Usage Scenarios and Roles

Activity

This represents the collection of simple definitions of a set of decisions and knowledge sources in a business area.

Use

In the simplest instance, this description could just be the decision names. These decisions can be mined from corporate processes, documents or systems, or they can be obtained by interviewing subject matter experts. Once captured, these decisions may be organized in a certain way (for example, using a diagram or spatial arrangement to group together related decisions), and they may be accompanied by metrics to identify high-value opportunities for further analysis or decision automation. Decisions are typically identified by name, the questions they raise and the possible answers.  The goal of this scenario is only to explicitly recognize, differentiate and define key decisions in terms of semantics and outcomes.

Business Value

The business benefits include: establishing clear goals, raising awareness of the key business decisions, improving consistency and reducing complexity and redundancy (e.g., through decision disambiguation and removal of duplicates). Because business analysis has historically been dominated by processes, the inventory process can often ‘discover’ implied decisions that were being made implicitly or taken for granted by companies. Once recognized, these decisions can be clearly formulated and managed.

Participants

Decision inventories may be conducted as group activities with decision owners and team members.

Output

Ultimately, decision inventories are about being explicit concerning decision-making and the degree of automation taking place.  They can be a simple list of decisions listed on a white board, a document or in a DMN Decision Requirements Diagram (DRD).

Audience

Decision inventories disclose to a larger audience what decisions are in place and their relationship to key business functions.  Managers and executives will understand their decisions’ impact on revenue, costs, efficiency and alignment. Decision inventories are therefore useful for prioritization. Enterprise architects will rely on these models for their planning activities while they map these decisions into processes and software assets.  Budget owners will understand resourcing aligned around each decision for planning.  Executives will understand which decisions directly align with strategic initiatives.

Activity

This is the need to draw a diagram to sketch out the basic relationships between a set of decisions and knowledge sources without necessarily adding any additional details.

Use

The business goals are to capture decision relationships and hierarchies in more detail than in the Decision Inventory scenario, and reason about the dependencies between them. Dependencies can include the need one decision has for data provided by another or the need a decision has for external sources of knowledge.

Business Value

This process also helps to detect accidental duplicates and opportunities for refactoring decisions to improve consistency.

Participants

This requires an analyst in consultation with subject matter experts (SMEs).

Output

Simple Decision Requirement Diagrams (DRDs) with decisions, relationships and annotations.

Audience

The broader team project team and enterprise architects.  Business leaders might prefer this level of DRD detail while looking for decision automation opportunities. These diagrams also help to explore relationships between different decision owners within a company, especially those owning one decision which is dependent on another owned by a colleague.

Activity

This scenario is about more detailed and organized documentation of decisions and knowledge sources. Detailed documentation can be used to ensure consistency, to improve communication with other stakeholders and for the purposes of training.

Use

This scenario is where decision hierarchies are captured in detail: decisions are broken down into sub-decisions, some of which may be reused. The business goals are deep understanding, dependency control, decision factorization, identifying reuse opportunities, improving decision consistency and staff education. Sometimes this process is used as the first step toward determining which decisions need to be automated and how to achieve this.

Business Value

Typically, this process captures the detailed intent of each decision and its business value as textual descriptions or examples. It ensures the barrier between one decision and another is well understood so that resources can be utilized effectively.

Participants

Decision analysts, architects, SMEs

Output

More complete Decision Requirements Diagrams (DRDs) complete with textual definitions and examples.

Audience

The broader team touching the full lifecycle of the Decision that is about to be modeled or defined.  Enterprise architects and business leaders looking for opportunities for investment.

Activity

Decision definition for the purpose of evaluating what input data is required to make a decision.

Use

This activity is essential to assist with data provisioning. In cases where the available data is in question, this activity can help us to understand how much automated decision making or decision support can be done with the available data and what impact provisioning additional data will have. This scenario can be useful even in the absence of automated decisions.

Business Value

The business value of data requirements capture is to understand the cost and impact of data provisioning (especially in cases where the data has to be purchased or obtained from many other systems). It addresses the data tradeoff: what data must I have to make the decision and how much of the decision can I make if I don’t have all the data?

Participants

Data source experts, data architects, data scientists and business owners focused on the same set of data that will drive the future of automation within the organization.

Output

Data models, Decision Requirements Diagrams (DRDs), architectural diagrams, conceptual models and vendor reviews.

Audience

Business leaders, data architects and other data team members.

Activity

Defining specific decision logic in the form of models, text, partially implemented artifacts such as decision tables or even executable languages.

Use

With logic expressed in this manner, decisions acquire some additional transparency, accountability, and rigor.

Business Value

The business benefit of this scenario is to capture detailed business knowledge (logic, intentions, vocabulary) about how complex decisions are made (e.g., to avoid loss of such information by staff churn and to provide a concise definition of decision semantics) and, potentially, to facilitate automation. It also captures the way in which decision hierarchies collaborate to achieve complex decisions.

Participants

Decision analysts, SMEs, and Decision Owners

Output

This is a Decision Requirements Diagram (DRD) with the inclusion of a declarative decision table or text definitions to some or all decisions in a hierarchy.

Audience

The broader team implementing the logic, decision developers and QA.

All decisions turn a set of data inputs into a defined outcome. In this respect, the outcome of each decision instance depends on its data inputs which represent an individual business scenario. However, Data-Driven decisions are those whose behavior depends on an entire dataset. For example, outlier detection of hospital stays to determine if a hospital length of stay is more than some standard deviation from the norm for a particular diagnosis. Typically, the datasets are the outcomes of earlier decisions or related historical data.  Data-driven decisions cover a spectrum of techniques, including but not limited to:

• Decisions dependent on statistical properties of a dataset (standard deviation, mean, inter quartile range, etc)
• Decision tables or trees induced from historical datasets (i.e., rule induction)
• Decisions featuring machine learning (ML) models trained on historical datasets
• Decision featuring generative AI components.
• Decisions using continuous learning (e.g., reinforcement learning, incremental stochastic gradient descent)

Activity

This is the integration of statistical parameterization, machine learning or constraint solving into decision-making. It allows decisions to learn from previous experience or analytical insight.  The combination of machine learning and deterministic decision making is powerful because it allows decisions to benefit from the lessons of historic data, allows them to adapt to new trends and affords an opportunity to improve the robustness of ML.

Use

Use historical data to influence decision design or train models and offset the need for manual entry of human expertise where this is onerous or impractical. Machine learning can introduce a statistical insight into decision-making of which decision tables alone are incapable.

Business Value

Decisions influenced by historical insights tend to yield more accurate and tailored outcomes and deliver more business benefit. In addition, decision modeling helps companies to understand precisely what they need from machine learning and this is vital to successful deployment. Decision models can also make ML deployments more robust by checking that the statistical models are being exercised within their limits and taking action otherwise.  Moreover, data-driven decisions result in analytical insight whereas other types of decisions represented as decision tables cannot. Finally, machine learning allows us to support decisions that learn from experience automatically, without continual manual adjustment.   

Participants

Data scientists, analysts, SMEs and decision owners. Data-driven decisions require different skills and almost always require a separate and distinct lifecycle from decisions authored by a human.

Output

The business goal is improved decision-making and business performance. The output is a Decision Requirements Diagram (DRD) with the inclusion of a declarative decision table or text definitions to some or all decisions in a hierarchy with integration of machine learning and optimization.

Audience

Because they are based on analytics, they provide significant value to business leaders, architects and data analysts.

Activity

This represents the need to expose the result of decisions as a reusable service with a well-defined, contractual interface.

Use

The public part of the service exposes its interface (inputs and output) benefitting users, while the internal part provides details on how the public part derives its result.

Business Value

The business benefit of this scenario includes the support for internal and external reuse of decisions.  Architects will be interested in these services from an asset perspective while they inventory candidates for orchestration, management and procurement for public/private consumers (API Management) and lifecycle visibility.  Business leaders may also take direct interest in decision services that result in revenue. Decision services also provide powerful integration points for generative AI such as large language models (LLMs).

Participants

Architect, Decision Developer

Output

DRDs containing services and possibly other artifacts not supported by DMN such as OpenAPI content.

Audience

Enterprise Architect, Business Leader, DevOps