South Asian auto lender gains a 2x uplift
in predictive accuracy

An emerging auto lender in South Asia providing loans to individuals and small and medium-sized enterprises (SMEs). The company set out to gain a competitive edge in its market by strengthening the accuracy and reliability of its credit decisions.

As an emerging auto lender in South Asia, the client was looking for a competitive edge while serving both individuals and SMEs. Its existing credit scoring system was holding it back: the legacy model returned a low Gini index of 0.4, signaling weak predictive power and a real risk of rising loan defaults down the line. The problem was compounded by data preparation issues (inconsistencies and gaps in the available data) that further undermined the reliability of credit decisions.

The GiniMachine team started with rigorous data preparation, resolving the inconsistencies and filling the gaps that had been limiting the existing system. On the cleaned dataset, the team built a custom AI-driven credit scoring model and applied machine learning strategies to maximize its predictive power. The model was then thoroughly validated and back-tested on hold-out datasets to confirm its accuracy and reliability before it was put to use.

The data work and machine learning approach delivered a marked improvement in performance. Predictive power, measured by the Gini index, rose from 0.4 to 0.75, which nearly doubled the original. GiniMachine also lifted the ROC-AUC to 0.88, sharpening the quality of predictions and reducing the False Positive Rate (FPR). Validated and back-tested on hold-out data, the model now gives the client a trusted, high-performing tool to grow its loan portfolio strategically and with confidence.