A Deep Dive into Credit Scoring with ATLAS FINTREST

In the ever-evolving landscape of finance and economics, staying ahead of the curve is imperative for both professionals and institutions. The Finance and Economic Club of ATLAS, FINTREST recently organised an enlightening session on ‘Explore the Latest in Credit Scoring: Innovations in Machine Learning & AI.’ The event featured an esteemed guest, Mr. Manauar Jawed, Chief Credit Officer (CCO) at Aditya Birla Finance, who delved into the fascinating world of innovation and automation in credit score modeling.

Exploring the Essentials of Score Card

Data Sourcing: A Wealth of Information

In a world saturated with data, the foundation of modern credit scoring lies in the ability to source diverse datasets. Financial institutions now have access to an extensive array of information, including banking, financial, GSI, bureau, and alternate data – all readily available in machine-readable formats.

Innovative Tools for Model Monitoring and Maintenance

Maintaining an effective scorecard is no small task. Modern credit assessment leverages innovative data visualization tools to monitor and maintain models. This dynamic approach ensures the accuracy and relevance of these models as they evolve with changing data.

Aligning Underwriting and Risk Strategy

Mr. Jawed emphasized the significance of aligning underwriting and risk strategies with the output of credit scoring models. The scorecard’s role is not only to assess creditworthiness but also to guide business decisions, ensuring optimal risk management.

Advanced Algorithms for Accurate Models

The session shed light on the implementation of advanced algorithms, including neural networks, XG Boost, and Random Forest, in credit scoring. These algorithms offer enhanced accuracy and efficiency, making them indispensable tools in the world of finance.

Embracing New Age Data Sources and Platforms

ONDC: The Unified Digital Commerce Platform

ONDC, the Open Network for Digital Commerce, provides a unified platform that connects stakeholders in the digital commerce value chain. Beyond connectivity, it presents significant lending opportunities for financial institutions.

OCEN: Democratising Credit Access for MSMEs

The Open Credit Enablement Network (OCEN) aims to simplify and democratise credit access for Micro, Small and Medium Enterprises (MSMEs). It offers a streamlined approach to credit assessments, benefiting both businesses and lenders.

India Stack: Enabling Consent-Based Data Sharing

India Stack enables the sharing of consent-based data, such as banking, GST and mutual fund data. This wealth of data facilitates a comprehensive assessment of creditworthiness.

Unconventional Data Sources: Customer Behavior, Geospatial and More

Beyond the established data sources, unconventional data, such as customer behavior and geospatial data, are increasingly being incorporated into credit scoring models. These sources provide a deeper understanding of an applicant’s financial standing.

Harnessing the Power of Advanced Algorithms

Neural Networks: Transforming Fraud Modeling and Credit Scoring

Neural networks are widely used in fraud modeling and customer credit scoring, optimizing risk assessment in today’s complex financial landscape.

XG Boost: Correcting Errors for Enhanced Accuracy

XG Boost corrects errors from previous decision trees, resulting in more accurate and reliable credit scoring models.

Random Forest: Reducing Variance and Boosting Accuracy

Random Forest algorithms reduce variance, providing more stable and accurate credit assessments, critical for responsible lending.

Exploring Advanced Scorecards

The credit scoring landscape offers a diverse array of scorecards, each with its unique purpose.

Application Scorecard: Demographics as the Foundation

Application scorecards rely on demographic data to assess creditworthiness. These are especially useful for understanding the financial capability of first-time borrowers.

Bureau-Based Scorecard: A Deeper Dive into Credit History

Bureau-based scorecards delve into trade line-level performance, utilisation, inquiries and more to provide a comprehensive credit assessment.

Banking Scorecard: Revolutionising SME Scoring

Innovative SME scoring models leverage bank data to assess the creditworthiness of small and medium-sized businesses.

Alternate Data Scorecard: The Non-Traditional Approach

Alternate data scorecards draw from non-traditional sources, offering a fresh perspective on an applicant’s financial situation.

Behavioral Scorecard: A Holistic View of Creditworthiness

Behavioral scorecards take into account regulatory data, financial and banking data, providing a holistic view of an individual’s creditworthiness.

GST Scorecard: Tapping into Historical GST Data

GST scorecards leverage historical GST data to assess an applicant’s financial stability and tax compliance.

Strategising Scorecard Implementation

The session also introduced effective strategies around scorecard implementation.

Zero Underwriting for Top Decile Customers

Excluding bottom deciles and offering standard loan amounts to top deciles without underwriting provides several advantages, including best-in-class TAT to customers and cost savings in loan processing.

Light Touch Underwriting: Streamlining the Process

Light touch underwriting for top decile customers streamlines the lending process, reduces risk, and enhances operational efficiency.

Model Maintenance and Monitoring

The session underlined the importance of evolving with data through data visualisation tools. Credit scoring models need to adapt to demographic changes to remain effective and reliable. Visualisations offer insights into model performance, ensuring that the scorecard remains a dependable tool for decision-making.

The session was a deep dive into the dynamic and ever-evolving world of credit scoring. As we embrace the latest in machine learning and AI, it’s clear that innovation and data-driven insights are shaping the future of finance.