The Tail That Wags the Dog

Where does business intelligence and decision support begin?

By Seth Grimes

Data analysis practices go by many different labels. I don't know who coined the one that forms this column's heading, but I do know that decision support (DS) practice has a long history, predating online analytic processing (OLAP), data warehousing, and data mining by decades. These latter disciplines contribute to efforts to turn data into "actionable" information, but DS subsumes them and more. Although business intelligence (BI), a new kid on the analytic block, is now perhaps the most commonly used catch-all term, I see it as yet other DS subdiscipline - albeit one that also encompasses many subfields and is rapidly growing in importance.

The distinction between DS and BI might seem unimportant for practitioners, but the two fields' vocabularies and toolsets do differ. BI tools generally have broader appeal with greater ease of use, while non-BI DS tools are generally more technically sophisticated and more focused on particular subject domains. Understanding the distinction can help you choose the best approaches and tools for the analytic computing tasks you have at hand.

Signs of Intelligence

When I write my Decision Support columns, I always think about the setting in which they appear, as one Intelligent Enterprise column among many. I try to choose topics that are relevant to the magazine's goal (in my view) of describing how organizations can and do integrate distributed operational and analytic systems to maximize capabilities, efficiency, and profitability.

Although I doubt that I'd characterize even the most hypercompetitive enterprise as "intelligent," data analysis clearly plays a central role for the more advanced among them. (I suppose you could consider an enterprise to be intelligent if it passes a sort of Turing test. The mathematician Alan Turing, who created much of the foundation of computing theory, proposed that a system could be considered intelligent when a person could converse with it and with another person for an extended time without it being apparent which interlocutor is a machine. When customers, suppliers, and business partners interact with your organization - with the enterprise relationship management tools and processes that govern dealings with the outside world - do they encounter the equivalent of a lucid, coherent individual or of a dim-witted amnesiac with multiple personalities?)

Classifying DS

I've seen DS systems classified in a number of ways. You can find a very helpful introduction to DS online at dssresouces.com, where Daniel J. Power, Professor of Information Systems and Management at University of Northern Iowa, summarizes and synthesizes more than 30 years of academic and professional thinking in light of his own experience in the field. Power categorizes DS into systems that are:

• Data driven, emphasizing manipulation of volumes of structured data, most often involving aggregation and calculation of simple measures
• Model driven, although data-based, using systems of parameterized equations and statistical techniques to create "actionable" business indicators
• Knowledge driven, relying on business rules, knowledge bases, and inference engines
• Document driven, attempting to gather, classify, and understand unstructured text
• Communications driven, also known as "group DS," facilitating collaborative decision-making.

(The categories are Power's; the descriptions are mine.) Of course real-world systems occupy multiple categories to varying extents and have characteristics that form new categories like "Web driven" that are of lesser importance in understanding analytic roles.