Decision Making

Chapter learning objectives

1. Classify a decision by its structure, information needs, authority, and context.
2. Evaluate dashboard measures and decision-model assumptions through examples.
3. Explain cognitive biases and assign human and system responsibilities.
4. Distinguish capturing, sharing, locating, and applying organizational knowledge.

Goals for this part

• Explain Decision Types.
• Explain Problems and Information Needs.
• Explain Managerial Roles and Decision Context.

Decision Types

A structured decision has a well-defined procedure for reaching an answer, such as applying an agreed reorder rule. An unstructured decision requires substantial judgment because the problem, criteria, or alternatives are not fully specified. Semi-structured decisions combine defined analysis with judgment.

Decision Types

These categories describe the decision, not the intelligence or seniority of the person making it. An employee can face an unusual customer exception requiring substantial judgment, while a senior manager can approve a routine action under a clear rule.

Decision Types

For the service manager, calculating how many requests arrived yesterday is structured. Choosing how to redesign the service involves judgments about customer needs, staffing, costs, and acceptable waiting time. Software can support calculations and comparisons without deciding every value or tradeoff. The degree of structure can also change as the organization learns and establishes procedures.

Decision Types

Decision making is the process of choosing and carrying out a course of action among alternatives. A decision support system supplies information, models, or interaction to assist that process. It does not have to automate the final choice.

Decision Types

The system's usefulness depends on the decision and the person using it, rather than on the presence of an impressive analytical feature.

Problems and Information Needs

A decision starts with a statement of the problem and the outcome sought. A symptom is an observed condition; a cause explains how it arose. Longer waiting times could reflect greater demand, absent staff, complicated cases, or rework. The same symptom can justify different responses depending on the cause.

Problems and Information Needs

A practical decision process identifies the condition, develops and compares alternatives, chooses an action, implements it, and reviews the result. These activities can repeat when new information changes the understanding of the problem.

Problems and Information Needs: comparison

Problems and Information Needs

The table makes information requirements follow the decision rather than the available dashboard.

Managerial Roles and Decision Context

Managers coordinate work, allocate resources, handle exceptions, communicate priorities, and interpret changing conditions. Their information comes from formal records and from conversations, observation, and experience. A system can improve access while still omitting important context.

Managerial Roles and Decision Context

Decision quality also depends on authority and timing. A service supervisor may see a staffing shortage but lack authority to change shifts. A senior team may approve a change after the relevant demand peak has passed. Better information alone cannot resolve those organizational constraints.

Managerial Roles and Decision Context

For the service problem, analysis should identify who can change staffing, who owns the process, and who is affected by the alternatives. It should also state constraints, such as available skills or service obligations. A recommendation becomes useful when an authorized person can act on it within the required period and obtain feedback about its result.

Goals for this part

• Explain Dashboards and Business Intelligence.
• Explain Decision Support Models.
• Explain Measures and Performance Feedback.

Dashboards and Business Intelligence

Building on Chapter 6, the management question is how to interpret measures for action. A key performance indicator is a selected measure connected to an important objective. Selection implies a judgment about what deserves attention; it is not a property automatically created by displaying a number prominently.

Dashboards and Business Intelligence

The service manager might monitor the proportion of requests completed within the promised time. To compare teams, the definition needs the same starting event, completion event, eligible requests, and treatment of exceptions. A team handling more difficult cases can appear worse under an unadjusted average.

Dashboards and Business Intelligence

A dashboard should help a user investigate a condition rather than conceal its context. Showing a trend and the relevant case mix may be more useful than a single red or green signal. Managers should also inspect whether a measure encourages the intended behavior.

Dashboards and Business Intelligence

A target for closing requests quickly can encourage premature closure if resolution quality is not examined.

Decision Support Models

A decision support model represents relationships relevant to comparing alternatives. A staffing model might estimate capacity from available hours and average handling time. A what-if analysis changes an assumption to examine its consequences. A sensitivity analysis examines which assumptions most affect the result.

Decision Support Models

Suppose a team has 60 staff-hours available and an assumed handling time of 15 minutes per request. Its simple estimated capacity is 240 requests. If the average is 20 minutes, the estimate falls to 180. The calculation does not include breaks, uneven arrivals, or differences among requests, so it should not be treated as a guaranteed service capacity.

Decision Support Models: comparison

Decision Support Models

The table shows that the handling-time assumption can change the recommendation. It creates a reason to investigate that assumption before committing resources.

Measures and Performance Feedback

A measure supports improvement when it connects an action with evidence about the outcome. Leading indicators concern conditions expected to influence later results; lagging indicators describe results already observed. The relationship should be justified rather than assumed from the label.

Measures and Performance Feedback

Training completion may be a leading measure for a service change, while correct resolution is an outcome measure. Completing training does not prove that employees can handle realistic cases. A stronger feedback process connects preparation, observed work, and customer outcomes.

Measures and Performance Feedback

Managers should define a baseline, the intended change, the review period, and plausible alternative explanations. If waiting time falls after a new tool is introduced during a period of lower demand, the tool may not explain the entire improvement. Feedback helps refine the decision when it includes both favorable results and evidence that challenges the original assumptions.

Goals for this part

• Explain Human Judgment and Organizational Context.
• Explain Bias and Decision Quality.
• Explain Human and System Responsibilities.

Human Judgment and Organizational Context

Judgment is the reasoned interpretation of evidence and alternatives when a procedure does not settle every question. It is necessary when objectives conflict, information is incomplete, or consequences extend beyond what the model represents. Judgment can improve a decision, but experience alone does not make it error-free.

Human Judgment and Organizational Context

The service manager may know that a temporary event changed demand or that a proposed shift requires skills the available staff do not have. Those facts can qualify a model's recommendation. The manager should state the reasoning so that others can examine it rather than invoking experience as an unreviewable answer.

Human Judgment and Organizational Context

Organizational context also shapes which alternatives are considered. A department may favor a proposal that protects its budget while shifting work elsewhere. A sound decision process makes those effects visible and includes the people needed to understand the whole outcome. Analytical evidence and organizational judgment should challenge and inform one another.

Bias and Decision Quality

Confirmation bias favors evidence that supports an existing view. Anchoring gives disproportionate influence to an initial value or proposal. Automation bias gives excessive weight to a system output because it appears automated or authoritative. These tendencies can affect how people interpret evidence, including evidence from sophisticated models.

Bias and Decision Quality

A manager convinced that staffing is the cause of every delay may overlook rework. A model's first cost estimate may become a reference point even after its assumptions are questioned. Simply adding more data does not necessarily correct the way the decision is framed.

Bias and Decision Quality: comparison

Bias and Decision Quality

The responses support better reasoning; they do not guarantee the correct outcome in every case.

Bias and Decision Quality

A cognitive bias is a systematic tendency in perception or reasoning that can influence judgment. Bias is not the same as a deliberate dishonest choice, and identifying a named bias is not enough to explain every disagreement. The decision record should show the evidence and reasoning at issue.

Bias and Decision Quality

This makes it possible to examine whether a proposed correction improves the decision rather than merely assigning a label to the person.

Human and System Responsibilities

A division of responsibility should identify what the system calculates, what people interpret, who authorizes action, and how errors are corrected. “Human in the loop” is an incomplete description unless the person has relevant information, sufficient time, and authority to disagree.

Human and System Responsibilities

For a staffing recommendation, the system might estimate workload under several assumptions. A supervisor checks skills, constraints, and unusual conditions. An authorized manager approves the schedule. Later review compares the actual service with the intended result. The decision record should preserve important assumptions and reasons for departures from the recommendation.

Human and System Responsibilities

Responsibility also includes deciding when the system should not be used. Missing data, a changed process, or an unfamiliar case may exceed the evaluated scope. A clear escalation route allows staff to obtain help rather than forcing every problem through a misleading score.

Human and System Responsibilities

The purpose is informed action with a traceable rationale, not automatic agreement with either a person or a model.

Goals for this part

• Explain Capturing Knowledge.
• Explain Sharing and Locating Expertise.
• Explain Reusing Knowledge in Work.

Capturing Knowledge

Explicit knowledge can be expressed in records such as instructions, explanations, and designs. Tacit knowledge is embedded in experience and practice and may be difficult to articulate fully. The distinction explains why a document can be useful without reproducing everything an experienced colleague knows.

Capturing Knowledge

Knowledge management supports acquiring, organizing, sharing, and applying relevant organizational knowledge. A useful capture process records the problem, context, reasoning, action, result, and limits of a solution. A short instruction without its conditions may be easy to reuse incorrectly.

Capturing Knowledge

For the service team, a resolved case can become a knowledge record when staff explain why the solution worked and when it should not be used. Capture should be part of a feasible workflow. Requiring long reports for every routine action can discourage contribution without creating proportionate learning value.

Sharing and Locating Expertise

Knowledge becomes available through search, classification, collaboration, and access to knowledgeable people. Taxonomy organizes concepts into a structured set of categories. Tags and metadata can improve discovery, but inconsistent labels and outdated records can make a large repository difficult to use.

Sharing and Locating Expertise

An expertise directory helps staff find people with relevant experience. It should describe the expertise, context, and a practical way to make contact. A list of job titles does not necessarily identify who can resolve a specific problem.

Sharing and Locating Expertise: comparison

Sharing and Locating Expertise

The table connects forms of knowledge to ways of making them usable. Collaboration can complement a repository when a question requires interpretation rather than another keyword search.

Reusing Knowledge in Work

Knowledge reuse applies existing experience or recorded guidance to a current task. The user needs to evaluate whether the context matches. A solution developed for an older product or a different customer arrangement may require revision.

Reusing Knowledge in Work

The service team can connect relevant guidance to the point where employees encounter the problem. It should also provide a way to report that the guidance failed or no longer applies. Review and retirement of obsolete material are as important as adding new content.

Reusing Knowledge in Work

An AI search or summarization tool can help locate material, but the underlying record still needs an authoritative source, current status, and an owner. A generated answer can omit the condition that makes a solution appropriate.

Reusing Knowledge in Work

Managers should evaluate successful application and reduced repeated mistakes, rather than treating the number of uploaded documents as proof that organizational knowledge has improved.

Reusing Knowledge in Work

An expert system represents specialized knowledge through explicit rules and an inference process. It differs from a repository that stores documents and from a model trained to generate language. A rule-based diagnostic aid can be useful when its knowledge is appropriate and maintained, but its conclusions remain limited by the encoded rules and inputs.

Reusing Knowledge in Work

Choosing a knowledge tool requires understanding those differences rather than treating every form of automated advice as the same technology.

Reading and assignments

• Chapter reading
• Curated references
• Preview, Case Study, and Homework