What does a decision tree describe in the context of decision analysis?

A decision tree is a structured, graphical model for representing decision-making processes. It is used in computer science, statistics, business administration, psychology and increasingly also in artificial intelligence. At its core, a decision tree depicts decision rules in the form of a tree structure. Each branch stands for a condition, each branch for a possible manifestation of this condition, and each leaf (end node) represents a result or a decision.

Decision trees are particularly popular because they represent complex relationships in a visually understandable way. They belong to the so-called supervised learning methods in the field of machine learning. At the same time, they have been used for decades in traditional decision analysis, for example in investment decisions, Risk assessments or medical diagnoses.

1. Basic idea of a decision tree

A decision tree is based on simple logic: If a certain condition is met, go to the left. If it is not fulfilled, go to the right. This principle is repeated recursively until a final result is achieved.

Formally, a decision tree consists of:

• Root node: Starting point of the decision process
• Inner nodes (decision nodes): Check a condition or feature
• Branches: Possible characteristics of the condition
• Leaves (Leaf Nodes): Final result or classification

Example: A company checks whether a project should be started. The first question is: Is the expected ROI greater than 12%? Depending on the answer, the decision-making process branches out into further checks.

2. Mathematical basis

Decision trees divide a decision space step by step. Mathematically, this is a recursive partitioning of the feature space.

Typical optimization criteria when training a decision tree are

• Gini index
• Entropy (information gain)
• Variance reduction (for regression trees)

The aim is to reduce the "impurity" of the data with each split.

3. Types of decision trees

Type	Type Description	Description Example
Classification tree	Assigns data to a category	Spam or no spam
Regression tree	Predicts numerical values	Sales forecast
CHAID	Statistically based allocation with chi-square test	Market segmentation
CART	Binary splits, widely used	Medical diagnostics

4. Advantages of decision trees

• High interpretability
• No linear assumptions necessary
• Can work with categorical and numerical data
• Can be displayed visually

5. Disadvantages of decision trees

• Overfitting
• Instability with small data changes
• Greedy optimization (local, not global)

The last point is particularly relevant: A classic decision tree only ever optimizes the best split locally. It does not check the entire decision space simultaneously.

6. Decision tree vs. complex decision space

If we evaluate 20 projects, for example, there are²²⁰ possible combinations. That is 1,048,576 combinations.

A classic decision tree would not search through all of these possibilities. It makes step-by-step decisions and follows a path.

This is the fundamental difference to modern decision intelligence.

7. StratePlan decision intelligence

While a decision tree structures a decision space hierarchically and optimizes it locally, stratePlan works with a global optimization logic.

Instead of evaluating individual splits, StratePlan analyzes the entire combinatorial space simultaneously. From around seven projects, the number of possible combinations increases exponentially (2^N). From 20 projects, we are already talking about over a million options. From 50 projects, we are talking about over a quadrillion.

A classic decision tree can represent these spaces structurally, but cannot calculate them globally.

StratePlan, on the other hand, uses mathematical optimization methods, to directly calculate the global optimum under constraints (budget, runtime, IRR, strategic restrictions) directly.

8. Example: Investment portfolio

A CFO has to choose from 15 projects. Each project has:

• Investment costs
• Expected ROI
• Duration
• Risk profile
• Strategic priority

A decision tree could filter step by step:

1. ROI > 10%?
2. Budget available?
3. Risk acceptable?

StratePlan calculates all²¹⁵ = 32,768 combinations simultaneously, taking into account all all constraints simultaneously and identifies the mathematically optimal portfolio.

9. Decision trees in practice

Decision trees are used in:

• Medical diagnostics
• Credit scoring
• Marketing segmentation
• Quality control
• HR decisions

In many of these use cases, interpretability is more important than global optimization.

10. Conclusion

A decision tree is a transparent, intuitive tool for structuring decisions. It is ideal for classification and forecasting tasks with clear characteristics.

Its weakness lies in the local, step-by-step decision logic. For high-dimensional portfolio or investment decisions with an exponential decision space, this structure is no longer sufficient this structure is no longer sufficient.

This is where modern decision intelligence such as StratePlan comes in: The search is not for the best next split, but the global optimum in the entire space.

FAQ

What is a decision tree in simple terms?

A decision tree is a decision diagram that checks questions one after the other and leads to a result.

Is a decision tree AI?

Yes, in machine learning it is a supervised learning process. It is one of the classic AI methods.

What is the difference between a decision tree and a random forest?

A random forest combines many decision trees to increase stability and accuracy.

Why are decision trees interpretable?

Because every decision is based on clear, comprehensible rules.

Where are the limits?

With exponentially growing decision spaces and complex constraints.

What makes StratePlan different?

StratePlan calculates the global optimum in the entire combinatorial decision space and considers budget, time and return constraints simultaneously.

When should you use a decision tree?

When transparency, rule-based and fast classification are more important than global portfolio optimization.