Fintech · AINorth America·Credit Intelligence · Decisioning2023

Predictive Credit Decisioning — SME Lending Platform


Situation & Challenge

An alternative lending platform was operating on a legacy scorecard with a high manual-review rate and approval cycles measured in days. The scorecard was incorrectly rejecting a meaningful segment of creditworthy applicants — the feature set did not reflect alternative repayment signals available through open banking, and there was no mechanism to learn from resolution patterns over time.

What Coadjutant Built

An ensemble model incorporating 140 features drawn from bank feed, open banking, bureau, and behavioural signals. Monthly automated retraining pipeline with drift monitoring. SHAP-based explainability layer generating decision rationale for every application — required for fair lending documentation, adverse action notices, and regulatory examination readiness.

Technology & Protocols

Gradient boosting ensembleOpen banking integrationSHAP explainabilityAutomated retrainingFair lending audit layerAdverse action generator