Principle
Warrant: Foundation
Whoever pays for an assessment is whoever bears the cost of it being wrong.
Never the party whose favourable result is being purchased, and the reason is structural rather than ethical.
Principle
Warrant: Foundation
A fee is a standing instruction about what to optimise.
A fee is not only a transfer of money. It is a standing instruction about what the party receiving it should optimise, and a market supplies whatever the instruction asks for without anybody deciding to supply it. The question is therefore never whether an assessor is honest. It is what the money in the room is quietly asking for.
Two mechanisms make the point and neither needed a villain. A rating paid for by the party being rated is not a rating: when a generous answer wins the next mandate, honesty becomes commercially irrational exactly where the stakes are highest, and the signal degrades without a single person choosing to degrade it. An auditor that also sells advice to the client it audits carries the same defect in a different shape, because the standing incentive not to jeopardise the larger relationship decides the smaller one. Both failures are structural, which is why neither can be fixed by hiring better people.
So the payer is fixed by rule rather than by opportunity. Whoever pays for an assessment is whoever bears the cost of that assessment being wrong: an insurer pricing deployment risk, a main contractor accepting liability for a subcontracted crew, an asset owner underwriting a project. Never the employer or the worker whose favourable result is being purchased. That party wants calibration, because a flattering answer costs them directly and soon, and calibration is exactly the quantity this firm exists to produce.
A pooled record is trustworthy only because unflattering outcomes went into it, so access to it is conditioned on contributing to it, including the outcomes nobody enjoys reporting. That rule is at reciprocity of contribution. Who bears the cost when a clearance turns out to be wrong is named before the first incident, not after it, at liability allocation.
Whoever pays decides what gets measured. So the payer is decided first.
01 · The rule, and the defence of it
Whoever pays for the answer must be whoever carries the loss.
The rule is one sentence long. Whoever pays for an assessment must be whoever bears the cost of that assessment being wrong. It is written first because everything else here is downstream of it, and because it cannot be revised later without unpicking every judgement made before the revision.
The defence begins with what a payment actually is. A fee is not only a transfer of money, it is a standing instruction about what the party receiving it should optimise. Whatever the payer wants more of, the market supplies more of, without anybody involved deciding to supply it. The question is therefore never whether an assessor is honest. It is what the money in the room is quietly asking for.
When the payer carries the exposure, the instruction is unambiguous. That party wants the assessment to be accurate, because an assessment that flatters costs them directly and soon. Severity and generosity are equally unwelcome to them: one wastes capacity they have paid for, the other injures somebody or stops a project. What they want is calibration, which is precisely the quantity this firm exists to produce.
When the payer is the party being described, the instruction inverts. They are not buying a description. They are buying a result, and the more favourable the result the more it is worth to them. The assessment stops being a measurement and becomes a purchase, and no methodology applied on top of that structure recovers the measurement, because the methodology is not what was bought.
A fee is an instruction. The only safe instruction is accuracy.
- The payer
-
The party who loses money, capacity or safety if the assessment turns out to be wrong. They want the answer to be right, because it is a forecast about a cost they will personally meet.
Fixed here, and used in this sense throughout.
- The subject
-
The party the assessment is about: the employer whose workforce is described, the worker whose capability is recorded. Their interest is in the result, which is a different thing from the accuracy of the result.
The subject is never the payer. That is the whole rule, stated from the other end.
02 · Who that is, and who it is not
Warrant: Derivation
The list is short, and it is decided before the first engagement, not after the first dispute.
Applying the rule is not a matter of taste. In any industrial deployment there is a party who meets the cost of a bad clearance, and that party is identifiable in advance, in writing, before anything is measured. Naming them is the first act of the relationship, and the answer decides who may commission the work.
They price the exposure and then meet the claim. A flattering description of the workforce costs them directly, so what they want from the record is that it be right.
The liability transfers to them at signature. If the crew is not what it was described to be, the rework, the delay and the incident are theirs.
They carry the outcome for the life of the asset, which is longer than any contractor's presence on the site. Nobody has a stronger interest in an unflattering true answer.
An employer commissioning a description of their own workforce is buying a description of themselves. What is worth most to them is the best defensible version, which is a different object from the accurate one.
The same structure at individual scale. A record purchased by its own subject is one whose value to the buyer rises with its generosity, and a generous record is worth nothing to anybody relying on it.
The parties named here are counterparties who carry risk. The full set, and the vocabulary each of them uses, is set out in who queries it.
03 · When the subject selects the judge
A rating paid for by the party being rated is not a rating.
The first failure structure arises when the party being graded also chooses the grader, pays the grader, and stays free to take the next mandate elsewhere. Every element of that arrangement is ordinary commercial practice. Together they produce an outcome nobody in it selected.
Follow the money one cycle at a time. A grader with a reputation for severity is expensive to approach: the subject knows in advance that the answer is more likely to be unwelcome, so the subject approaches somebody else. A grader with a reputation for generosity is cheap to approach, for the same reason in reverse. Neither grader has done anything, and neither has been asked for anything improper. The severe one simply receives fewer enquiries, quarter after quarter.
Repeat that over enough cycles and the outcome stops being a matter of individual conduct at all. It is a selection effect. The market allocates work toward generosity as reliably as if generosity had been the specified requirement, and the graders who survive the process are the most accommodating ones, which is exactly the population least able to say something expensive when it is true.
The signal that comes out of that structure is not weak. It is worse than weak: it is confidently wrong, in a known direction, at the moment the people relying on it can least afford it.
- 01 The subject selects the judge The party being described chooses who describes them, from a field of willing candidates.
- 02 The subject pays the judge The fee travels from the party with an interest in the result to the party producing it.
- 03 The subject can go elsewhere next time Nothing binds the next mandate to the last judge. The relationship is renewed by choice, repeatedly.
- 04 Severity loses work, generosity wins it No individual decides this. It is what the sum of independent, reasonable choices produces.
- 05 The market selects for generosity Over enough cycles the result is indistinguishable from a system designed to produce flattering answers.
04 · When the judge also sells advice
Two contracts with one party. Only one of them is adversarial.
The second failure structure needs no shopping around at all. It arises whenever the party that examines a client also sells that client advice. Two contracts then exist with one counterparty. The first is small and adversarial: its function is to say what is true whether or not the client enjoys hearing it. The second is large and cooperative: its function is to help the client with something the client wants.
Nobody has to trade one against the other explicitly for the trade to happen. Judgement on the small contract is priced, silently and continuously, against the risk of losing the large one. A marginal call that could go either way goes the way that keeps the room warm. A finding raised loudly by a party with nothing else at stake is raised quietly by a party with a great deal else at stake. Neither of those is a lie. Both are the structure moving the answer.
In more than one major market the two functions have now been separated by law, so that the party examining a company may not also be the party advising it. That separation is worth reading for what it concedes. It does not test individuals for integrity or ask for better procedures. It removes the arrangement, which is a formal admission that the structure rather than the people was at fault, arrived at after the failure rather than before it.
The remedy chosen was not better people. It was the removal of the structure.
| The structure | What it selects for | Who bore the cost |
|---|---|---|
| The subject selects, pays and may replace the judge | Generosity, by market selection rather than by anyone's decision. Severity is priced out one mandate at a time. | Everybody downstream who relied on the grade, none of whom chose the grader or paid the fee. |
| The examiner also sells advice to the party examined | Silence on the small adversarial contract, to protect the large cooperative one. The trade is continuous and never stated. | Those relying on the examination, and in the end both parties to the arrangement. |
In both structures, the party who paid was never the party who lost.
05 · Structural, not ethical
No arrangement of good people survives a structure in which honesty is commercially self-defeating.
It would be more comfortable to read the two structures above as stories about character. They are not, and reading them that way is how the arrangement gets rebuilt by people certain it could not happen to them. Both mechanisms work perfectly well on conscientious professionals. Neither requires a single participant to decide to do anything wrong.
What they require is only that honesty cost something and that the cost fall on the party being honest. Once that condition holds, ordinary attrition does the rest. The severe grader receives fewer enquiries. The examiner who raises the awkward finding has a harder conversation about next year. Nobody is corrupted. Who is still in the room simply shifts, quarter by quarter, until the population that remains is the one the structure was always going to leave behind.
This is why the rule is written as a structural constraint and never as a statement of values. A values statement is a promise about how people will behave inside an arrangement. A structural constraint changes the arrangement, so that behaving well costs nothing and there is no attrition to survive. The first is an assurance. The second is a design, and only the second can be checked from outside by somebody who does not know or trust anybody here.
The same reasoning runs underneath every other rule this firm binds itself to: the claim dated when it leaves and revisited when the answer exists, in the calibration ledger; a stall in verified coverage published rather than smoothed, in the plateau principle; the loss allocated by a named rule before the first incident, in liability allocation. It is also the same absent feedback path that empties the advisory relationship, in advisory work is priced as though reasoning were scarce.
Integrity that depends on nobody being tested is a forecast, not a property.
06 · The one door left, and the rule that closes it
Warrant: Derivation
Advice sold back to the party assessed inherits the ledger's discipline, or it is not sold.
One route remains by which the old unaccountable shape could re-enter, and it is worth naming precisely because it looks harmless. An operator who has been assessed will frequently want to know what to do about the result. Answering that is useful work, the operator will pay for it, and the payment comes from the subject of the assessment rather than from the party carrying the risk. That is the door.
It is not closed by refusing the work. It is closed by a condition on the work. Where remediation advice is sold back to the party assessed, it inherits the discipline of the ledger in full: a specific intervention, named in advance, with a stated effect on a stated measurement, re-measured afterwards on a date fixed before the work begins. The intervention can therefore be shown to have failed, and if it failed the record says so.
What is refused is the alternative shape: the unfalsifiable strategic narrative, the recommendation composed so that no later observation could contradict it, the document that closes the engagement before the outcome it concerns has happened. It has no measurement attached, which is exactly what makes it comfortable to buy and impossible to grade.
Nothing in the assessment moves as a result of this work, which is the second half of the condition. The advice does not adjust the record, it is measured against it. The party paying for a remedy never becomes the party deciding whether the remedy worked, because that is the first failure structure wearing a helpful expression.
A remedy that cannot be shown to have failed is not a remedy. It is a narrative.
Principle
Warrant: Foundation
The money decides the answer long before the method does.
Method is visible and arguable, and it is the part everybody inspects. The payer is neither, and it sets the direction of every marginal call made under it. Choose the payer first, in writing, and the method is then allowed to matter.
The rule, as it binds
The party who pays is the party who loses.
Whoever pays for an assessment is whoever bears the cost of that assessment being wrong. It is a condition of the work rather than a preference about it, it is settled in writing before anything is measured, and it binds from the next engagement forward.
A rule a firm binds itself to is a stronger statement than an assurance about its own character, because an assurance costs nothing and a rule turns away revenue. This one turns away the easiest revenue available to a firm of this kind: the subject who would happily pay for a description of themselves. Declining it is the price of the record being worth anything to the parties who rely on it.
It does not stand alone. Access to the pooled record is conditioned on contributing real outcomes to it, including unflattering ones, in reciprocity of contribution. Every claim is dated when it leaves and revisited when the answer exists, in closing the loop. What is sold, and the architecture that prices it, is in what is sold. What is not yet known is stated rather than resolved by optimism, in open questions.
The cheapest revenue available is the revenue this rule refuses.