AfriScore is a B2B credit intelligence platform that converts mobile money transactions, airtime patterns, and device data into explainable credit decisions for financial institutions serving underbanked African markets. Its machine-learning engine delivers risk scores, tiers, and recommended loan terms with SHAP-based explanations for regulatory compliance. The platform has processed over 3,500 loans in a Nairobi pilot, with 700+ previously unbanked individuals gaining credit access.
Funding
Funding not disclosed
Founders
Product
Problem
Traditional credit bureaus cover less than 5% of the population in most Sub-Saharan African countries, leaving over 300 million creditworthy adults and small businesses without formal financial histories. This credit-invisible population is locked out of formal lending despite an estimated $100 billion annual credit gap, not because they are risky, but because they lack the traditional data required to prove creditworthiness.
Solution
AfriScore provides a B2B credit decisioning layer that transforms mobile money transaction histories, airtime purchase patterns, and device metadata into explainable, compliant credit decisions. The platform ingests alternative data via secure API, scores applicants using a calibrated Random Forest ensemble model with isotonic calibration, and returns risk scores, risk tiers, and recommended loan terms in real time. Every decision includes SHAP-based feature contributions and counterfactual explanations, ensuring auditable and regulator-friendly outcomes. The system is designed for on-premise deployment to meet strict data residency requirements common among African financial institutions.
Target Audience
Primary customers are financial institutions across emerging African markets, including credit unions, microfinance institutions, and banks serving informal-economy borrowers in regions like the CEMAC area and East Africa.
Features
- Python-based ML pipeline using Random Forest ensemble with isotonic calibration and SHAP explainability
- Real-time scoring engine that processes mobile money, airtime, and device data via secure API
- SHAP-based feature contributions and counterfactual explanations for every credit decision
- Self-hostable architecture supporting on-premise deployment for data residency compliance
- Live interactive demo allowing users to run hypothetical credit assessments
- API reference and documentation for institutional integration