What Kinds Of Problems Lend Themselves To Edm?

by FICO

(Posted by guest blogger, James Taylor)

I was talking with Neil Raden and Tom Davenport today on the subject of decisions - what are the various kinds of decisions and how do companies make them and think about making them. Afterwards I was reminded on something Rob Meredith posted - IT Archaeology: Whatever Happened to SDS? - a couple of weeks ago. It made me think about how we classify problems suitable for Enterprise Decision Management or EDM.

I often talk about a spectrum of decisions - from strategic to tactical to operational - and describe EDM as increasingly important as one moves along this spectrum. While this helps clarify the difference, for instance, between a strategic decision like "should we do business in China"(not suitable for EDM) and "what should be cross-sell this customer (very suitable), it has a serious limitation. The problem is that how high-volume, operational decisions are taken can be strategic to a company. Indeed the core principle of EDM is that this is the case - that companies should take control of how they make these decisions so as to improve their business.

Rob, thankfully, has a more useful taxonomy. He reminds us of Herbert Simon's work on decision types:

Structured - well understood, repetitive, lend themselves to well-defined rules and steps
Unstructured - difficult, ambiguous, no clear process for deciding
Semi-structured - somewhere in between

This taxonomy works much better as structured decisions are ideal for automation and improvement using EDM, semi-structured decisions can often be supported by EDM, while unstructured are poor choices for the use of EDM. Rob points out unstructured decisions are best supported by what are classically called Decision Support Systems and that previous writers have talked about Structured Decision Systems for supporting the more structured decisions. Rob's perspective is that BI tools are being used to build such Structured Decision Systems while the truly unstructured decisions are left unsupported with Excel and custom hackery being applied piecemeal.

I can see his point here in that BI tools are often very poor at supporting the highly unstructured, collaborative decisions that executives must take. I would suggest, though, that the primary use of BI tools and related technologies is in supporting semi-structured decisions. These still require human intervention in the minds of most and yet have some well defined boundaries, source information, ways to view the problem etc. BI tools are all about getting information to people, not to systems, and require this kind of structure to be usefully applied. I don't think BI tools do any better developing the kind of systems you need to support highly structured decisions because a modern organization increasingly needs these decisions to be automated and, as I have noted before on this blog and others, BI tools don't do that well, if at all. to automate structured decisions you need ways to manage the rules and policies that drive the decision, ways to structure the information into executable predictive analytics and ways to bring all this to bear in a modern service-oriented architecture. You need EDM.