Portfolio impact
The same 4,000 applicants, scored two ways — a bureau-only baseline (what a bank can assess today without alternate data) vs. this engine. The model ranks risk better, so it buys you a choice: fewer defaults or more approvals.
−2.0pp
Default rate cut
19.2% → 17.2% at equal volume
+6.0%
More approvals
+240 businesses at equal risk
14
Credit-invisible included
no track record a bureau could assess
19.2%
Risk appetite held
default rate kept at baseline
Same volume → better quality
Approve the same number of businesses, chosen by the model.
Bureau-only baseline1,910 approved · 19.2% default
This engine (same volume)1,910 approved · 17.2% default
Same risk → more inclusion
Hold the default rate; approve as many as the score safely allows.
Bureau-only baseline1,910 approved · 19.2% default
This engine (same risk)2,150 approved · 19.2% default
Beyond equal-risk, the engine can approve further into the credit-invisible segment — with risk-based pricing (rate scales with PD, see the Health Card) pricing for the residual risk, so expansion stays profitable.
Baseline = approve only established businesses (≥3y vintage + healthy balances) — the few signals a bank has without alternate data. Synthetic population with ground-truth outcomes.