Developing a Credit Scoring Model for a Leading NBFC in Sri Lanka

Our client, a leading NBFC in Sri Lanka, partnered with NeenOpal to enhance its gold loan disbursement process by building a machine learning-based customer scoring model. Instead of traditional credit checks, the model enabled smarter loan-to-value (LTV) decisions, allowing the client to confidently offer higher LTVs to trustworthy customers based on historical gold loan and transaction behavior.

Developing a Credit Scoring Model for a Leading NBFC in Sri Lanka
10-20%KPI Arrow
Higher LTV for Low-Risk Customers
40%KPI Arrow
Faster Approvals (1.5 Hrs)
15%KPI Arrow
Increase in Customer Retention

Customer Challenges

The client struggled with outdated, manual processes for risk assessment and loan approvals. This led to inefficiencies, inconsistent decisions, and missed opportunities for personalization.

Unstructured Risk Evaluation Process

Unstructured Risk Evaluation Process

The client relied on static rules and manual judgment to assess customer eligibility, leading to delayed approvals and inconsistent turnaround times.

Inefficient LTV Decisions

Inefficient LTV Decisions

A few high-risk customers received excessive loans, while some loyal borrowers were underfunded, impacting portfolio quality and growth.

Lack of Personalization

Lack of Personalization

All customers were treated with uniform risk criteria, missing opportunities to tailor LTV offerings based on individual behavior and history.

Solutions

NeenOpal developed a customized, end-to-end machine learning–based credit scoring system specifically designed for the client’s gold loan operations. The solution aimed to automate and enhance risk evaluation, improve loan approval efficiency, and enable data-driven, personalized LTV decisions.

A comprehensive data pipeline was built to extract and cleanse historical loan data, including ticket-level gold loan transactions, repayment timelines, pledged gold quality, previous LTV ratios, and customer demographics. This structured dataset formed the foundation for building reliable predictive credit scoring models.

01

Key behavioral and financial indicators were extracted and engineered into meaningful features. These included metrics such as loan application frequency, overdue duration trends, timely loan closures, gold valuation consistency, and borrower repayment patterns, all contributing to predictive accuracy.

02

A linear regression model was implemented to derive a weighted scoring equation that balanced multiple customer attributes. This scoring mechanism provided each borrower with a numerical risk score, enabling objective and scalable evaluations.

03

Based on the generated scores, customers were segmented into low, medium, and high risk tiers. This allowed the client to implement loan strategies, offering higher LTV ratios and faster approvals for low-risk customers, while applying stricter criteria to high-risk segments.

04

Why choose NeenOpal?

NeenOpal brings deep expertise in financial services, specializing in machine learning solutions for credit risk modeling, customer scoring, and loan optimization. Our team has built predictive models using complex datasets, including gold loan portfolios and repayment patterns. With strong capabilities in supervised learning, feature engineering, and model interpretability, we help financial institutions automate risk assessment, reduce defaults, and make smarter lending decisions, while staying aligned with regulatory and business goals.

Services Used

AWS CodePipeline
AWS CodePipeline
pandas
Pandas
scikit-learn
Scikit-learn
matplotlib
Matplotlib
seaborn
Seaborn
SQL
SQL
SQL Server Management Studio
SQL Server Management Studio
EventBridge
Jupyter Notebook
Tableau
Tableau

Benefits

The solution helped the client to make faster, smarter, and more personalized lending decisions. By embedding data science into core operations, it improved efficiency, consistency, and customer satisfaction across the gold loan business.

Conclusion

The deployment of a machine learning–powered customer scoring engine transformed our client’s gold loan operations. By leveraging predictive analytics, the model enabled more precise LTV decisions and significantly reduced approval turnaround time by 40%. This led to faster, more tailored disbursals, improved portfolio quality, and a scalable foundation for long-term, data-driven lending growth.

Authors

Yash Khare

Senior Data Scientist

LinkedIn

Madiha Khan

Content Writer

LinkedIn
Contact Us

Contact Us To See How We Can Help You Achieve Your Goals

Libraries

Related Case Studies