Hosting 10+ Behavioral Scorecards in Bank’s BRMS

Behavioral scorecards are an integral part of the banks’ decision-making system, helping the institution assess the risk level of every service user. With the help of behavior scorecards integrated into a business rule management system (BRMS), a bank can implement machine learning algorithms for risk prediction. This case study examines the implementation of behavior scorecards in the 4IRE client’s existing BRMS to enable automated decision-making and advanced data processing.  

Client’s Profile 

The client is a large international bank covering the Asia-Pacific, European, North American, and MENA regions with many financial services. The current service range of the bank includes deposits and loans, foreign exchange support, trade finance, and e-banking services. 

Client’s Requirements & Goals 

The client acquired a BRMS from a U.S. provider to implement predictive modeling of risks with machine learning algorithms. The BRMS also served as a vital tool for business rule configuration and complex calculations. However, the behavior scorecard management process was conducted by risk managers. The process required automation, so the client wanted to integrate scorecards with the BRMS to achieve greater end-to-end automation of risk management processes. 

The client’s major project requirements and expected outcomes included:

• Setting up an automated process of scorecard hosting and deployment 
• Smart data quality control
• Web-based behavior scorecard updates without the IT department’s participation 
• Coverage of all banking services by scorecards 
• Autonomous control of the server workload with a batch engine 

Business Challenge 

The 4IRE team was tasked with a comprehensive project spanning numerous processes and elements within the client’s existing infrastructure. The business challenge included: 

• Setting up an algorithm for scorecard import to the BRMS, risk model adjustment, and variable calculation. 
• Data retrieval from the bank’s database in collaboration with the IT department and its transformation into scorecard input variables.
• Configuring the logic of exception processing in the banking data processing. 
• Working on the integration task within the required CPU load restrictions (a maximum of 80%). 
• Improving the UI of the BRMS web interface to simplify its use by risk managers. 

Implementation 

The work on this project involved 1 Java developer and 1 BRMS developer. The tech stack applied in this project included FICO Blaze Advisor for rule and scorecard setup and Java for integration and batch wrap-up. 

The project’s implementation focused on creating a scorecard import mechanism for risk managers with lay knowledge of IT specifics and setting up a batch engine for BRMS operation without constraining the entire bank’s infrastructure. These goals were achieved by:

• Successful import of 12 behavioral scorecards for all banking products into the existing BRMS 
• Web interface modifications in line with the model adjustment requirement, enabling the bank’s staff to customize scorecards and manage risk profiles more effectively. 
• Implementing the exclusion processing mechanism. 
• Developing and implementing the Batch Web Application architecture. 
• Setting up the automated model batching algorithm.  
• Providing continuous technical support to the client’s tech team throughout implementation and user acceptance test (UAT).  

Solution Delivered 

The 4IRE team gave the client a working mechanism for simplified scorecard import that could be done even by non-IT staff members. The batch engine for the BRMS was set up to coordinate all data processing operations. The 4IRE experts chose the Batch Web Application as an autonomous multi-threaded application to better organize BRMS processes without cumbersome customization of the existing BRMS infrastructure.

Project Outcomes 

The project took four months of the 4IRE team’s work on the technical solution and its integration. The work started in January 2020, with UAT submission in February 2020 and production in May 2020. 

The resulting BRMS infrastructure thus enabled advanced and speedy data processing, easy scorecard adjustment without IT staff involvement, and automated batching of the models. The project’s outcome complied with the client’s initial specifications by enabling efficient data processing within the bank’s process window using the on-site hardware.