How To Calculate Decision Tree

Calculate expected values and optimal decisions using probability-weighted outcomes

Comprehensive Guide: How to Calculate Decision Trees

A decision tree is a powerful visual tool used in decision analysis to help individuals and organizations make optimal choices under uncertainty. By systematically evaluating possible outcomes, their probabilities, and their associated values, decision trees provide a structured approach to complex decision-making scenarios.

Fundamental Concepts of Decision Trees

1. Decision Nodes

Represent points where you make a choice between different alternatives. Typically depicted as squares in the tree diagram.

2. Chance Nodes

Represent points where outcomes are uncertain. Depicted as circles, with branches showing possible outcomes and their probabilities.

3. Terminal Nodes

Represent the final outcomes of each possible path through the tree. Show the payoff or value of that particular outcome.

The Decision Tree Calculation Process

1. Define the Decision Problem: Clearly articulate the decision to be made and the alternatives available.
2. Identify Possible Outcomes: For each alternative, determine all possible outcomes and their probabilities.
3. Assign Values to Outcomes: Quantify the value (monetary or utility) of each possible outcome.
4. Calculate Expected Values: For each alternative, calculate the expected value by multiplying each outcome’s value by its probability and summing these products.
5. Compare Alternatives: Select the alternative with the highest expected value as the optimal decision.
6. Perform Sensitivity Analysis: Examine how changes in probabilities or values affect the optimal decision.

Mathematical Foundation of Decision Trees

The expected value (EV) calculation is central to decision tree analysis. For a given alternative with multiple possible outcomes, the expected value is calculated as:

EV = Σ (P_i × V_i)

Where:

• P_i = Probability of outcome i
• V_i = Value of outcome i
• Σ = Summation over all possible outcomes

Practical Example: Product Launch Decision

Consider a company deciding whether to launch a new product. The decision tree might include:

Alternative	Market Response	Probability	Net Profit ($)	Expected Value ($)
Launch Product	High Demand	0.30	500,000	150,000
	Moderate Demand	0.50	200,000	100,000
	Low Demand	0.20	-100,000	-20,000
Don’t Launch	Status Quo			0
Total Expected Value:				230,000

In this example, launching the product has an expected value of $230,000, which is higher than not launching ($0), making it the optimal decision.

Advanced Decision Tree Techniques

1. Risk Profiles

Graphical representations showing the distribution of possible outcomes for each alternative, helping visualize risk.

2. Value of Information

Calculates how much additional information (like market research) would be worth before making the decision.

3. Sequential Decisions

Models decisions that unfold over time, where later decisions depend on outcomes of earlier ones.

Common Applications of Decision Trees

• Business Strategy: Evaluating market entry, product development, or investment decisions
• Finance: Assessing investment portfolios or capital budgeting decisions
• Healthcare: Determining optimal treatment plans based on patient responses
• Project Management: Evaluating different project approaches and their potential outcomes
• Public Policy: Assessing the impact of different policy alternatives

Decision Trees vs. Other Decision-Making Tools

Tool	Best For	Strengths	Limitations	Complexity
Decision Trees	Sequential decisions with uncertainty	Visual, handles uncertainty well, shows decision paths	Can become complex with many branches	Moderate
Decision Matrices	Simple multi-criteria decisions	Easy to create, good for weighted criteria	No probability consideration, less visual	Low
Monte Carlo Simulation	Complex systems with many variables	Handles vast uncertainty, probabilistic	Requires specialized software, less transparent	High
Cost-Benefit Analysis	Financial evaluation of projects	Quantitative, standardized approach	Ignores qualitative factors, static analysis	Moderate

Building Decision Trees: Step-by-Step

1. Start with the Decision Node:

   Place your primary decision at the far left of the diagram. Draw branches for each possible alternative.

2. Add Chance Nodes:

   At the end of each alternative branch, add chance nodes representing uncertain outcomes. Draw branches for each possible outcome.

3. Assign Probabilities:

   Label each outcome branch with its probability. Ensure probabilities for all branches from a chance node sum to 1 (or 100%).

4. Add Terminal Values:

   At the end of each outcome branch, add the value (monetary or utility) associated with that outcome.

5. Calculate Expected Values:

   Work backward from the terminal nodes, calculating expected values at each chance node and selecting the best alternative at each decision node.

6. Interpret Results:

   The path with the highest expected value represents the optimal decision strategy.

Common Mistakes in Decision Tree Analysis

• Ignoring All Possible Outcomes: Failing to consider all relevant outcomes can lead to incomplete analysis.
• Incorrect Probability Assessment: Overconfidence or bias in probability estimates can skew results.
• Overlooking Time Value of Money: For financial decisions, not discounting future cash flows appropriately.
• Double Counting Risks: Including the same risk factor in multiple parts of the tree.
• Neglecting Sensitivity Analysis: Not testing how changes in assumptions affect the optimal decision.
• Overcomplicating the Model: Creating trees with too many branches that become unmanageable.

Software Tools for Decision Tree Analysis

TreePlan

Excel add-in that creates decision trees within spreadsheets. Good for financial analysis and sensitivity testing.

PrecisionTree

Advanced decision tree software with Monte Carlo simulation capabilities. Integrates with Excel.

Analytica

Visual software for building and analyzing decision models with influence diagrams and decision trees.

Academic Research on Decision Trees

Decision tree analysis has been extensively studied in decision science and operations research. Key academic contributions include:

• Raiffa’s Decision Analysis: Howard Raiffa’s work at Harvard established foundational principles for decision trees in his 1968 book “Decision Analysis: Introductory Lectures on Choices Under Uncertainty.”
• Von Neumann and Morgenstern’s Utility Theory: Their 1944 work “Theory of Games and Economic Behavior” provided the theoretical basis for quantifying preferences in decision trees.
• Keeney and Raiffa’s Multi-Objective Decisions: Their 1976 book “Decisions with Multiple Objectives: Preferences and Value Tradeoffs” extended decision tree methodology to complex multi-criteria problems.

Limitations and Criticisms

While decision trees are powerful tools, they have some limitations:

1. Cognitive Biases: The quality of a decision tree depends on the accuracy of probability and value estimates, which can be affected by cognitive biases like optimism bias or anchoring.
2. Complexity Management: Real-world decisions often have numerous possible outcomes and alternatives, making comprehensive trees impractical to construct and analyze.
3. Static Nature: Traditional decision trees represent a single point-in-time analysis and don’t easily accommodate dynamic situations where probabilities or values change over time.
4. Qualitative Factors: Decision trees primarily handle quantitative data, making it challenging to incorporate important qualitative considerations.
5. Interdependencies: They struggle to model situations where outcomes are interdependent or where one decision affects the probabilities of another.

Enhancing Decision Trees with Other Techniques

To address some limitations, decision trees can be combined with other analytical methods:

Influence Diagrams

Visual representations that show relationships between decisions, uncertainties, and values before constructing the decision tree.

Monte Carlo Simulation

Can be used to model uncertainty in probability and value estimates, providing probability distributions for outcomes.

Sensitivity Analysis

Systematically varies key assumptions to identify which factors most influence the optimal decision.

Real-World Case Studies

The following examples demonstrate practical applications of decision tree analysis:

1. Pharmaceutical R&D:

   A major pharmaceutical company used decision trees to evaluate whether to proceed with clinical trials for a new drug. The analysis considered:

   • Probability of success at each trial phase (60% for Phase I, 40% for Phase II, 25% for Phase III)
   • Development costs ($50M for Phase I, $100M for Phase II, $200M for Phase III)
   • Potential revenue if approved ($1.2B over 10 years)
   • Alternative option to license the drug to another company

   The decision tree revealed that proceeding with trials had an expected value of $180M, while licensing had an expected value of $120M, leading to the decision to proceed with internal development.

2. Oil Exploration:

   An energy company used decision trees to evaluate drilling options for a potential oil field. The analysis included:

   • Probability of finding oil (30%) or dry well (70%)
   • Drilling costs ($20M)
   • Potential revenue if oil found ($150M)
   • Option to sell drilling rights for $10M
   • Option to conduct seismic tests first ($2M) to improve probability estimates

   The decision tree showed that drilling immediately had an expected value of $13M, while conducting tests first had an expected value of $15M, making testing the optimal initial strategy.

Ethical Considerations in Decision Tree Analysis

When applying decision tree analysis, several ethical considerations should be kept in mind:

• Transparency: All assumptions, probabilities, and values should be clearly documented and justified to stakeholders.
• Stakeholder Impact: The analysis should consider impacts on all affected parties, not just the decision-maker.
• Value Assessment: Monetary values should not be the sole consideration; ethical and social values should be incorporated where possible.
• Bias Mitigation: Steps should be taken to identify and mitigate cognitive biases in probability and value estimates.
• Uncertainty Communication: The level of uncertainty in the analysis should be clearly communicated to decision-makers.

Future Directions in Decision Tree Analysis

Emerging trends and developments in decision tree methodology include:

• Machine Learning Integration: Using machine learning algorithms to estimate probabilities and values based on historical data.
• Real-Time Decision Trees: Dynamic decision trees that update probabilities and values in real-time as new information becomes available.
• Collaborative Decision Trees: Web-based tools that allow multiple stakeholders to contribute to and review decision tree models.
• Visualization Enhancements: Interactive visualizations that allow decision-makers to explore different scenarios more intuitively.
• Behavioral Decision Trees: Incorporating behavioral economics insights to better model how people actually make decisions under uncertainty.

Learning Resources for Decision Tree Analysis

For those interested in deepening their understanding of decision trees, the following resources are recommended:

• Books:
  • “Decision Analysis for Management Judgment” by Paul Goodwin and George Wright
  • “Smart Choices: A Practical Guide to Making Better Decisions” by John Hammond, Ralph Keeney, and Howard Raiffa
  • “Making Hard Decisions: An Introduction to Decision Analysis” by Robert Clemen and Terence Reilly
• Online Courses:
  • Coursera’s “Introduction to Decision Making” (University of Michigan)
  • edX’s “Decision Making in a Complex and Uncertain World” (University of Queensland)
  • MIT OpenCourseWare’s “System Optimization and Analysis for Manufacturing”
• Professional Organizations:
  • INFORMS (Institute for Operations Research and the Management Sciences)
  • Decision Analysis Society
  • Society for Risk Analysis

Authoritative References

For more in-depth information on decision tree analysis, consult these authoritative sources:

• National Institute of Standards and Technology (NIST) – Provides guidelines on risk assessment methodologies including decision trees
• U.S. Food and Drug Administration (FDA) – Uses decision analysis in drug approval processes and provides case studies
• Harvard Decision Science Laboratory – Conducts research on decision-making under uncertainty and offers educational resources
• Stanford University’s Decision Analysis Program – Offers courses and research on advanced decision analysis techniques