Skip to main content
Separate new and used vehicles, purchase loans and leases, and prime and non-prime cohorts. The underwriting unit is the borrower, facility and vehicle at a dated evidence cutoff.

​
Financial-risk measures

​
Public benchmark sources

Use the auto-loan series in CMHC’s non-mortgage delinquency tables, retaining its quarter and 90-day threshold. The Bank of Canada indicators provide another defined credit-performance reference. Match the published product scope before comparing a lender’s auto book. These aggregates do not supply vehicle-level valuation, repossession cost or recovery labels. Establish those benchmarks from an appropriately authorized, dated portfolio cohort rather than assigning a market-wide loss severity.

​
AMLOK evaluation

Test whether extracted contracts, payments and income reconcile to the facility. Measure amount, date, term and rate extraction separately. Evaluate collateral calculations only where a valid valuation is supplied; missing collateral evidence must remain missing. For outcome testing, hold out borrowers, vehicles where identifiable and later origination vintages. Report discrimination and calibration separately for new and used vehicles and for each supported market. For Signal, evaluate missed-payment and cash-flow change signals without exposing repossession or borrower-contact commands. Follow the common benchmark contract and report task-specific results through the evaluation protocol.