A bank has 40 models: 8 complex capital and pricing models, 20 moderate scoring and reporting models and 12 simple tools. Tier them, set validation cycles and estimate the validation team's annual workload.

Put the 8 capital and pricing models in tier 1 on a yearly cycle, the 20 moderate models in tier 2 every two years, and the 12 tools in tier 3 every three years: about 376 validator days a year. That is 240 plus 120 plus 16. At about 200 working days a validator, it needs roughly two people for periodic reviews alone, before new models and changes.

A hospital does not give every patient the same scan every week; it triages by how serious the condition is and how fast it could turn. A model's tier comes from what a mistake in it would cost and how likely a mistake is: its materiality, its complexity and how it is used. Capital and pricing models are all three: they drive large numbers, involve heavy maths and feed regulatory filings, so an error is both likely and expensive. A spreadsheet that formats a monthly report has none of those. Write the criteria down, so a model's tier is a rule anyone can apply, not a negotiation with its owner.

Annualise each tier by dividing its effort by its cycle. Tier 1 costs 8 x 30 = 240 days every year, tier 2 costs 20 x 12 / 2 = 120, and tier 3 costs 12 x 4 / 3 = 16, about 376 days a year. Two thirds of the effort goes on one fifth of the models, which is the point. Stagger the tier 2 and tier 3 reviews across the cycle; if all twenty tier 2 models fell due in the same year, that year would need 528 days rather than 376.

The 376 days cover periodic reviews only. New models, material changes to existing ones and annual performance monitoring for tier 1 models all come on top, and in many banks they take as much time again. So a team of two covers the cycle but not the bank; three or four validators is a more honest estimate. Tiering is also not fixed: a tier 3 tool that starts feeding a capital number moves to tier 1 the day it does, which is why the inventory needs an owner who reviews tiers at least yearly.

Say what the tiering protects against. Without it, a team validates whatever the business asks for first, and the busiest model owners get the most attention while quiet high-impact models go years without review. With it, the bank can show a supervisor that effort follows risk. Check the current supervisory guidance on model risk management for any minimum review frequencies before fixing the cycles.