≡ Menu

First Look – Blaze Advisor 7.1 Update


I wrote about Blaze Advisor 7.0 back in September. FICO is about to release Blaze Advisor 7.1 and key themes for 7.1 are a focus on increased business control and flexibility, faster time to deployment and more integrated simulation. The key driver for this comes from the steady increase in business user engagement – the typical Blaze Advisor business user wants more control over the rules and more of them also want to handle the whole lifecycle from creation to deployment. This has resulted in an investment in HTML 5, especially for an all new Decision Table editor, fewer requirements for customization and features to allow business users to achieve more without IT resources. More and more projects are being driven out of the business side and the environment is shifting from an assumption of a technical environment backed by a partial business user environment to one in which the business user environment allows 100% of the tasks to be performed.

  • The new Decision Table is completely HTML 5 based. No applet is required so access to the metaphor is instant with no download and accessible from any browser including mobile browsers. Pagination is supported to handle large tables, the size of the table in the browser can be expanded and rules can be uniquely identified for tracking purposes in the table even as they are re-ordered. Data dependency (Make:Model for instance) can be defined between columns so that a row must contain values that are compatible, rows can be filtered using a quick filter interface (like Excel’s) and typeahead is automatically supported for list values. One key feature of the new Decision Tables is support dynamic binding of columns to object properties so that changes in the underlying object model once a system is deployed can be rapidly and easily managed by non-technical users. This is important as one of the most common causes of a harder-than-usual rule update is a change in the underlying meta model and this could be mitigated using this dynamic approach.
  • Prepackaged approval workflow and user flows for use in the web based business user environment are now bundled with 7.1. This is completely customizable in the business environment and is focused on business user control and approval and the workflow for maker/checker etc that they need. This workflow is now completely delivered in the web-based rule management environment. Permissions and roles, project by project, can be defined by an administrator in the business user interface. This now includes an approver role that can review, compare and comment but not edit. The web based interface responds to the roles and completely integrates the review and approval task management needed by the workflow. Differences between versions can be reviewed visually in the business user environment as 7.1 has brought some nice existing visual compare capabilities into the review and edit context of the business user environment while merging of changes is supported in edit mode.
  • 7.0 brought most functionality into the Eclipse environment and 7.1 adds Decision Simulator configuration in the Eclipse environment. This allows simulation and impact analysis to be defined in Eclipse and then performed inside either the Eclipse environment or the web based environment. Additional reports have been defined as simulations have become an increasingly important part of approval for rule changes.
  • Finally the decision metaphors (decision tables, graphs and decision trees) have been reengineered so they do not generate the implied rules first and then execute them. This results in faster sequential execution 50-500x commonly but especially in bigger tables, graphs and trees where it can be 1,000x-3,000x. Particularly for typical FICO customers who have very large trees, tables and graphs this can be a big deal.

Decision Optimizer 6.3 has also just been released and this now supports the direct generation of a decision tree from optimization results (in addition to being to export the optimization results to a data mining tool to identify a decision tree). Such a tree can be imported directly into Blaze Advisor for execution, giving a set of rules that mimics the optimal assignment determined by the optimization algorithm.

You can get more information on FICO here and they are one of the vendors in our Decision Management Systems Platform Technology report.


Comments on this entry are closed.

  • Paul Vincent September 18, 2012, 2:46 am

    Congrats to the FICO team (their original decision table and tree implementation I recall was always considered very neat in the early 2000s – not that I am biased at all 🙂 ). Especially interesting to note the performance difference in decision execution going to a custom decision engine rather than mapping to rules. AFAIK of the main tool vendors (and I may be wrong) TIBCO and now FICO use custom decision table code generators / engines, and IBM and Oracle map to their rules engines still?

    Of course Progress Corticon and Bosch don’t have inference engines so must also use custom decision engines…