≡ Menu

Live from InterACT – Building a Decision Engine


Next up was Chris Collard of Dell talking about building a decision engine. Chris had done an implementation at Dell Financial Services and was sharing some of his experience with replicating that at Dell. Chris talks about decision engines as full decomposed applications – data, process and logic all externalized. Chris’ central thesis is

Effective Decision Management Rule Engine implementations require a higher degree of organizational orchestration and coordination than typical systems implementations

He believes there are ways to:

  • Quantify and characterize the complexity
  • Manage existing rules implementations
  • Plan for new rule implementations

Chris made some interesting assumptions, all of which seem reasonable:

  1. Business rules architecture is one part business architecture and one part IT architecture. Being embraced by business and IT is critical
  2. Focus on enterprise decision management – not just “niche” rules implementations but use of business rules to manage core decisions
  3. Rules technology is well applied for marketing, financial services but not ubiquitous e.g. in account or asset management
  4. Integration is required and SOA worth while.

Chris presented a general framework for decision engines that included data translation, messaging, objects that can be used by rules, rule services themselves, workflow and UI, model and rule development, all built on a rule repository and integrated test and development environment. His point was that this complexity is inherent and unavoidable and a “quick and easy” demo is not reasonable. He then developed a set of scores to see how ready a project or organization would be for this kind of, potentially quite complex, project. These scores came in different areas:

  • Governance
    The existence of a coherent Project Management Office, requirements traceability, BPM and compliance all contribute to a governance score. One of the core reasons for a rules project is to support change and this requires management and alignment.
  • User Interface
    An ability to integrate the user interface of the various elements makes rules adoption easier.
  • IT infrastructure
    Need the IT organization engaged because multiple systems provide data and must be integration. SOA can really help here but scoring this makes it clear how hard it is going to be.
  • Authoring
    Different people with different skills, and cross-over skills, must be available
  • Data and Analytics
    Lack of a decision sciences organization and data integration/federation are big potential barriers.
  • Testing
    An ability to rapidly and effectively test rules with an integrated suite of tools is critical for large scale rules implementation. Chris identified strategy, rules, integration, live production and champion/challenger testing as the 5 layers needed.

I will see if I can get Chris to give me a copy of the spreadsheet to post here.BTW Chris is a member of SABAP – the Strategic Alliance of Blaze Advisor Professionals, a neat organization focused on professional qualifications for rules and Blaze Advisor.