— SERVICES · INDEPENDENT MODEL VALIDATION
THE SECOND OPINION THAT HAS TO HOLD.
Before a board signs a number, an auditor tests it, or a regulator probes it, someone independent should have asked the hard questions first. We are that someone — the effective challenge that finds the problem while it is still yours to fix, not theirs to find.
— THE PROBLEM
THE TEAM THAT BUILT IT CANNOT BE THE TEAM THAT CLEARS IT.
A model built in-house is too often reviewed by people who report to the same executive, share the same assumptions, and have every incentive to see it pass. That is not a criticism of them — it is precisely why the standard requires validation to be independent of development, and why a reviewer who is not truly independent tends to confirm rather than challenge.
And the requirement is not occasional. An IRB bank must obtain the Prudential Authority’s prior written approval before it changes a credit-risk model; an insurer’s head of actuarial function must be independently peer-reviewed at least every three years under SAM. When the challenge is weak, the problem is not avoided — it is deferred to the worst possible moment: the exam, the audit, the deal, the loss. Independent validation moves that discovery earlier, to when it is still cheap to fix.
— WHAT WE DO
EFFECTIVE CHALLENGE, TO THE STANDARD.
We provide the independent review that satisfies regulatory expectations for effective challenge — assessing the credit, actuarial, and capital models your board, auditor, and regulator rely on, across their full lifecycle and against the recognised three-pillar validation standard.
Conceptual Soundness
The first pillar of validation: is the model theoretically right for its purpose? We assess methodology, assumptions, data, and design — whether an IFRS 9 PD curve, an IFRS 17 risk adjustment, or a capital model’s calibration genuinely captures the risk it represents, whether the assumptions are justified, and whether the developmental evidence supports the choices made. The questions a regulator asks first, asked first by us.
Independent Actuarial & Capital Review
The review the insurance regime specifically calls for: independent peer review of the head of actuarial function’s work, IFRS 17 measurement, and the appropriateness of SAM standard-formula or internal-model capital. A qualified, independent second opinion on technical provisions, the SCR, and economic capital — grounded in actuarial training, not applied from a checklist.
Outcomes Analysis & Back-Testing
The second pillar: does the model actually work? We back-test outputs against realised experience, benchmark against alternatives, and stress the model to find where it breaks. A model that has never been tested against reality is an assumption; outcomes analysis turns it into evidence — or exposes where the evidence isn’t there.
Third-Party & Vendor Model Validation
Buying a model does not transfer the risk. Vendor and third-party models must be brought inside your model-risk framework and validated to the same standard as in-house builds — even where components are proprietary. We assess what the vendor will not, so a bought model is one you can still defend as your own.
Ongoing Monitoring & Model Risk
The third pillar: is it still working? Models degrade as the world moves away from the data they were built on. We review monitoring, sensitivity, and override analysis, and assess whether the model remains fit for use — the continuous discipline the Prudential Authority increasingly expects, not a one-off tick before the deadline.
– HOW WE WORK
INDEPENDENCE IS NOT A POSTURE. IT IS THE POINT.
The value of a validation is exactly equal to how independent and rigorous the challenge is. Ours is built to be both — because a review that only confirms is worth nothing to the board that relied on it.
01
Establish scope and materiality
We tier the review to the model’s risk and purpose. A model driving capital or provisioning demands deeper scrutiny than a peripheral tool — validation effort is matched to consequence, not applied as a flat checklist.
02
Reproduce, don’t just read
Where it matters, we reproduce the model independently rather than accepting the developer’s account of it. Replicating the result is how you find the error the documentation didn’t mention.
03
Challenge across all three pillars
We assess conceptual soundness, outcomes, and ongoing performance together — because a model can be theoretically elegant and empirically wrong, or accurate today and fragile tomorrow. Validation that tests only one pillar isn’t validation.
04
Grade the findings
We produce formal, graded findings — what must be fixed, what should be, and what to watch — with clear recommendations. A validation that finds nothing is usually a validation that looked at nothing.
05
Deliver a report and file that stand up
The output is documentation an auditor, a board, and the Prudential Authority can rely on: independent, evidenced, and detailed enough for an expert to follow — written to survive the examination it exists to pre-empt.
— WHO THIS IS FOR
FOR THE PEOPLE WHO NEED TO BE CHALLENGED — BEFORE SOMEONE ELSE DOES.
This work is for heads of model risk who need genuine independent capacity, or a second line that is actually second; for CROs and heads of credit risk who need their IRB and ECL models challenged before the regulator does it for them; for heads of actuarial function and statutory actuaries who require the independent peer review the regime mandates; for CFOs and audit committees who must sign numbers they need independent assurance on; and for institutions running vendor models they have never truly validated.
It spans banking, insurance, mining and resources, retail credit, and project finance — where a model carries enough consequence that someone independent should have challenged it before it was trusted.
— THE ALETHEIA DIFFERENCE
WE ARE THE CHALLENGE YOU WANT TO FAIL — WHILE IT STILL COSTS NOTHING.
The point of independent validation is not to be told your model is fine. It is to have someone genuinely try to break it while breaking it is still cheap — before the auditor, the examiner, or the loss does it for you. A validation that flatters is worse than none, because it converts unexamined risk into false confidence.
Our independence is real: we did not build the model, we do not report to the team that did, and we have no incentive but to find what is actually there. Grounded in actuarial science, with the perspective of the audit chair and cross-sector depth across insurance, banking, mining, and credit, we bring effective challenge that meets the standard the regulator now expects — and the standard the number deserves. It does not flatter. It illuminates.
”
WHO HAS GENUINELY TRIED TO BREAK YOUR MOST IMPORTANT MODEL?
A 30-minute discovery call. No obligation. A scoped proposal within 48 hours.
ALETHEIA PARTNERS
Analytics. Modelling. Truth.
A specialist analytics and modelling consultancy, grounded in actuarial science. Based in Johannesburg, working with corporates, project sponsors and investors across Africa.
© 2026 Aletheia Partners (Pty) Ltd · All Rights Reserved
SPACE