How does the Red/Amber/Green model validation framework work?

Question

AcadiFi · Accepted Answer

The Red/Amber/Green (RAG) framework is a tiered classification system that model validation teams use to communicate the severity of findings and drive remediation actions. It provides a common language between model developers, validators, risk management, and senior leadership.

**The Three Tiers**

| Rating | Meaning | Impact | Remediation Timeline |
|--------|---------|--------|---------------------|
| **Red** | Critical deficiency | Model may produce materially incorrect results; significant capital or P&L misstatement risk | Immediate action required (30–90 days); model use may be restricted |
| **Amber** | Significant concern | Model has notable weaknesses but results are broadly acceptable; compensating controls may mitigate | Action required within 6–12 months; enhanced monitoring |
| **Green** | Minor observation | Best-practice improvement; no material impact on results | Address in next model update cycle; informational |

**Practical Examples**

Consider Blackmere Bank's annual validation of their Internal Ratings-Based (IRB) credit model:

**Red Finding Example:**
"The model's PD calibration uses a data sample that excludes the 2020 pandemic period, resulting in PD estimates that are 35% lower than out-of-sample actuals. This understates risk-weighted assets by approximately $800 million."
- **Why red:** Material capital understatement, regulatory non-compliance risk.
- **Action:** Immediate PD recalibration with extended data; possible management overlay until fixed.

**Amber Finding Example:**
"The LGD model uses a linear regression that does not capture the non-linear relationship between collateral values and recovery rates at low LTV ratios. Backtesting shows a 12% deviation in the 0–30% LTV bucket."
- **Why amber:** Affects a specific segment, not the entire portfolio; partial impact on capital.
- **Action:** Develop non-linear model enhancement within 9 months; apply conservative overlay to affected segment.

**Green Finding Example:**
"Model documentation does not include a sensitivity analysis for the choice of correlation matrix estimation window."
- **Why green:** Documentation gap; does not affect model output or capital.
- **Action:** Update documentation in the next annual review.

**Governance Workflow**

```mermaid
flowchart TD
    A[Model Validation Completed] --> B[Findings Classified RAG]
    B --> C{Red Findings?}
    C -->|Yes| D[Escalate to CRO & Board Risk Committee]
    D --> E[Restrict model use / Apply overlay]
    D --> F[30-90 day remediation plan]
    C -->|No| G{Amber Findings?}
    G -->|Yes| H[Report to Model Risk Committee]
    H --> I[6-12 month remediation plan]
    H --> J[Enhanced monitoring]
    G -->|No| K[Green: Track in findings log]
    K --> L[Address in next model cycle]
```

**Exam Tip:** The FRM exam may give you a model validation finding and ask you to classify it. Focus on **materiality** (does it affect capital, P&L, or risk limits?) and **immediacy** (is there a current incorrect output vs. a theoretical weakness?).

For more model risk management content, visit our FRM Part II materials.

How does the Red/Amber/Green model validation framework work?

Master Part II with our FRM Course

Related Questions

Practice Questions

Rating	Meaning	Impact	Remediation Timeline
Red	Critical deficiency	Model may produce materially incorrect results; significant capital or P&L misstatement risk	Immediate action required (30–90 days); model use may be restricted
Amber	Significant concern	Model has notable weaknesses but results are broadly acceptable; compensating controls may mitigate	Action required within 6–12 months; enhanced monitoring
Green	Minor observation	Best-practice improvement; no material impact on results	Address in next model update cycle; informational