Monday, May 18, 2009

Database Design and the Demands of Operational Business Intelligence

Anant Gupta, SOA Practice Director

When hearing the term Business Intelligence the first thought that springs to mind is that of reporting and trend analysis based on historical data. So what exactly is Operational Business Intelligence (BI)? It almost seems to be an oxymoron. In one sentence, it is the mechanism of enabling everyone, and not just the strategy team as in the case of BI, to make decisions that improve the top-line, bottom-line or enhance customer satisfaction. These decisions are not made quarterly or yearly but on a transactional basis. A very simple example of operational intelligence is your GPS system telling you to take a different route to work because of a bad traffic condition on your regular route. To make this more interesting, let’s assume that your system not only understands the traffic conditions, but also gas prices, your schedule, tolls, etc., and needs to present you the most effective option based on all of these factors. To do this, it needs to perform a lot of analysis in real-time. Similarly, systems need to look at the product / promotion info and customer purchase history / status and determine cross-sell, up-sell opportunities.

So far we have been talking about transactional databases or data warehouses / data marts, one for OLTP and the other designed strictly for post-facto reporting and analysis. The new trend will be a middle path where we will be designing data marts that will have the capability to be queried and performing analysis in real time. There will also be a trend that will emerge to update the operational data stores with updates from the data warehouse. Essentially, data warehouses will be able to continuously analyze data, recognize and send events that operations will subscribe to. For example, I might be interested in subscribing to an event if merchandise returns for one of my product lines exceed 5% of normal returns. So, databases will need to start understanding events and perform continuous analysis to determine the occurrence of any of those events.

All of this imposes challenges to the database design as both the depth of analysis and performance will have to be delivered simultaneously.

Anant Gupta was recently named the SOA Practice Director at Prolifics after serving as a Senior Business Integration and J2EE architect Anant has with extensive experience in IBM's SOA software portfolio and specializes in delivering business integration and business process management solutions. He has worked for major clients in the banking, insurance, telecommunications and technology industries.