Every decision is influenced by regulation, market dynamics, and the data & information, so it requires constant monitoring and updating. The law and regulations change over time, in some industries more frequent than the others. Market dynamics change as new customers and consumers demand more, and competition matures and expands. The companies’ access to data and information for different situations changes every day. Therefore, keeping decisions updated and aligned with business objectives is necessary as they determine the quality of business operation.
A single undivided, complex decision is difficult to update if not impossible in some situations. This behavior leads to inconsistency and inaccuracy of decisions since it becomes complicated to update and incorporate all the situations and parameters into a single decision logic or algorithm. As a result, this inconsistency and inaccuracy lead to compliance and audit problems. On top of this, it will increase the reputational risks of providers, e.g. lenders or insurers in the market. Subsequently, it will lead to losing a business to competition and loss of the market share.
The multistep decision is a technique to divide a complex decision into lesser complexity and manage them individually and enable companies to build composable, manageable, complex business decisions.
How does a Multistep Decision work?
A multistep decision or multistep decisioning refers to a technique where a complex decision in decision-making processes is divided into multiple simpler, more understandable, reusable decision components with lesser complexity by applying the “Decision Decomposition” pattern. These less-complex decisions then work in collaboration to satisfy the required outcome of the complex decision.
Multistep decisioning is a very effective technique to reduce the modeling complexity of a decision and its decision logic (algorithm) and enable the complex decision to incorporate different methods and technologies to deliver the required outcome.
Process vs. Graph
There are two distinct methods in applying a decision decomposition pattern to a complex decision to produce a multistep decision model: process approach or graph technique.
Traditionally, the process approach seems natural, simply because many people are familiar with this line of thinking. In process modeling, it is straightforward to add conditional variations and create an overall process or procedure that involves multiple decisions to produce an outcome. Also, this approach enables organizations to use another technology stack to implement multistep decisioning. For instance, if you have a BPM solution, you can use it as an orchestration mechanism to build a multistep decision.
With the flexibility of the process approach, comes a very big drawback. It is called “disconnected decisioning” which has multiple angles:
- Disparate branching conditions: multiple branching conditions in the process model routes to different decisions; rather, those conditions must be part of the decisions themselves and make them self-aware of situations and self-sufficient. This leads to making decisions less reusability and introduces conditions duplication.
- Disparate context: A decision must satisfy all stages of the decision cycle, which using other platforms to build the context make decisions impossible to fulfill this requirement.
- Convoluted logic: As the process approach can do more than orchestrating decisions, it becomes tempting to use shortcuts and implement other types of logic i.e. data logic, application logic, security logic… as part of the process model rather than focusing only on decision orchestration. This leads to a big, unmanageable process model that cannot be updated as frequently as you need to change decisions.
- Unnatural sequencing: The nature of business decisions is not based on order but is based on dependencies between other business decisions’ outcomes. In the process approach, the nodes (or steps) follow each other, determining the order of execution. In any business decision, even with the slightest complexity, this leads to a problem of finding the correct prerequisites of decisions execution and finding the right order will become a big problem and error-prone.
And as a result, the process approach tends to deviate from all the significant outcomes that we wanted to achieve in the first place. It contradicts why we started managing decisions separately from processes and applications by recognizing them as self-existence entities and concepts.
The graph is a combination of different types of nodes (vertex) and transition lines (edge) connecting them.
In this particular type of modeling, decomposing a complex decision is very simple and has many advantages over the process approach.
A decision model using the graph technique is more natural for composable decisions for decision scenarios because business decisions by nature should satisfy the requirements of dependencies.
A graph of business decisions models the dependencies between multiple decisions, the dependencies between data requirements and decisions, and the dependencies between various graphs of decisions. This approach eliminates all the problems of disconnected decisioning explain above. Also, it offers a straightforward, easy-to-understand and explains the why’s behind the decisions’ outcomes.
Multistep decision divides a complex decision into multiple less complex, self-sufficient, reusable business decisions and putting them back as the decision-making process to deliver the decision outcomes. Despite the name of “decision-making process”, the process approach introduces significant drawbacks and contradicts the whole idea of separating the decisions from processes and applications to build consistent and accurate business decisions and make it easier to adapt to required business and regulatory changes.
Published July 15th, 2021 at 05:02 pm