Keith Gile wrote an interesting article on operational business intelligence - The Time is Right for Operational Business Intelligence in Business Integration Journal - it is not available online yet, sadly. In it he says that BI solutions have "less of a history with tactical decision-making" and that "The next evolution is toward operational decision-making".
"It's at the operational level - where line workers and customer-facing employees have to make individual transaction-level decisions - that BI is often excluded from the discussion"
He goes on to use an example of a call center representative trying to decide which promotional product to offer each customer during a typically short conversation.
Now it seems to me that what he is discussing here is the difference between "traditional" BI and Enterprise Decision Management, something I discuss at length here. I don't disagree with any of his recommended steps except the one he has omitted - selecting the right technology platform. He identifies four steps:
- "Timing is Everything" about avoiding latency issues
- "Align with processes" about improving or optimizing existing processes
- "Build off what you already have" about extending existing applications
- "Build BI applications in line with the corporate hierarchy" about matching different kinds of BI to different organizational levels
If you consider these four items as it applies to the transactional or operational systems being discussed here the difference between what Keith describes as "Operational BI" and what I describe as EDM comes in the last step. I agree that you must get the timing and latency right, must incorporate the automation of these operational decisions with your business processes and should extend existing applications by adding decisioning services to enhance them. I disagree that the operational layer is simply another kind of BI application.
The call center person in Keith's example does not need a report or an analysis or access to an OLAP cube or anything even remotely similar. They need access to a decision that can drive a script or conversation. This means using business rules to automate this decision and using the same data and techniques you use to build BI reports to instead build predictive analytic models that can be executed in the rules-based decision to analytically enhance it.
This article does a great job summarizing the whys of operational BI. I firmly believe that the how's of operational BI need new technology not found in today's BI solutions - business rules and executable analytics or EDM.
