Project Overview

The project focused on implementing Predictive Data Models to reduce Bank-Debt Recovery risks using data collected over Recovery Calls. The objective was to develop a machine learning-based solution that would enable the BPO Call Centre to formulate a predictive data model for reducing the Non-Performing Assets (NPA) rate seen by banks. By utilizing audio call records and internal call feedback data, the model aimed to predict lender defaulting and provide an alternate recovery strategy.

The Problem

Prior to the implementation of the Predictive Data Models, the BPO Call Centre faced challenges in efficiently recovering debts for lenders who were delaying or defaulting on their bank loans. The existing process lacked an effective means of identifying defaulting lenders early on, leading to increased NPA rates and financial risks for the banks.

The Goal

The goal of this case study was to develop a Predictive Data Model that leveraged machine learning techniques to analyze audio call records and internal call feedback data. The objectives included:

• Reduce the NPA rate experienced by banks by identifying lenders at risk of defaulting on their loans.
• Provide an alternate recovery strategy to mitigate financial risks associated with defaulting lenders.
• Enhance the overall efficiency and effectiveness of the debt recovery process for the BPO Call Centre.

User Research

• Conducted data analysis to understand patterns and trends in audio call records and internal call feedback data.
• Performed machine learning algorithms on historical data to train the Predictive Data Models.
• Evaluated the accuracy and performance of the models through rigorous testing and validation.

Through user research, it was identified that the BPO Call Centre faced challenges in effectively recovering debts from lenders. By analyzing audio call records and internal call feedback data, there was an opportunity to develop a predictive data model that could reduce the NPA rate and provide an alternate recovery strategy.

Pain Points

The main pain point experienced by the BPO Call Centre was the inability to accurately identify lenders at risk of defaulting on their loans. This lack of early detection resulted in increased NPA rates and financial risks for the banks.

Solution Brief

Predictive Data Model: Our team formulated a Predictive Data Model using machine learning techniques. The model utilized audio call records and internal call feedback data to predict the likelihood of lender defaulting. This enabled the BPO Call Centre to create an alternate recovery strategy, reducing NPA rates and mitigating financial risks for the banks.

Impact and Lessons Learned

The implementation of the Predictive Data Model had a significant impact on the BPO Call Centre and the banks it served. The key outcomes of the project were:

• Reduced NPA rates: By accurately identifying lenders at risk of defaulting, the Predictive Data Model helped reduce the NPA rates experienced by the banks.
• Improved recovery strategies: The model provided insights that allowed the BPO Call Centre to create alternative recovery strategies, increasing the effectiveness of the debt recovery process.
• Enhanced risk management: By mitigating financial risks associated with defaulting lenders, the banks could allocate resources more efficiently and minimize losses.
• Increased operational efficiency: The Predictive Data Model automated the identification process, saving time and effort for the BPO Call Centre.

Through this project, we learned the value of utilizing machine learning and predictive analytics in the debt recovery process. The successful implementation of the Predictive Data Model showcased the impact of data-driven insights in reducing NPA rates and improving recovery strategies, ultimately benefiting the BPO Call Centre and the banks it served.