Business processes tend to be fairly stable things, specifying how information, money and materials flow throughout the organization. Decisions however tend to be quite volatile, and in industries where policy and regulatory changes are frequent, they can be very volatile. And so it makes sense to separate out processes from decisions, but at the same time provide integration mechanisms.
The traditional format for modeling and specifying decisions has been the decision table. Closely related to this is the decision tree, where decisions are represented in an easily understood tree format. These decision representations provide a good starting point for adding intelligence to business processes. Existing manual decision processes can be captured and embedded into trees and tables, which are in turn loaded into a business rules engine. This will deliver rules to applications as they need them, usually via some widely used protocol such as Service Oriented Architectures (SOA). Using a Business Rules Management System is crucial. It is not uncommon for organizations to find they have thousands of rules, and a rigorous rule management environment is crucial. A central repository of this nature means that business users can directly access rules (given relevant permissions) and that a rule can be defined once and used many times. Changing rules is also very easy, and usually doesn’t require programmers or other technical staff, but can be accomplished by someone with responsibility for business rule management.
Creating a business rules management facility has been repeatedly shown to boost productivity, agility and accuracy in decision making. However decision trees and tables are really just the tip of the iceberg. Predictive analytics has recently gained a significant profile and is already widely used for customer focused applications (churn reduction, upselling, targeted marketing etc.). Predictive analytics uses data mining methods and tools to trawl through historical data with the intention of finding patterns which might be useful in the future. Many types of technique can be used. In some industries it is crucial the way a model works can be understood by humans, and so methods such as clustering and decision trees prevail. Other techniques are available when this isn’t a requirement – such as neural networks and support vector machines. The crucial issue is that processes which can take advantage of these predictive models should have easy access to them. This where an integrated decision management platform becomes important, along with methods such as the Decision Model and Notation (DMN), and PMML (Predictive Model Markup Language).
Other techniques are also available for creating decision models. Optimization methods will create the best plan for resource deployment given a set of constraints and well defined objectives. Statistical methods are also increasingly employed in decision making. Regardless of the techniques used the resulting models need to be available to business processes and other activities (not all business activities can be captured in process) via well-defined interfaces.
The 80-20 rule tends to apply to the creation of decision capability within business processes. Eighty per cent of the benefit will come from twenty per cent of the rules, and unlike traditional business process initiatives, the returns from adding decision models to business processes can be very high – an order of magnitude higher.
The term iBPM (intelligent business process management) has recently been created to capture the added capability that decision models bring to processes. It gives the unfortunate impression that intelligence is some sort of add-on. The decision is a core unit of work almost every person in an organization executes day after day. Decisions need decision management tools and methods if they are to be managed effectively. The alternative is a half-baked solution that will lead to decision model anarchy with associated poor agility, productivity, transparency and performance. We have to be serious about decisions – their automation and management represent the next big jump in business efficiency and efficacy.
The previous article in this series is Business Process Management Meets Decision Management
The next article in this series is Intelligent Business Process Methods
Get the full report:
