Business decisions should be treated and managed as fundamental entities in organizations across the enterprise. That means we must be able to represent them and specify their details.

There are many techniques to model a business decision, such as decision trees, influence diagrams, causal loop diagrams, Bayesian networks, etc., but they cannot capture the true nature of business decisions because of their complexity and versatility. Additionally, each of these modeling methods and techniques are relevant to a very specific problem context, such as cost and risk analysis, SWOT analysis, etc., and they cannot be used as a go-to modeling technique for a broader range of business decisions as a generic decision modeling practice across organizations.

There are, however, two techniques that can be used as a generic go-to modeling technique for business decisions:

  • The Decision Requirements Diagram (DRD) is part of the Decision Model and Notation (DMN) standard, which is used mostly for rule-driven decisions.
  • Decision Graph – An extended version of DRD with richer behavior yet simple and concise without making modeling too much of an effort. Complex decisions require combining different techniques such as machine learning, rules, optimization, and so on.

Decision Graph

A Decision Graph is a model that represents a multistep, hierarchical, and situation-aware business decision. It enables organizations to model complex, transient, and long-running business decisions.

To build a Decision Graph, you will use a decomposition technique, which means the below process:

  • Break down a complex business decision into its components and subcomponents (frame a business decision). Each of the components and subcomponents is called a Decision Unit.
  • Determine the dependencies between the decision units and create the link between them by connecting one decision unit to another.
  • Identify and specify the metrics of each decision unit in relation to the Business Decision's KPI.
  • Determine in what situations and circumstances each of those decision units are relevant.
  • Specify the Data Input requirements for each of those decision units.

By following the above 5-step process, you will create a hierarchical, multistep Decision Graph that represents a holistic view of a business decision.

This model not only visually represents a complex business decision but also, is the live specification of how a decision is made and executed.
Here's an example of a Decision Graph created using our Decision Intelligence Platform:

Example of a Decision Graph: Decision Requirement Diagrams (DRD)

Example of a Decision Graph modeling hierarchical decision with using Sub-Graph (light-blue nodes)

Multistep Decision

Almost every business decision is too complex to be made in one go. The outcome of one decision unit should be an input of another that is called Multistep Decision. The Decision Graph works out this behavior based on the dependencies of the decision units. Therefore, in a graph, decision units with no dependencies are activated, and their outcomes are passed to the next connected unit until the whole graph is executed and outcomes are created.

In a complex business decision, dependencies work better for decomposition, however, Decision Graph allows you to look at the complex business decisions in terms of order, which can represents a decision flow if that's what you prefer.

Composite and Hierarchical Model

Decision Model and Notation (DMN) is generally used for rule-driven decisions. It can also integrate PMML (score cards) as part of the decision. However, business decisions are typically far more complex. Decision Graphs allow you to build many different types of decision logic using business rules, machine learning, statistical analysis, optimization, data query and processing, activity flow, etc.

Each decision unit can be use different technique, and the Decision Graph coordinates between the units to create the outcome based on different types of logic. This is called Composite AI – a single model allows you to combine multiple techniques to execute a business decision.

What is interesting about Decision Graph is it enables hierarchical decisions. Means a Decision Graph can be based on other Decision Graph. This hierarchical view makes decision graph concise and easy to understand.

Long-Running Decision

As the complexity of business decisions grows, there are times that a decision outcome becomes inconclusive. Therefore, at the time of decision execution, the final outcomes cannot be created. This means the decision should be suspended and resumed later when it is possible to create the outcome. In some scenarios a domain expert should be able to provide more information to continue the decision execution. Or even override the decision outcome when needed.

In a long-running decision, the execution of the decision model can be temporarily suspended (or paused) based on some criteria or conditions and resumed based on some internal or external events or inputs.

Situation-aware Decision

In a Decision Graph, there are many connected decision units based on their dependencies. However, at any point in time, based on the input data to the graph, not all nodes are activated. This node activation makes the decision graph very effective as it can present a holistic view of business decisions (as opposed to a narrow view), and based on different circumstances, different nodes on the graph will be active for execution. Additionally, each decision node based on different circumstances may need a different set of data as their input for execution.

In contrast, in a Decision Requirements Diagram (DMN), all the nodes in the DRD are always active, and those circumstances are pushed to the rule's implementation. This makes rule implementation complex and couples the decision situation and rule logic.

Business Decision KPIs

In any business operation, the team cares about the specific outcomes and the impact of their work. When you model a business decision using the Decision Graph technique by decomposing complex business decisions into smaller, easier-to-understand decision units, you can specify metrics for each individual decision unit. This means that when executing a business decision using a Decision Graph, the outcomes of the decision are captured, and the metrics around those outcomes are collected.

This will enable you to understand the impact of each decision unit as part of the overall decision outcomes. With decision units' metrics you can simply understand how those units are impacting the overall result of what matters for the team in relation to the business KPIs and business objectives.

Book a Custom Demo

First or last name is too short





Conclusion

Decision Graph builds cohesive and concise model of business decisions where you capture and model everything about business decisions. Including the decision units, states, transitions, situations and techniques, information access, and so on.

This comprehensive method for modeling business decisions ensures that decision models are executable and manageable from different aspects and avoids a disconnected decisioning experience. It also creates a Composite AI model that integrates business rules, machine learning, optimization, and other techniques for decision execution.

Last updated July 9th, 2024 at 02:27 pm Published June 13th, 2024 at 10:43 am