Model Risk

Governance of expert judgement

How to make expert judgement in risk models auditable, consistent and challenged.

By Jonas Osman AbdelfourPublished July 20, 2025

Summary Expert judgement is used across risk models — in scenario design, overlays, low-default portfolios, and stress. Treating it as an unavoidable weakness misses the point; treating it as ungoverned is the actual weakness.

Where judgement is used Common locations include PD assignment in low-default portfolios, stress scenario severity, behavioural assumption setting, and post-model overlays.

Making judgement auditable Each judgement should be recorded with the reasoning, the range considered, the evidence weighed and the decision-maker. A one-line note is not documentation.

Consistency Judgements on comparable questions should be consistent across time and across assessors. Calibration exercises and second-line review support consistency.

Challenge Independent challenge — from validation, audit, or a designated challenge function — is the primary control. Where challenge produces material adjustments, the frequency of judgement review should increase.

Aggregation Individually reasonable judgements can compound into materially aggressive or conservative aggregate positions. Portfolio-level review of judgement impact is essential.

Limitations Judgement cannot be eliminated in areas where data is thin. It can be governed so that its role is transparent and its impact bounded.

Related expertise See [Model Risk and Validation](/expertise/model-risk).

Frequently asked questions

Where judgement is used?

Common locations include PD assignment in low-default portfolios, stress scenario severity, behavioural assumption setting, and post-model overlays.

What should risk leaders know about making judgement auditable?

Each judgement should be recorded with the reasoning, the range considered, the evidence weighed and the decision-maker. A one-line note is not documentation.

What should risk leaders know about consistency?

Judgements on comparable questions should be consistent across time and across assessors. Calibration exercises and second-line review support consistency.

What should risk leaders know about challenge?

Independent challenge — from validation, audit, or a designated challenge function — is the primary control. Where challenge produces material adjustments, the frequency of judgement review should increase.

What should risk leaders know about aggregation?

Individually reasonable judgements can compound into materially aggressive or conservative aggregate positions. Portfolio-level review of judgement impact is essential.