Model documentation
For yourmodel inventory
Supervisory guidance on model risk management treats a risk rating calculation as a model. This page is written so your bank can drop it straight into its model inventory and answer the examiner question without a research project.
Model description
- Purpose
- Calculate a composite borrower risk grade from spread financial data using grading limits set by the bank, and record that grade on the credit memo alongside the policy set that produced it.
- Inputs
- Balance sheet, cash flow, debt schedule, collateral, and tax return data entered or imported into the workbook. No external data source, no market feed, and no vendor supplied scoring data.
- Calculation logic
- Each ratio is computed with standard credit formulas visible in the worksheet cells. Each ratio is compared to the bank's own policy limits on the policy tab and assigned a grade. The composite grade is a weighted combination of those individual grades, with the weights also set on the policy tab.
- Ownership of parameters
- Every threshold and every weight lives on a tab inside the bank's own file. The vendor does not supply, tune, or hold the parameters. Changing credit policy means editing that tab, and the change applies to files graded after it.
- Transparency
- There is no compiled scoring engine and no proprietary black box. Every formula is a readable worksheet formula and the supporting VBA project is unlocked, so internal audit or a validator can trace any grade back to its inputs by hand.
- Output and record keeping
- The generated credit memo carries the composite grade and a stamp identifying the policy set in force when the file was graded. A file graded last year still shows the limits that graded it, which is the record an examiner asks for.
- Human judgment
- The calculation does not approve or decline credit. The loan officer writes the narrative and the committee makes the decision. The grade is an input to that decision, not a substitute for it.
- Known limitations
- Output quality depends on the accuracy of the data entered or imported. Scanned tax returns of poor quality are refused rather than estimated. The consolidated relationship workbook has an entity cap. Limits set inconsistently with written credit policy will produce grades inconsistent with written credit policy.
- Validation approach
- Because the parameters and formulas belong to the bank, validation is a review the bank can perform itself: confirm the policy limits match written credit policy, recompute a sample of ratios by hand from source documents, and confirm the composite weighting matches what the credit committee intended. Repeat when policy changes.
This description is provided for your convenience. It is not a validation, and no regulator has reviewed or endorsed it. Your bank remains responsible for the model inventory entry and for validating the parameters it sets.
Related
The rest of the vendor file
Data handling, file integrity, versioning, continuity, and why there is no SOC 2 report are all written out for your compliance team.
