Decision Making

Learning Objectives

  • Understand the rational decision-making process.
  • Identify heuristics and biases that affect decision making.
  • Differentiate between programmed and non-programmed decisions.
  • Recognize the conditions under which decisions are made (Certainty, Risk, Uncertainty).
  • Apply quantitative tools like Payoff Matrices and Decision Trees.

The Decision-Making Process

Engineering managers are essentially paid to make decisions. The quality of a manager's decisions directly dictates the success or failure of their team and projects. A rational decision-making process involves these sequential steps:

Decision Making

The logical process of identifying a problem or opportunity, generating alternatives, and choosing the optimal course of action based on the values, preferences, and available data of the decision-maker.

Rational Decision-Making Steps

  1. Define the decision: Describe the gap between the current and desired conditions and distinguish known facts from assumptions.
  2. Set criteria and constraints: Identify the relevant measures (such as cost, time, safety, and quality) and any requirements an alternative must satisfy.
  3. Develop alternatives: Generate feasible courses of action and gather evidence relevant to each.
  4. Evaluate alternatives: Compare options against the criteria, constraints, uncertainty, and effects on affected parties.
  5. Select and document: Record the chosen option, rationale, decision authority, and assumptions.
  6. Implement: Communicate the decision and carry it out within assigned authority.
  7. Review effectiveness: Compare actual results with the intended objective and revisit the decision when evidence or conditions warrant.
Rational Decision Review

Classify the decision, select an evidence method suited to the uncertainty, check hard constraints, and review the result.

Use an existing rule only when it is authorized and actually fits the case. Expected value requires supportable probability estimates and does not replace safety, contractual, ethical, or other mandatory constraints. Under uncertainty, state the payoff-matrix criterion before comparing alternatives. Review outcomes and revisit assumptions when results miss the intended objective.

Rational Decision ReviewClassify the decision, select an evidence method suited to the uncertainty, check hard constraints, and review the result.. Use an existing rule only when it is authorized and actually fits the case. Expected value requires supportable probability estimates and does not replace safety, contractual, ethical, or other mandatory constraints. Under uncertainty, state the payoff-matrix criterion before comparing alternatives. Review outcomes and revisit assumptions when results miss the intended objective.. Start → Define the decision, objective, constraints, and affected parties; Define the decision, objective, constraints, and affected parties → Does an authorized rule fit this recurring case?; Does an authorized rule fit this recurring case? — Yes → Apply the rule within its stated scope; Does an authorized rule fit this recurring case? — No → Gather relevant evidence and develop feasible alternatives; Apply the rule within its stated scope → Implement within assigned authority; Gather relevant evidence and develop feasible alternatives → Set mandatory constraints and weighted decision criteria; Set mandatory constraints and weighted decision criteria → Can outcome probabilities be reasonably supported?; Can outcome probabilities be reasonably supported? — Yes → Use a decision tree and expected value with stated assumptions; Can outcome probabilities be reasonably supported? — No → Compare a payoff matrix using a declared uncertainty criterion; Use a decision tree and expected value with stated assumptions → Compare alternatives against evidence, criteria, and constraints; Compare a payoff matrix using a declared uncertainty criterion → Compare alternatives against evidence, criteria, and constraints; Compare alternatives against evidence, criteria, and constraints → Does an alternative satisfy every mandatory constraint?; Does an alternative satisfy every mandatory constraint? — Yes → Select and record the rationale and assumptions; Does an alternative satisfy every mandatory constraint? — No → Revisit the objective, evidence, or alternatives; Revisit the objective, evidence, or alternatives → Gather relevant evidence and develop feasible alternatives; Select and record the rationale and assumptions → Implement within assigned authority; Implement within assigned authority → Review outcomes against the intended objective; Review outcomes against the intended objective → Did the decision meet objectives and constraints?; Did the decision meet objectives and constraints? — Yes → Close decision review; Did the decision meet objectives and constraints? — No → Gather relevant evidence and develop feasible alternatives

Start → Define the decision, objective, constraints, and affected parties; Define the decision, objective, constraints, and affected parties → Does an authorized rule fit this recurring case?; Does an authorized rule fit this recurring case? — Yes → Apply the rule within its stated scope; Does an authorized rule fit this recurring case? — No → Gather relevant evidence and develop feasible alternatives; Apply the rule within its stated scope → Implement within assigned authority; Gather relevant evidence and develop feasible alternatives → Set mandatory constraints and weighted decision criteria; Set mandatory constraints and weighted decision criteria → Can outcome probabilities be reasonably supported?; Can outcome probabilities be reasonably supported? — Yes → Use a decision tree and expected value with stated assumptions; Can outcome probabilities be reasonably supported? — No → Compare a payoff matrix using a declared uncertainty criterion; Use a decision tree and expected value with stated assumptions → Compare alternatives against evidence, criteria, and constraints; Compare a payoff matrix using a declared uncertainty criterion → Compare alternatives against evidence, criteria, and constraints; Compare alternatives against evidence, criteria, and constraints → Does an alternative satisfy every mandatory constraint?; Does an alternative satisfy every mandatory constraint? — Yes → Select and record the rationale and assumptions; Does an alternative satisfy every mandatory constraint? — No → Revisit the objective, evidence, or alternatives; Revisit the objective, evidence, or alternatives → Gather relevant evidence and develop feasible alternatives; Select and record the rationale and assumptions → Implement within assigned authority; Implement within assigned authority → Review outcomes against the intended objective; Review outcomes against the intended objective → Did the decision meet objectives and constraints?; Did the decision meet objectives and constraints? — Yes → Close decision review; Did the decision meet objectives and constraints? — No → Gather relevant evidence and develop feasible alternatives

  • Start: terminator
  • Define the decision, objective, constraints, and affected parties: process
  • Does an authorized rule fit this recurring case?: decision
  • Apply the rule within its stated scope: process
  • Gather relevant evidence and develop feasible alternatives: process
  • Set mandatory constraints and weighted decision criteria: process
  • Can outcome probabilities be reasonably supported?: decision
  • Use a decision tree and expected value with stated assumptions: process
  • Compare a payoff matrix using a declared uncertainty criterion: process
  • Compare alternatives against evidence, criteria, and constraints: process
  • Does an alternative satisfy every mandatory constraint?: decision
  • Select and record the rationale and assumptions: document
  • Revisit the objective, evidence, or alternatives: process
  • Implement within assigned authority: process
  • Review outcomes against the intended objective: process
  • Did the decision meet objectives and constraints?: decision
  • Close decision review: terminator

Bounded Rationality and Heuristics

While the rational decision-making process is ideal, human beings rarely make perfectly rational choices due to cognitive limitations and time constraints, a concept known as Bounded Rationality.

Heuristics and Biases

Managers often rely on heuristics (mental shortcuts or "rules of thumb") to quickly process complex information. However, these shortcuts can lead to cognitive biases:

  • Availability Heuristic: Basing decisions entirely on information that is readily available in memory (e.g., heavily weighting the risk of a recent, highly publicized bridge collapse, even if statistically rare).
  • Confirmation Bias: The tendency to seek out only information that explicitly supports one's pre-existing beliefs while ignoring completely contradictory data.
  • Sunk Cost Fallacy: Continuing to invest heavy capital into a failing engineering project simply because a massive amount of money has already been spent, rather than cutting losses based on future utility.

Group Decision Making

Engineering problems are often too complex for one individual, requiring collective expertise. Group decision-making leverages diverse perspectives but must be managed carefully to avoid "groupthink."

Group Decision-Making Techniques

  • Brainstorming: A highly interactive technique to generate an enormous volume of creative alternatives in a short period. Criticism of ideas is strictly forbidden during the generation phase to encourage wild, "out-of-the-box" thinking.
  • Nominal Group Technique (NGT): A structured method in which individuals first record ideas independently, then share and rank them; this can reduce the influence of dominant voices.
  • Delphi Technique: An iterative method that collects expert judgments through rounds of questionnaires and feedback. Anonymity can reduce some social pressure, but it does not guarantee consensus or eliminate bias.

Types of Decisions

Managers face a spectrum of decisions, categorized broadly by their frequency and complexity.

Programmed vs. Non-Programmed Decisions

  • Programmed Decisions: Routine, repetitive, and highly structured decisions. Procedures, rules, or policies are already in place to handle them. The information needed is readily available.
    Example: Approving standard employee leave requests or reordering raw materials when inventory hits a specific par level.
  • Non-Programmed Decisions: Unique, one-shot, and completely unstructured decisions. Information is often ambiguous or incomplete, requiring custom-tailored solutions and high-level judgment.
    Example: Deciding whether to acquire a competing engineering firm or selecting a site for a new, first-of-its-kind manufacturing plant.
Programmed and Non-Programmed Decisions

Classify decisions by whether a repeatable procedure fits the case, not only by how consequential the choice is.

A novel decision may use parts of a policy, but incomplete criteria or new conditions call for explicit analysis and judgment.
Programmed
  • Recurring situation
  • Known criteria or authorized procedure
  • Routine evidence and review
Non-Programmed
  • Novel or substantially changed situation
  • Criteria or outcomes are uncertain
  • Custom analysis and documented judgment

Decision-Making Conditions

The level of information available to the manager dictates the condition under which the decision is made.

Conditions of Decision Making

  • Certainty: The decision-maker treats the outcomes of alternatives as known for the stated model.
  • Risk: Outcomes are uncertain, but probability estimates are available and defensible enough for the analysis.
  • Uncertainty: Probabilities cannot be supported well enough to use as known inputs; the decision-maker states which criterion will guide comparison.
What Is Known About the Outcome?

Separate certainty, risk, and uncertainty by the information available about outcomes and probabilities.

These are modeling conditions for a decision, not guarantees about the real world. Probability quality depends on evidence, assumptions, and the decision context.
Certainty — outcomes treated as known
Risk — outcomes uncertain; defensible probabilities available
Uncertainty — probabilities not supportable from available evidence

Quantitative Tools for Decision Making

When making decisions under risk or uncertainty, quantitative models can make assumptions and trade-offs explicit. They do not remove judgment or bias and should not override mandatory constraints or qualitative effects.

Payoff Matrices

A Payoff Matrix is a table that displays the possible financial payoffs (profits or costs) for different alternatives under various unpredictable future events (called "states of nature").

When operating under strict Uncertainty (no probabilities known), managers use psychological strategies to pick an alternative from the matrix:

  • Maximax (Optimist's Choice): The manager assumes the best possible state of nature will occur and chooses the alternative that yields the absolute highest maximum payoff.
  • Maximin (Pessimist's Choice): The manager assumes the worst possible state of nature will occur. They look at the worst possible outcome for each alternative and choose the "best of the worst" to guarantee a minimum baseline outcome.
  • Minimax Regret: The manager seeks to minimize the maximum "regret" (opportunity loss)—the difference between the actual payoff they received and the best possible payoff they could have received had they known the state of nature in advance.

Decision Trees and Expected Value

A Decision Tree represents sequential choices and chance events. The model assigns probabilities to uncertain outcomes only when those estimates are supported. Expected value is one comparison criterion; it does not by itself capture risk tolerance, distribution of outcomes, or nonfinancial constraints.

Expected Value (EV)

Calculates the long-run average value of a decision by multiplying each possible financial outcome by its probability of occurring, and summing the results.

EV=∑i=1n(Pi×Xi)EV = \sum_{i=1}^{n} (P_i \times X_i)

Variables

SymbolDescriptionUnit
PiP_iThe probability of outcome i occurringdimensionless
XiX_iThe financial value (payoff) of outcome icurrency

Expected Value Context

To solve a decision tree, managers calculate the Expected Value at each chance node. EV is the long-run average value of a decision, calculated by multiplying each possible financial outcome by its probability of occurring, and summing the results.

Interactive Simulation

Explore decision trees and expected value dynamically with the simulation below.

Decision Tree: Launch vs. Don't Launch

Market Conditions

(60%)
Prob. of Failure:40%

Payoffs

$
$
$

Expected Value (EV)

Launch:$52,000
No Action:$0
Rec: Launch Product
Launch
Don't Launch
Strong (60%)
Weak (40%)
Decision
Market
EV: $52,000
END
Success
$100,000
END
Failure
$-20,000
END
Status Quo
$0
Match the Tool to Its Assumptions

Choose an analysis method that fits the information available rather than treating a numerical output as certainty.

A decision tree uses probability-weighted branches when probabilities are defensible. A payoff matrix can compare alternatives when probabilities are unavailable, but its selection rule (such as a conservative, optimistic, or regret criterion) must be stated.
Risk with supported probabilities
  • Decision tree
  • Expected value from stated probabilities
  • Test other constraints and sensitivity
Uncertainty without supported probabilities
  • Payoff matrix
  • Declare the criterion before comparing
  • Test how the decision changes under other criteria
Key Takeaways
  • A rational Decision-Making Process defines the decision, criteria, constraints, and alternatives; documents the choice; and reviews the result.
  • Managers handle routine, structured issues with Programmed Decisions (using rules and policies) and tackle unique, complex problems with Non-Programmed Decisions (requiring judgment and custom analysis).
  • Decisions are categorized by the information available: Certainty (outcomes known), Risk (probabilities known), and Uncertainty (neither known).
  • Expected value and payoff-matrix criteria make assumptions explicit; results still require checks against constraints and decision context.

References