IMPLEMENTED A SCIENTIFIC SOLUTION TO UNDERWRITE GOLD LOANS FOR A CENTURY OLD BANK
CASE STUDY
A CASE STUDY ON
AI-DRIVEN CREDIT
UNDERWRITING SOLUTION
One of the oldest banks in India with a history of over 100 years, offering a wide range of banking services, deposits, loans, saving/current accounts, wanted an intelligent technology solution to transform the current underwriting approach for their gold loans.
The bank’s objective was to reduce the dependency on manual underwriting of gold loans and implement a scientific approach to improve risk assessment accuracy with deeper insights for their existing and new potential borrowers.
A CASE STUDY ON
AI-DRIVEN CREDIT
UNDERWRITING SOLUTION
One of the oldest banks in India with a history of over 100 years, offering a wide range of banking services, deposits, loans, saving/current accounts, wanted an intelligent technology solution to transform the current underwriting approach for their gold loans.
The bank’s objective was to reduce the dependency on manual underwriting of gold loans and implement a scientific approach to improve risk assessment accuracy with deeper insights for their existing and new potential borrowers.
The bank operates from more than 750 branches and provides a wide range of loan products to service the financial needs of small individual customers and large industries. And one of the loan products is a gold loan.
For a gold loan, the bank lends money to a borrower by pledging their gold articles as collateral. And based on the current market value and quality of gold, the loan amount is provided.
According to our client, the critical challenge that the bank faced was to identify the right set of creditworthy borrowers and defaulters. Before reaching out to us, the bank’s risk assessment was solely based on the hands of a branch manager, who evaluates the customer risk appetite based on documents, jewel, and financial status.
PROBLEM STATEMENT
PROBLEM STATEMENT
The bank operates from more than 750 branches and provides a wide range of loan products to service the financial needs of small individual customers and large industries. And one of the loan products is a gold loan.
For a gold loan, the bank lends money to a borrower by pledging their gold articles as collateral. And based on the current market value and quality of gold, the loan amount is provided.
According to our client, the critical challenge that the bank faced was to identify the right set of creditworthy borrowers and defaulters. Before reaching out to us, the bank’s risk assessment was solely based on the hands of a branch manager, who evaluates the customer risk appetite based on documents, jewel, and financial status.
PROBLEM STATEMENT
Adding, the bank did not provide any additional benefits to loyal and good credit-worthy borrowers. Also, no risk premium was added to loan defaulters. Instead, the bank just offered a plain vanilla interest percentage to both ideal borrowers and loan defaulters.
Again, in the case of a top-up loan, the customer must be physically present in a branch, where the decision to provide a top-up loan was solely based on a manual risk assessment by a bank manager.
To overcome these challenges, the client wanted a very scientific solution that can help them improve their risk assessment accuracy and speed up their operations by cutting down the time-to-yes.
PROBLEM STATEMENT
Adding, the bank did not provide any additional benefits to loyal and good credit-worthy borrowers. Also, no risk premium was added to loan defaulters. Instead, the bank just offered a plain vanilla interest percentage to both ideal borrowers and loan defaulters.
Again, in the case of a top-up loan, the customer must be physically present in a branch, where the decision to provide a top-up loan was solely based on a manual risk assessment by a bank manager.
To overcome these challenges, the client wanted a very scientific solution that can help them improve their risk assessment accuracy and speed up their operations by cutting down the time-to-yes.
Mitigating credit risk and bias between borrowers
Predict interest rates based on risk assessment
Provide faster credit decisions
Streamline and standardize all manual processes
Role-based access to all stakeholders
Anytime-anywhere information access
BUSINESS GOALS
BUSINESS GOALS
Mitigating credit risk and bias between borrowers
Predict interest rates based on risk assessment
Provide faster credit decisions
Streamline and standardize all manual processes
Role-based access to all stakeholders
Anytime-anywhere information access
SOLUTION
The data scientists of Zuci initiated this project by understanding the existing loan approval workflow and the various business challenges in the loan cycle to define the success metric beforehand.
With this information, our data engineers collected all the relevant data fields required to help determine a good borrower from a defaulter and fed them into our home-grown credit underwriting solution for feature extraction.
Zuci’s AI-driven credit underwriting solution, HALO, analyzed this fed data and extracted patterns and behaviors. These patterns were then automated and were brought in the correct form to attain data integrity.
After repeated data cleansing, HALO’s machine learning algorithm created a unique credit underwriting model that exploits these patterns and behaviors to identify high-risk and creditworthy borrowers when underwriting a new customer.
Finally, the model was continuously trained to attain the highest accuracy in predicting credit risk for a gold loan before deployment in the bank’s environment.
SOLUTION
The data scientists of Zuci initiated this project by understanding the existing loan approval workflow and the various business challenges in the loan cycle to define the success metric beforehand.
With this information, our data engineers collected all the relevant data fields required to help determine a good borrower from a defaulter and fed them into our home-grown credit underwriting solution for feature extraction.
Zuci’s AI-driven credit underwriting solution, HALO, analyzed this fed data and extracted patterns and behaviors. These patterns were then automated and were brought in the correct form to attain data integrity.
After repeated data cleansing, HALO’s machine learning algorithm created a unique credit underwriting model that exploits these patterns and behaviors to identify high-risk and creditworthy borrowers when underwriting a new customer.
Finally, the model was continuously trained to attain the highest accuracy in predicting credit risk for a gold loan before deployment in the bank’s environment.
Our team started with a discovery phase by conducting a 3-day workshop with the stakeholders
Understood overall requirement, end-to-end loan life cycle, challenges, and business goals
Post-discovery phase, the bank provided a sample set of customers with the pre-defined data fields for creating a sample model
With API calls, our team of data engineers fed the sample set into HALO for data analysis, feature extraction and curation, and finally built an underwriting model
Presented a demo to all stakeholders with all the necessary KPI’s and addressed all questions from stakeholders
HOW ZUCI SYSTEMS HELPED
HOW ZUCI SYSTEMS HELPED
Our team started with a discovery phase by conducting a 3-day workshop with the stakeholders
Understood overall requirement, end-to-end loan life cycle, challenges, and business goals
Post-discovery phase, the bank provided a sample set of customers with the pre-defined data fields for creating a sample model
With API calls, our team of data engineers fed the sample set into HALO for data analysis, feature extraction and curation, and finally built an underwriting model
Presented a demo to all stakeholders with all the necessary KPI’s and addressed all questions from stakeholders
HOW ZUCI SYSTEMS HELPED
Predicted good borrower and the bad ones by scoring the sample set customers and post all the stakeholders were satisfied, the bank provided a more extensive sample set of banking customers to train the model for better accuracy further
Once approved and stakeholder feedbacks were addressed, our team deployed HALO to the bank’s production environment
Provided HALO product documentation (including a system administration guide), 24/7 product support, oversight for production rollout, and post-production support to ensure successful business adoption
HOW ZUCI SYSTEMS HELPED
Predicted good borrower and the bad ones by scoring the sample set customers and post all the stakeholders were satisfied, the bank provided a more extensive sample set of banking customers to train the model for better accuracy further
Once approved and stakeholder feedbacks were addressed, our team deployed HALO to the bank’s production environment
Provided HALO product documentation (including a system administration guide), 24/7 product support, oversight for production rollout, and post-production support to ensure successful business adoption
