Model risk tiering for your inventory

Answer six questions about one model to see its tier, validation frequency and monitoring checks.

Model questions

These made-up answers describe an application scorecard.

Risk tier

Tier 1
ValidationEvery 12 months
MonitoringMonthly

This model is Tier 1 because its materiality is high and its complexity is medium.

Materiality and complexity

Monitoring checks for this tier

NeededPopulation stability (PSI)Open tool
NeededRanking power (Gini)Open tool
NeededGoods and bads (KS)Open tool
NeededDefault rate calibrationOpen tool
NeededOverrides, input quality
NeededChallenger modelOpen tool

Tier 1 needs 6 of the six checks.

Get the result by email

Run the model inside a decision flow, where its inputs, scores and decisions are logged for every application, so the monitoring checks can run on real decisions. Start your trial

How the tier is calculated

Each answer is worth 1 to 3 points. The three answers for materiality and the three for complexity place the model in the grid.

Materiality = Decision impact + Book covered + Outside use
Complexity = Output review + Model type + Data used
Low = 3–4 · Medium = 5–6 · High = 7–9
Tier 1: Every 12 months · Monthly
Tier 2: Every 24 months · Quarterly
Tier 3: Every 36 months · Yearly

The sums below use the answers shown above.

Materiality = 3 + 3 + 2 = 8 · High
Complexity = 2 + 2 + 1 = 5 · Medium
Risk tier = Tier 1

The points, cut-offs and review intervals are one common approach to tiering. A model risk policy sets its own.

Model risk tiering explained

Model risk tiering groups a model into a tier based on materiality and complexity. Materiality uses decision impact, book covered and outside use. Complexity uses review, model type and data.

Models can have different effects on a lender’s book and decisions. Use the tier to set a validation interval and monitoring checks for each model in the inventory.

Common questions

What is model risk tiering?

Model risk tiering assigns a tier based on a model’s materiality and complexity. This tool works out both from six answers about the model.

How do I rate a model’s risk tier?

The tool adds decision impact, book covered and outside use for materiality. It adds review, model type and data for complexity, then uses both levels in the grid to set a tier.

How often should a credit model be validated?

The tool sets validation every 12 months for tier 1, every 24 months for tier 2 and every 36 months for tier 3. These intervals are one common approach, and a model risk policy sets its own.

What belongs on a model monitoring checklist?

PSI and Gini apply to every tier. KS and PD calibration apply to tiers 1 and 2, and override and input data checks plus a challenger model apply to tier 1.

How does model inventory tiering work?

Answer six questions about a model to get its materiality, complexity and tier. Use the result to set its validation interval and monitoring checks in the inventory.