Two Competing Decision Model standards or one Composite Model?


Subscribe
Suleiman Shehu
Posted by Suleiman Shehu on Jan 16, 2013

Decision Model Management

Welcome to our blog. This is an archived post, most of our knowledge and advice remain valid but some material or links may be outdated. Click here to see our most recent posts.

In order for Decision Management to be successfully implemented within an organization the following capabilities and technologies must be present:

  • A graphical modelling language that enables business decisions to be identified and all associated business decision logic identified or created and modified by the business in a language that they can understand. These models are known as decision models.
  • The business should be permitted to use business friendly names for the data that they want to use in their decision models. The use of a glossary is required to map these business friendly-names into the fact or object models known to IT. This mapping is done by IT concurrently with the creation of decision models by the business.
  • While the business is modelling the business decisions, IT can be busy re-designing business processes (using BPMN) to eliminate from process models any business logic that has been embedded in the process models from previous process design efforts. IT could then be work with the business to define the BPMN rule tasks that will consume the decision models created by the business – deployed as decision services.
  • Once the business has completed the modelling and testing of the decision models two things can now happen:
    • The decision models can be handed over to IT to manually convert into code that can run in any technology and deployed as decision services = that will be consumed by business process engines. Also IT will ensure that the business results, from the execution of each business decision, are logged and stored in the corporate data warehouse. The logged data can be used to drive key performance indicators dashboards.  Managers will now have the data to verify the effectiveness of each decision model which they can now iteratively refine as required.
    • Alternatively if the technology exists the decision models are automatically converted into code (including logging capabilities outlined above) that can be executed and deployed as decision services without a lengthy IT development cycle. In reality some intervention is always need by IT however with the deployment of Continuous Integration and Deployment technologies the business can make, modify and deploy their decision models automatically several times a day. This results in very high levels of business agility.

The decision model

In 2009 the seminal book “The Decision Model: A business logic Framework linking business and technology” by Barbara von Halle and Larry Goldberg was published. 

This book defined a graphical notation called “The Decision Model” that was easy to understand by business people as well as IT. The book also defined 15 principles that ensured that the business logic was without logical errors and was aligned with the business decision.

The Decision Model provides all the Decision Management capabilities outlined above, in fact The Decision Model (TDM) could be argued to be the mother of Decision Management. Figure I show the high-level structure of a single decision model and Figure 2 shows the underlying decision table containing the business rules for each of the green clipped Rule Family shape.

The-Decision-Modeler-skeleton-form_450x320

The_Decision_model_rule_families_img2

The interesting thing about The Decision Model is that it provides not only a graphical notation but a technique or methodology that can be used by the business to easily create decision model. Figure 3 shows how decision models created by the business can be linked and integrated with other models used by IT. It is should be clear by now that Decision Management requires a very strong collaboration between the business and IT.

The_Decision_model_integration_img_3

Figure 3 © Lux Magi Ltd

The decision model limitations

There is no doubt that The Decision Model (TDM) is an excellent business logic framework to get started in Decision Management implementation but it suffers from the following limitations:

  • The Decision Model cannot model decisions that use types of logic that cannot be represented by a decision table, such as analytical models, mathematical models, algorithms, etc
  • The Decision Model is focused on modelling a single business decision. Standard TDM (as defined in “the Decision Model book”) has nothing to say on how groups of decisions that are connected should be modelled. Although TDM has been expanded to cover these advanced topics, nothing has be published publically about these proprietary extensions to the standard TDM
  • The Decision Model blue octagon shape does not actually make the decision; the decision is made by the conclusion of the Decision Rule Family.  This means that there can only be one Decision Rule family.

Decision Model and Notation (DMN)

These limitations were solved when in February 2014 when the OMG introduced the “Decision Model & Notation” Beta 1 (DMN) standard which is expected to be ratified in December 2014.  The DMN standard is supported by many organizations including IBM, Oracle, FICO, KPI (inventors of The Decision Model).

Decision_Model_and_Notation_img_4

In the DMN decisions are shown as rectangles. And decision logic as clipped rectangles.

 

These are called Business Knowledge Model objects and can contain decision table logic or other types of logic such as analytical models.

 

Note the Decision Requirement Duagram (DRD) showing relationship between three decisions as well the BPMN business rule task that calls the routing decision, input data. The decision table shown is created at the decision logic level,

 

Figure 4© OMG - Taken from OMG DMN specification.

The DMN solves the above TDM limitations by providing the following capabilities:

  • A Decision Requirements Diagram (DRD) that enables the complex inter-connection between decisions to be modelled
  • Analytical models (using PMML) are an important part of many business decisions. The DMN standard permits decision models to be created that integrate analytical models, probabilistic & mathematical models, algorithms, application code, in fact any type of logic that can be represented by a function.
  • The DMN defined an Expression Language with a formal semantic model plus a Simple version of that language
  • In the DMN there is no limit to the number business model knowledge objects that can provide input to a DMN decision.  This requires that the DMN contains an expression or a decision table that will determine how the decision will combine all its inputs to produce a decision.

However the DMN suffers from the following limitations:

  1. The DMN is a standard primarily for tool vendors; it does not define any methodology for its use.
  2. The DMN does not have a rigorous formal logic for decisions based on based on decision tables.

So which decision model standard (TDM or DMN)  should one use

The solution is not to consider these two standards as competing standards but rather they should be considered as complementary standard that should be integrated into a Composite Decision Model. 

For example TDM would benefit from the DMN Expression language, the Decision Requirements Diagram whereas the DMN would benefit from TDM rigours logic principles. See Figure 5.

TDM_and_DMN_img_5Figure 5 © Lux Magi Limited

How to get started in Decision Management

The best way to get started with Decision Management is:

  • Conduct an on-site training course in The Decision Model and DMN and get as many people from the business and IT to attend the course (typically lasting 3 days duration). Best to hold the course in a hotel away from office distractions. An ‘on-site course’ is best (when compared with a public course) as it enable confidential corporate problems to be discussed as well as help to create a corporate-wide excitement for a subsequent BDM pilot project.
  • Conduct a short 30-day pilot proof-of-concept project. Get a business sponsor and select an area of the business that has some real pain and identify a business process that contains key business decisions. If possible, get some mentoring help to ensure that your decision model training is being applied correctly.  You should use a decision modelling tool to support your pilot project.

One of the advantages of the OMG DMN standard is that it has legitimised Decision Management. The fact that DMN is supported by heavy weight giants such IBM, Oracle, FICO and the inventors of The Decision Model – KPI should encourage more companies to undertake BDM projects.

Get BPMN Certified

SUBSCRIBE TO BIZZDESIGN'S BLOG

Join 10.000+ others! Get BiZZdesign's latest articles straight
to your inbox. Enter your email address below:

 

Subscribe to Email Updates

comments powered by Disqus