Alternative credit scoring

In West Africa, a majority of adults have no formal banking history. Without a comprehensive credit bureau, microfinance institutions and fintechs either forgo a vast market or lend blind, with high default rates and provisions that weigh on profitability.

Yet these same customers leave a rich digital footprint: mobile-money transaction regularity, telecom top-ups, geographic stability, counterparty network. The challenge is not a lack of data, but the ability to turn it into a reliable, explainable risk score that meets prudential requirements.

ADST builds an alternative scoring engine trained on behavioural signals from mobile money and telecom. We engineer several hundred variables — transaction recency, frequency and amount, income seasonality, location stability, financial social-network density — cleaned and validated through reproducible pipelines.

The model core is a calibrated gradient boosting engine whose every decision is made transparent through SHAP values: the loan officer sees exactly which factors push the score up or down. This explainability is essential for regulatory audit and for guarding against discriminatory bias.

We deliver a real-time score via API, together with a lending threshold optimised to your risk appetite, a recommended credit limit, and a strategy to gradually raise the ceiling based on observed repayment behaviour.