10 Principles of the New Business Intelligence

Business intelligence--and its predecessor concepts decision support, executive information systems, and so forth--have been circulating for several decades in business. However, I don't think it's ever fully worked. What we've done is to throw data (often in the form of difficult-to-navigate data warehouses) and software tools at business users, and said "Go at it." That's simply been too hard. We used a lot of terms like "ad hoc queries" and "drill down," but it simply didn't happen very often.

I've argued for a while that organizations need to increase their focus on decision-making. In particular, they need to think again about the relationship between information and decision-making. I recently completed a study on this topic, with the sponsorship of IBM's Information Management business unit, in which I looked at 26 efforts to improve decision-making in organizations. I concluded the following ten things about how business intelligence (BI) needs to evolve:

1. Decisions are the unit of work to which BI initiatives should be applied.
2. Providing access to data and tools isn't enough if you want to ensure that decisions are actually improved.
3. If you're going to supply data to a decision-maker, it should be only what is needed to make the decision.
4. The relationship between information and decisions is a choice organizations can make--from "loosely coupled," which is what happens in traditional BI, to "automated," in which the decision is made through automation
davenport triangle.jpg5. "Loosely coupled" decision and information relationships are efficient to provision with information (hence many decisions can be supported), but don't often lead to better decisions.
6. The most interesting relationship involves "structured human" decisions, in which human beings still make the final decision, but the specific information used to make the decision is made available to the decision-maker in some enhanced fashion.
7. You can't really determine the value of BI or data warehousing unless they're linked to a particular initiative to improve decision-making. Otherwise, you'll have no idea how the information and tools are being used.
8. The more closely you want to link information and decisions, the more specific you have to get in focusing on a particular decision.
9. Efforts to create "one version of the truth" are useful in creating better decisions, but you can spend a lot of time and money on that goal for uncertain return unless you are very focused on the decisions to be made as a result.
10. Business intelligence results will increasingly be achieved by IT solutions that are specific to particular industries and decisions within them.

There was a story in this week's InformationWeek that I believe strongly supports this set of ideas. One quote is particular perspicacious:

BI historically has been about dashboards and scorecards developed for specific uses, says AMR Research analyst John Haggerty. But that's changing. "All of a sudden it's about integrated analytics within applications," he says. "The conversation is starting to shift to looking at information in the context of specific decisions and roles."

My sense is that this shift will be a major one, and both vendors and users of BI will be transformed by it. What do you think?

Tom Davenport

Tom Davenport holds the President’s Chair in Information Technology and Management at Babson College, where he also leads the Process Management and Working Knowledge Research Centers. His books and articles on business process reengineering, knowledge management, attention management, knowledge worker productivity, and analytical competition helped to establish each of those business ideas.