Strategic, tactical and operational business decisions all call for different decision-making processes and analytics techniques. There are two steps that data and analytics leaders should take to determine the role of data and analytics teams in improving each type of decision.
Key considerations
Data and analytics leaders pursuing strategies that support better business decisions should follow these two steps:
Reengineer business decisions by first investigating how each fit into the organization’s operating model and business processes. Look at the purpose of the decision, who makes the decision and what time constraints exist.
Examine the structure of the decision process and determine the role of data and analytics in decision making.
Background
Organizations make multiple kinds of decisions - each with different characteristics - so they need multiple approaches to decision making.
A strategic decision (such as deciding whether to add a new line of business to process credit cards) is made through a largely ad hoc decision process spanning months. These decisions often involve:
People from different parts of the organization, and perhaps some outside experts
Numerous meetings and emails
Iterative investigations and data collection
Multiple versions of spreadsheets or other analytics
By contrast, a real-time operational decision (such as whether to approve a customer’s credit card transaction) is executed entirely by a computer in a few milliseconds based on business rules and pretrained machine learning (ML) models. Between these extremes are tactical and operational management control decisions that involve various decision processes and analytics.
Data and analytics are used in different ways in each of these kinds of decisions. The role of the data and analytics team will therefore vary according to the different nature of the decisions.
In this article, “data and analytics teams” refers to business intelligence (BI) teams, data science teams, data and analytics centers of excellence (COEs), management science (that is, operations research) teams and similar groups.
Approach Strategic, Tactical and Operational Decisions Differently
Strategic decisions identify the objectives and principles of the organization and provide a broad outline of its priorities. Senior executives and boards of directors make these decisions — typically with the assistance of their staff, outside experts and the data and analytics team. Strategic decisions are made at a corporate level and at a departmental level in some departments (for example, IT strategy).
Examples of corporate-level strategic planning decisions include:
Should we redirect our sales strategy from brick-and-mortar to e-commerce channels?
Will we be a low-price and high-volume provider, or a high-price and high-quality provider?
Should we enter a new territory?
Do we target near-term profit or long-term market share?
Should we acquire a certain company that will change our product portfolio?
High-level strategic decisions provide guidance for thousands of tactical decisions that will fulfill the strategies (see Figure 1).
Decision Relationships
Figure 1. Decision Relationships
Tactical decisions define policies, business processes, product features and price lists. They also determine the algorithms (that is the decision criteria and parameters) for making subsequent operational decisions. They are metadecisions (decisions about decisions) that specify how operational decisions will be made.
For example, an organization may make the abstract, strategic decision to emphasize e-commerce rather than brick-and-mortar sales. This will trigger concrete projects to revise websites, redesign supply chains, repackage and reprice some products, adjust marketing campaigns and make other changes. Each of those projects will entail multiple tactical decisions such as:
What criteria will be used to personalize next best offers?
What should be our customer service procedures for each set of circumstances?
How do we configure complex products?
What will our credit approval policies be?
How can we reduce customer churn?
Operational decisions are the numerous day-to-day and minute-to-minute decisions that put tactical decisions into action. Some operational decisions are for management control (that is, managing resources that perform work). They are about obtaining, allocating, scheduling and managing people, machines, buildings, inventory and other resources. Management control decisions may be made by any level of management. For example, supervisors make the hiring decisions for workers, but senior executives make hiring decisions for high-level managers.
Examples of management control decisions include:
How many people should be scheduled for the contact center next week?
Should we hire this person?
Is it best to outsource these tasks?
Should we replace or repair this machine?
Is it time to order more supplies?
Production decisions are also operational decisions. These are made in the course of producing individual goods or services. They apply to specific instances of a business process. They may be part of one business transaction, customer interaction or activity for manufacturing or transporting goods. Production decisions are typically made by the resource (person, computer system or machine) that is performing the work. Many production decisions are automated. For example, determining the algorithm for personalizing customer offers is a tactical decision. However, the runtime execution of that algorithm, which generates an offer for a particular customer at a certain point in time, is a production decision that may be made by a software personalization engine.
Production decisions include things like:
Is this transaction an acceptable credit risk?
Should this order be shipped by ground or air?
What is the next best content to display on our website for this customer?
Is this person eligible to receive these welfare benefits?
Feedback on the outcomes of decisions is used to improve subsequent decisions. The results of production decisions are reported and used to modify production decision algorithms and, on an aggregate level, to improve strategic, tactical and management control decisions. Similarly, the results of management control and tactical decisions can form part of the input that is used to shape new strategic, tactical and management control decisions.
Senior executives do not get involved in minor operational or tactical decisions because their efforts need to be focused on the big picture. However, they do get involved in high-value operational and tactical decisions. Furthermore, some decisions affect multiple levels. For example, acquiring another company is a high-value management control decision because it deals with obtaining resources. However, it is also a strategic decision if it expands the product portfolio and takes the company into new markets.
The boundaries between these four types of decisions are not always precise. Table 1 summarizes common characteristics of decisions, but these are generalities with exceptions. Moreover, the increasing pace of business is compressing the timelines for making all types of decisions and reducing the duration of the actions that result from decisions. World and market events are driving companies to adjust their strategies and tactics more frequently. There is increased pressure on executives and managers to decide faster. Similarly, organizations can often save money or gain competitive advantage by responding to operational opportunities and threats earlier, or anticipating them before they happen.
Table 1: Common Characteristics of Decisions
| Decision subject | Decisions per year in a large company | Typical decision maker | Duration of decision-making process | Duration of resulting actions |
Strategic | Objectives, principles, priorities | <50 | Senior executives and boards of directors | Three to nine months | Years or decades |
Tactical | Policies, procedures, products, price list | 10,000s | Middle managers | Days or weeks | Weeks or months |
Operational Management Control | Obtaining and managing resources | Millions | Any level of management | Minutes, hours or days | Hours, days or weeks |
Operational Production | Producing and delivering goods or services | Billions | Workers or automated systems | Subseconds, seconds or minutes | Seconds, minutes or hours |
Use the Decision Structure to Determine the Role of the Data and Analytics Team
A decision is structured if an analyst can develop a decision model that documents the decision process and logic. In this context, we are talking about decision models that specify the input data, decision-making algorithm (meaning the recipe or set of formulas) and output. A decision model is different from other kinds of analytical models because it is prescriptive — its output specifies the response action to take. Other analytical models are descriptive or predictive. Organizations use different kinds of decision models depending on the circumstances.
A developer can build a system that automates a fully structured decision by writing code or using business rules, optimization, ML or other artificial intelligence (AI) techniques to implement the algorithm in software. However, a person could make a fully structured decision by adhering to a set of rules in a policy manual.
Unstructured means that the decision process and logic are unclear. A decision may be unstructured because the criteria are subjective or because the algorithm is too complex to model. An unstructured decision process is ad hoc. People invent the process as they learn more about the problem, or the algorithm is generated by AI software dynamically. People make choices using undefined criteria (such as judgment and gut feel).
Many business decisions are partly structured. The logic within some steps (or “subdecisions”) in the decision process may be understood and modeled while other subdecisions may be internally unstructured. All of the subdecisions may be identified in advance or people may create new subdecisions spontaneously. Analysts can build software components that calculate structured subdecisions to provide decision support to people who make the remaining, unstructured subdecisions.
For example, financial institutions that offer mortgages use automated systems to generate recommendations for whether to approve or reject mortgage applications (see Figure 2). The structured parts of this production decision involve subdecisions that evaluate criteria such as:
The applicant’s credit risk
The appraised value and other characteristics of the property
The proposed amount, down payment and term of the mortgage
However, the decision process, especially for large or unusual mortgages, typically includes a final, unstructured manual subdecision in which a human loan officer or committee considers the computer’s recommendation and other background information to make the final decision. This approach is decision augmentation — a type of decision support that uses prescriptive analytics to assist human decision makers.
Simplified View of a Multistage Mortgage Decision Process
Figure 2. Simplified View of a Multistage Mortgage Decision Process
Strategic
Strategic decisions are inherently relatively unstructured. They require substantial amounts of human analysis and judgment because the problem space has many uncertainties. Many issues are not visible at the beginning of the decision-making process, so the process takes shape dynamically over a period of months. Senior executives orchestrate the decision process, choose the decision criteria and make the final decision, subject to the approval of a board of directors. Data and analytics teams can play a significant supporting role by:
Providing analytical databases
Recommending and supporting analytical tools
Facilitating data literacy among executives and their staffs
They also may build and execute ML, optimization or other analytical models that implement some structured subdecisions (see Table 2).
Table 2: Determining the Role of the Data and Analytics Team
| Decision Subject | Degree of Structure | Typical Decision Maker | Role of Data and Analytics Team |
Strategic | Objectives, principles, priorities | Mostly unstructured, ad hoc, dynamic | Senior executives and boards of directors |
|
Tactical | Policies, procedures, products, price lists | Exploratory, dynamic, some structure | Middle managers |
|
Strategic decisions are made with the help of spreadsheets, analytics and BI platforms (with augmented analytics features) and other tools that support what-ifs and creative, ad hoc exploration of alternatives. They also emphasize user interface capabilities, such as visualization, search, storytelling and narratives that facilitate human decisions.
Tactical
Tactical decisions are more structured than strategic decisions and typically make substantial use of data warehouses, reports, ML and “what-if” analysis. They sometimes use simulation, optimization and other management science techniques. Data and analytics teams play a prominent role in determining tactical decision processes and providing data, tools and models used for decision support. They work closely with subject matter experts and business managers who provide business expertise.
Operational Management Control
Operational decisions are largely structured. Unlike strategic and tactical decision processes — which are inherently exploratory, iterative and dynamic in nature — most operational decision processes are repeatable. The decision process, and most subdecision logic, can be designed before a new instance of the decision is triggered. Most operational decisions don’t have the time or need to invent a modified decision process and algorithm every time a decision must be made. Operational decisions sometimes execute ML, optimization and other models but, unlike strategic and tactical decisions, people rarely develop new models when it is time to make a decision.
Managers and first-level supervisors make most operational management control decisions with decision support from dashboards, analytics and BI platforms, augmented analytics capabilities, and, sometimes, machine learning, optimization and other AI systems. Data and analytics teams work with business managers, business analysts and subject matter experts to provide the data, analytical tools and data literacy support for businesspeople and analysts. Some users develop their own self-service analytics but, in many organizations, a BI team still develops reports and dashboards, or a data science team develops ML and other models on behalf of the business units.
Operational Production
Operational production decisions are the most structured, repeatable and time-sensitive decisions. Production decisions often form part of an organization’s customer interactions, transaction processing systems, and manufacturing, supply chain and other operational processes. The decision logic may be embedded in an application or configured to run as a separate step in a business process. Not all production decisions are automated, but almost all fully automated decisions are production decisions.
Data and analytics teams play a different role in operational decisions than in strategic and tactical decisions because of the way analytics and rules are used. A data and analytics team’s role is to help application architects, business analysts, and application developers build systems that automate decisions or provide decision support.
Conclusion
Many data and analytics teams have limited or no experience with decision modeling, business rule processing and optimization. They may need to partner with business process modelers or management science experts who have expertise in these areas in order to create a virtual or real decision intelligence team. A decision intelligence team works with business managers, data engineers, ML engineers and application developers to model decisions and implement and deploy runtime software decision services.
Organizations are finding ways to automate more production decisions (and some management control decisions) by reengineering unstructured or partly structured decisions into fully structured decisions as part of their digital transformation and hyper automation initiatives. Automated decisions:
Reduce staffing costs.
Are more consistent and auditable.
Are made faster than decisions that have people in the loop.
Assuming the developer has implemented the analytics and rules well, an automated system will make fully structured decisions as well or better than a person can. They are also less likely to be biased than human decisions because their rules and analytical models are generally visible, even before the systems are deployed. However, some decisions can’t be automated because they need human judgment.