of the text is met, and the boundary depends on the interlocutor as much as on the file. The second: a judgement made on a set of compensations travels poorly from one authority to another, and practitioners report divergent outcomes on comparable files, as discussed on the FL330 forum.

The applicable text, the certification basis fixed for the product concerned, and the approved documentation prevail over this article, which describes a mechanism and not an enforceable procedure.

Questions fréquentes

Must an alternative means of compliance be accepted by the authority, or is it enough to declare it?

It must be accepted before it is used. The published means benefits from an acceptance announced in advance; any other path must be presented, reviewed and agreed, normally when the certification programme is fixed. An applicant who demonstrates first and proposes afterwards risks having the method rejected after the fact, in which case the entire campaign is lost.

What is the practical difference between an alternative means and an equivalent level of safety?

In the first case the letter of the requirement is met and only the demonstration path changes: it must be proven that this path covers the same cases. In the second, the letter is not met and it must be established that identified compensating factors bring it back to the intended level of safety. The two files do not have the same content, and a file reclassified from one to the other must be rebuilt, not merely completed.

Can a finding obtained by another manufacturer serve as a precedent?

It usefully informs on what the authority has judged acceptable and on the expected format, but it confers no right. Each finding is tied to a design, a scope and a set of compensating factors specific to it. Citing it without demonstrating that the configuration presented is comparable point by point weakens the file rather than strengthening it.

Is a flight manual limitation enough as a compensating factor?

It constitutes an admissible vehicle, because it is approved and subsequently binds the operation. It still needs to be tied to a specific uncovered case, be verifiable, and not duplicate a compensation already credited elsewhere. A vague limitation, or one that assumes crew vigilance with no means of guaranteeing it, compensates nothing.

What happens to an equivalence finding when the specification is amended?

For applications made after the amendment, the case may be absorbed by the new text and the gap disappears. For the product already certified, nothing changes automatically: its certification basis remains the one fixed at its date, and the finding continues to appear on the data sheet along with the limitations it generated. A deliberate step by the holder is required to align the product with the new text.

Commentaires

Leave a Reply

Découvrez l’écosystème FL330

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