Barb von Halle and Larry Goldberg from KPI and one of their clients presented to the Object Management Group session on the KPI Decision Model (Barb recorded a webinar on the Decision Model for me some time back). KPI has been working on the Decision Model for some time and published a book on this topic – The Decision Model: A Business Logic Framework Linking Business and Technology – back in 2009.
The basic approach for the Decision Model is to manage the decisions separate from a process, identifying the tasks within the process that is a decision-making task and linking those to a formal model of the decision – a Decision Model. Each Decision Model has a set of related Rule Families, each represented in a rule family table that contains a set of atomic logic statements that follow a rigorous format and can be normalized. The approach is a top-down one, starting with decisions that are driven by the business needs and tracked to the measures and metrics of the book (something very in line with my approach of beginning with the decision in mind).
The Decision Model has a notation (you can see an example and more information on the Decision Model page on the KPI site) as well as a set of rules for completeness and correctness of the various elements involved. Each rule family has a single conclusion and builds on dependent rule families that deliver interim conclusions to the overall decision. This dependency network of rule families is easy for non-technical folks to use to see what contributes to the decision. Rule family tables look very like a decision table – and could be described as such if you have a broad definition of what a decision table is – but are a little different from the way people think about decision tables. The rule family tables follow a number of principles for verification and validation, don’t contain ORs and ELSEs etc.
Clients have been using the model for a few years now and KPI say they get a lot of feedback from folks that it is quicker and easier to use than other ways of describing decision logic. Since the book was published the folks at KPI have created some additional extensions of the model around Decision Views and automated testing (though they did not talk about either of these so I can’t tell you what they mean by this).
Their client reported that they found the tabular format and the grouping of rule families into decisions was very intuitive, much more so than their previous use of natural language rules. The principles and normalization techniques also help ensure that multiple people working on the same problem come to the same answer, or at least that there is a coherent way to discuss the right way to model the decision. As they have become more familiar with the Decision Model they also find that changes to policy are being discussed and designed in terms of the Decision Model. This in turn leads to the business seeing the rules and decisions in the model as assets and being willing to determine what changes are required themselves. Finally, as the ability to load these decision models directly into the business rules management system has come online they can also push validated and correct rules (based on the rule families) directly into production. The power of the model to improve business/IT collaboration is what was most impressive– only a collaborative approach to decisions works so this is critical.