Overview: Business intelligence provides a snapshot of past operations. While the insight might be generated close to real-time, it remains a view of the past. Taking this guidance to drive the way you do business in your systems is akin to driving by looking in the rear-view mirror.
In order to update your systems to operationalize the insight gained in your dashboard, you would typically need to actually alter the code, undergoing a complete software development lifecycle. The task of capturing, analyzing, implementing and testing the changes can take weeks, months or sometimes years. By the time the effort is complete, more change requests might have accumulated and caused the never-ending backlog that your IT organization is cursed with.
In order to accelerate the operationalization of business intelligence, it is critical to create a fast link to production systems, removing the programming tasks which are causing the delays. One proven approach is the use of business rules management systems or more generally speaking decision management systems. By exposing the decision logic to the business owners or business analysts, they allow IT to focus on providing more data sources or making more actions available. Business users can then tweak the systems in a safe environment without going through the complete software development lifecycle, while keeping the lifecycle safeguards in place. As insight is gained, business users can make the proper changes and test them against historical data, submit an approval request and get the changes pushed within hours, accelerating significantly the process. The business control offered by decision management systems allows changes to take place fast. But because changes for the changing are not enough, it is key to measure the impact of the changes. These measurements can happen at various times in the process: very early on when the changes are captured, later on when they are ready and need to be tested against a large test set, and finally on-going in the production system closing the loop.
This approach has been proven to add agility and increase business performance as part of the systems. For some domains like marketing or fraud, changes can occur much faster though, pushing the pace of changes to the extreme. In the class, we will talk about complementary techniques leveraging machine learning to complement the business user's expertise with data insight in a more automated fashion. The first one is called experimental design and allows different strategies to be pushed into production for simultaneous experimentation with live data. The second one leverages predictive analytics algorithms to uncover new trends and turn them into understandable business rules that can be massaged. The last one relies on unsupervised learning for the most chaotic systems.
Areas Covered in the Session:
Who Will Benefit: