Comprehensive MSME Credit Assessment Platform
A multi-source, explainable underwriting platform combining statistical PD modelling with expert credit judgment.
- Multi-source view: financials + banking behaviour + bureau
- Explainable logistic-regression PD models with ML challengers
- Governed workflow combining model output and expert credit judgment
From business problem to measurable impact.
Business Problem
Corporate-style assessment based mainly on annual financial statements and broad benchmarks did not adequately reflect the heterogeneity, information latency and faster transmission of shocks in MSMEs.
My Role
Worked across business, credit, analytics and technology to translate MSME credit knowledge into features and model logic, shape the predictive assessment approach and connect scoring, policy, credit judgment, workflow and monitoring into a governed platform.
Solution
The platform combined financial statements, banking behaviour, bureau signals and engineered trends. Explainable logistic-regression PD models formed the backbone, with decision trees and random forests used as challengers where non-linear relationships added signal. Model output was embedded into a governed credit-decision architecture rather than treated as an autonomous decision.
Technology & Methods
Logistic regression · probability of default · decision trees · random forest · feature engineering · temporal validation · financial ratios · banking behaviour · bureau data · scorecard / risk bands · monitoring
Leadership Lesson
“In credit, predictive power is necessary but not sufficient. Explainability, domain logic, temporal validation, governance and expert accountability must be designed with the model from the start.”
Detailed Case Study
The infographic above is the executive summary. The full portfolio document contains the methodology, operating model, governance and implementation narrative at an externally shareable level.
View detailed PDF ↗