In previous articles I introduced Business Decision Management (BDM), answered some of the most frequently asked questions about it and discussed operational business decisions. BDM involves the automation of operational business decisions and this leads us to the topic of this article – what technology do you need (and can you use) to implement BDM?
The technology for automating operational decisions using BDM falls into four main categories – a Decision Execution Environment, a Business User Environment, an Analytic Environment and a Data Platform. All of these must work together, though many of them can have standards-based and fairly “arms-length” relationships. In addition these elements must fit into the broader platform used to develop the composite applications and business processes for which the decision services being created with BDM – the building blocks of automated decisions – are required.
Decision Execution Environment
The first critical element is some kind of Business Rules Management System (BRMS) or Decision Management Platform (DMP). This must be capable of taking a declarative definition of how a decision should be made and executing that decision as and when required. This typically means an ability to execute business rules or business rules-based metaphors like decision trees and decision tables. The environment must plug into your event processing/business process management environment – though it can use web service interfaces – and it must also execute on your systems platform of choice. DMPs and some BRMS also allow for the execution of analytic models as part of a decision, not just business rules, and these will support some BDM scenarios more effectively than those that don’t.
Decision design is typically linked to the execution platform in the sense that most BRMS and DMP products have a development environment – an IDE – that is used to define the decision, its data inputs and outputs, the rules behind the decision and more.
Business User Environment
One of the critical elements of BDM is the engagement of business users in the definition, management and testing of decisions. Most BRMS and DMP products provide separate environments targeted at less technical users to enable this. These rule or decision maintenance applications are typically web-based and expose some of the logic of the decision so that it can be managed, updated and controlled by business users. Increasingly these environments also support simulation so that business users can see what impact a change might have before putting it into production and analysis and testing of updates to help business users manage the whole lifecycle of the decision.
DMPs, and some BRMS, also support adaptive control explicitly, allowing a business user to define multiple approaches or strategies for a decision and then reporting on the relative effectiveness of the different approaches. Even where no explicit support is provided, most BRMS products can be used to develop adaptive control or champion/challenger infrastructure. For more sophisticated BDM scenarios this support for champion/challenger is important.
Analytic Environment
While DMPs and some BRMS execute predictive analytic models, the development of those models typically takes place outside the product in an Analytic Environment. The analytic environment typically consists of a variety of data mining and predictive analytic tools. Some tools are fully integrated suites, others point solutions good for particular kinds of analytic work. These tools are designed for analytically sophisticated users – statisticians and “quants” – to analyze large volumes of data and develop analytic insight into that data. Analytic and data mining tools increasingly offer the ability to push the models they develop into the execution platform to streamline deployment. Some DMPs offer analytic model development tools but most focus on the consumption and execution of models produced elsewhere.
Many companies getting started with BDM may find it easier to outsource the analytic model development piece of the puzzle and simply provide data to and take models from an outsourced provider. This avoids the need to install and use an Analytic Environment.
Data Platform
Last, but by no means least, there is a need for an effective Data Platform. Decision performance management and reporting against the decisions to see how well they worked, what the economic impact of changes was and how the current decision performance is mapping to established KPIs is part of this. However, the data for analytics and data mining must also be available and usable and this can be high volume. Data Warehouse and relate infrastructure components like Extract-Transform-Load must support the operational data required by the decisions and either the information quality/data management processes of the organization must support the non aggregated data required for analytics or a separate analytic data process must be established to provide clean, usable data for analytics.
All these pieces need to work together to deliver effective BDM. Standards for BDM are beginning to emerge and BDM can take advantage of existing general purpose standards. For instance, most BRMS can deploy a standard web service and allow for web services to be called from within rules. Analytic models can sometimes likewise be deployed as services, allowing them to be used by the Decision Execution Platform. Additionally, the Predictive Model Markup Language (PMML) is increasingly well established as a way to move predictive models from one environment to another and rules standards such as the Rule Interchange Format (W3C) and the Production Rule Representation (OMG) are well underway. In a short article such as this only a brief overview can be given but there is lots more on all of these components on this site and others.
