Decision Tree

A decision tree is a decision-making model widely used in business, science, and engineering — a tree-like structure representing a series of decisions and their potential consequences, useful specifically when there are multiple possible outcomes or paths and the best one isn't immediately obvious.

Structure

The root node at the top represents the initial decision. Each branch from a node represents a possible outcome or choice, connecting to further nodes representing the next decision in that path — so a decision tree makes explicit the full space of possible sequences, not just the single path someone happened to think of first.

Where it's applied

• Business — comparing scenarios such as marketing strategy options, pricing strategies, or product-development choices.
• Medicine — diagnosing a condition based on a patient's symptoms, branching through differential possibilities.
• Finance — evaluating investment strategies or financial plans across different outcome scenarios.

Types (in the data-science sense)

• Classification trees — classify data into discrete categories.
• Regression trees — predict a continuous value (a price, a temperature) rather than a category.
• Decision trees with continuous variables — handle input data that's continuous rather than purely categorical.

These automated, data-driven variants share the same branching structure as a manually-built business decision tree, but are learned from data rather than authored by hand.

Why it's a useful tool

Easy to interpret, even for people without a technical background — the branching structure itself explains the reasoning, unlike many other predictive models. Also flexible: a decision tree can be updated as new data or new options become available without needing to rebuild the whole structure from scratch.

Relationship to ideation

A decision tree is a structured evaluation tool — it maps out and compares already-identified options. It's typically used after Ideation's judgment-free idea generation has produced the options worth comparing, not as a substitute for generating those options in the first place.

Common pitfalls

• Building a decision tree before the option space is actually explored — see Ideation; a tree built from a narrow, prematurely-judged set of options only evaluates within that narrow set, missing whatever wasn't considered.
• Omitting a real branch/outcome because it seemed unlikely — a decision tree's value depends on genuinely covering the plausible outcome space; silently dropping an inconvenient branch defeats the purpose.
• Confusing a hand-built business decision tree with a machine-learned one — the data-science variants (classification/regression trees) are algorithmically fit to data and evaluated on predictive accuracy; a manually-authored business decision tree is a reasoning tool, judged by whether it captures the real decision space, not by a fit metric.

Learn more

• Ideation for the generative phase that typically precedes decision-tree evaluation.
• PlantUML Diagram for text-based diagramming (including tree-like structures) if a version-controlled representation is preferred.