My next session was on Flexibility, Scalability and High Performance with James Reid of Equifax. The application being discussed is the Equifax InterConnect Platform (part of their risk solutions, described here), a hosted SaaS solution that handles risk-based decisioning, especially credit risk. InterConnect is their decisioning platform for credit and lending and is designed to have rapid speed to market, flexibility, quality and low cost to customize. InterConnect is a very component-based application with a core decisioning component. InterConnect is focused on a pattern based approach that handles things like legacy access and third party data sources, bureau failover, handle Fair Credit Reporting Act (FCRA) rules, generate system events, handle duplicates etc. What if scenarios are supported off line and the platform implements champion/challenger online. Equifax built some custom plug-ins for Eclipse to use with JRules. These include
- A Business Object Model plug-in that handles all the various data sources
- A Rule Flow plug-in to update ILOG’s rule flow to replace their workflow solution by handling workflow to go get data etc.
- An attribute plug-in that handles statistical functions that are used to create characteristics that are used in a scorecard. This extends the business-centric rule language
- A testing plug-in for audit etc. using JUnit
All these plug-ins help move them towards pattern-based development of solutions for customers. Good software development practices, like using a robust object model and interfaces, make a difference in rules too – these plug-ins are not a substitute for them. Additionally a rich fact model allows them to add statistical aggregation features to their rule language and this allows for more sophisticated score cards.
A typical credit policy might contain 1,000-2,000 rules and these are exposed, through the Rule Team Server (RTS). RTS provides role and status definitions that are used to build rule governance. There is a plug-in for RTS that handles adaptive control (wiki) or hypothesis testing in conjunction with Cognos. Their scorecards are designed statistically, using predictive modeling tools from SAS.
Clearly Equifax did a nice job extending the BRMS in a sensible way, using APIs and building on existing features. The overall pattern-driven approach to rule development and maintenance as well as their use of rules with predictive analytic models was very compelling. They are also a great Enterprise Decision Management or EDM story – business rules, predictive analytics, decision services and adaptive control all being used in combination.
Don’t forget to look for more posts on DIABLOG.