About the Process Optimization tool

Basics

You can use the optional Process Optimization tool to perform probability analysis on assignments to identify common flow patterns and trends. For each assignment, the tool generates an array of outcomes showing the likelihood that a work item, given set of input conditions (properties), will result in a given output (flow action). The tool uses this information in its analysis:

• A data set of resolved work items within a defined date range.
• The assignment(s) you want to analyze. Choose an assignment that has two or more flow actions.
• The work item properties you want to use in the optimization analysis. Identify properties that are considered relevant to the outcome of the flow actions.

You can use the tool to automatically generate decision trees or decision tables based on the statistics. You can then employ the rules in Decision shapes, declarative expressions, or in activity calls to help automate and expedite processing in your design.

For instructions on using the tool, see How to use the Process Optimization tool.

How process optimization works

The optimization tool uses a statistical classification algorithm that examines, for a given assignment, a set of work item properties that you define, and sorts them into subsets according to associated flow actions. The subsets comprise property values that represent the most homogenous results (a subset that contains the highest number of similar values). The optimization process lists the subset conditions and tests them as true or false against the target flow action. In this way, the wizard discovers an historical flow pattern that you can save as if/then logic in a decision tree or decision table.

Optimization may be run on single value properties. You may not optimize using embedded properties.

Why use optimization

Business processes often have common flow patterns, but the large volume of data makes it difficult to identify those patterns. It may not be immediately obvious, given a variety of work item properties, why a certain group of work items are almost always processed the same way. The optimization tool helps you discover the patterns are and why they occur. With this information, you can streamline the process by eliminating, reconfiguring, or automating steps based on the anticipated outcomes.

For example, assume you are designing an auto insurance application process and that you want to optimize the assignment named Policy Decision, which has three resulting flow actions named Eligible (to Underwriting), Not Eligible (resolve), and Requires Review (to Review Application). You want to base the analysis values on four significant properties: .CustomerLevel, .CustomerAge, .VehicleLocation and .NumberOfViolations. The tool may discover that about 100% of applicants who share the following attributes are Eligible:

• Are at Standard or above level status
• Are older than 25 but younger than 50
• Garage their vehicles in a low-crime area
• Have no more than one driving violation

By saving these conditions as a decision tree or table rule, you can reduce work volume and eliminate operator handling in the Policy Decision assignment. Preceding it with a Decision shape using a decision rule generated by the wizard, the process automatically routes insurance applications with those properties directly to the Underwriting assignment.

Before you begin

Do the following before you use the optimization tool:

• Verify that the work type for the flow rule to be analyzed inherits from parent class Work- and is concrete (has saved work items).
• The property value pxFlowAction on the work item's history record must match those by the flow actions defined in the optimized assignment. If you replace or add a flow action, these work items will not be recognized by the tool.
• Because the Process Optimization wizard uses historical data to predict future behavior, using a large set of relevant data provides more accurate results. However, optimization of large data sets may require long run times. Consult your system administrator for guidance on JVM memory setting adjustments that can help reduce processing overhead.
• Familiarize yourself with the process. Identify the properties that are input or updated at each assignment and, of those, the ones that are most relevant to the analysis. For example, assume you have a restaurant reservation flow and are analyzing an assignment called Make Reservation that has three flow actions, the outcome of which depends upon the number in the party and the date requested. Use these properties for optimization. Including properties such as the party's phone number or names input earlier in the process are not useful in determining the assignment's outcomes and will slow processing time.

As a best practice, re-run the tool on a regular basis to ensure that as new data is added, the statistics represent an up-to-date pattern.

	decision shape
	How to use the process optimization tool Flow form — Editing in Process Modeler — Interacting with the Diagram tab About Decision Trees Property-Map-DecisionTree method
	Atlas — Standard Decision Trees

Tools — Process and Rules