Developing Alternative Data-Based Credit Scoring for Thin-File Indian Borrowers

Objective: To design and validate a credit-scoring model using alternative data sources to assess creditworthiness of thin-file borrowers in India.

Millions of Indians remain underserved by formal credit because they lack traditional credit histories, making them 'thin-file' borrowers. Standard credit scoring models relying on bureau data exclude gig workers, new-to-credit individuals, and informal sector participants, which hampers financial inclusion and growth for leading fintechs and NBFCs.

This project will identify, source, and analyze alternative data—such as mobile payment behavior, utility bill payments, e-commerce transaction history, and digital footprint—from public datasets and partner sources. Students will benchmark global and Indian approaches before selecting suitable features and modeling techniques for the Indian context.

Deliverables include data engineering pipelines, an end-to-end credit scoring model (machine learning or statistical), validation metrics (e.g., AUC, Gini), and a business case report. The team will demonstrate results with real or synthetic datasets, including explainability analysis and actionable dashboards.

Business impact: The project enables fintechs and NBFCs to responsibly extend credit to underserved populations, reduce default rates, and comply with evolving RBI guidelines. The model’s insights will inform lending policy, partnership design, and inclusion strategy.