Introduction

A banking and financial services company who aim to take financial products to every part of the country to build self reliance. The main objective was to provide individual, higher ticket size loans to its customers.

Business Objective

• Assess existing platforms & solution scalability

• Build a future ready data lake

• Implement a predictive model for loan disbursement

• Recommend ML models that can be used

• A data model was designed & a data mart was built to aggregate data from all sources.

• Python connectors were setup in data server.

• Automated solutions were implemented to enable data quality & periodic refreshes.

• KPIs were defined.

• A thorough risk analysis of the portfolio was conducted to identify drivers of behavior.

• Feature engineering was implemented. Univariate, Bivariate & Multi-variate analysis enabled statistical modelling.

• Training & validation samples were developed.

• Customers were segmented. Regression models for each segment were identified to predict likelihood of loan default.

Impact

• The suggested architecture empowered the client to build analytical solutions on top of their data

• Identification of customer segments that were more responsive to cross sell opportunities

• Reduce credit risk by enabling data driven loan disbursement.

Technological Framework

Why this setup?

Python and PySpark provide the processing power to transform raw financial data into clean, enriched datasets, while Kafka handles real-time data streaming for continuous model updates and data quality refreshes.

Why this setup?

Interactive dashboards built on AngularJS and NodeJS give business users real-time access to portfolio KPIs, risk metrics, and cross-sell opportunities without requiring technical expertise.

Why this setup?

Integrating bureau data, CRM, and core banking feeds into a unified data mart ensures a 360-degree view of customer behaviour for accurate risk scoring and cross-sell modelling.

Takeaway 

Zapcom built a full analytics lifecycle solution — from strategy and infrastructure to production-grade ML models enabling the client to make data-driven loan disbursement decisions and reduce credit risk.