Decision Trees: A Simple Tool for Complex Choices

A decision tree is a diagram that represents a decision as a branching sequence of choices and outcomes. Each branch represents a possible choice or event; each terminal node represents an outcome. The diagram makes the structure of the decision visible, which is often more than half the work.

Decision trees are simple enough to draw with pen and paper for most everyday choices. They become more powerful with quantified probabilities and outcomes, but even the unquantified version is useful for making sure no major branches have been ignored.

A worked example

Suppose you’re considering whether to accept a job offer in a different city. The decision involves multiple uncertainties and tradeoffs.

The first branch is the decision itself: accept or decline. Below “accept,” there are further branches: the new job might go well or might not. The relocation might be smooth or might not. Your personal life might benefit or might suffer. Below “decline,” there are similar branches: stay where you are and continue at the current pace, or accept that declining itself has consequences for future opportunities.

Drawing the tree forces you to enumerate the scenarios. Each terminal node represents a specific possible outcome. You then estimate, however roughly, how likely each scenario is and how much you would value each outcome. The combination gives you an expected value for each top-level choice.

Why it’s useful even when imprecise

The numbers in a decision tree are almost always approximate. The probabilities are guesses. The outcome values are subjective. Many people, on seeing the imprecision, conclude that the exercise is pointless.

This is wrong. The exercise is useful because it forces explicit enumeration of scenarios that are otherwise lumped together. “Accept the job” collapses dozens of distinct possible futures into a single decision label. The tree breaks that collapsed concept back into its components.

The numbers do not need to be precise. They need to be reasoned. If you assign a 60% probability to the new job working out, you should be able to explain why — what evidence, what comparable cases, what specific concerns. The reasoning matters more than the precise number.

Common errors

The most common error in decision-tree thinking is failing to enumerate branches that don’t fit the preferred narrative. If you’re excited about a job change, you may not draw the branches where the change goes badly. If you’re anxious about it, you may not draw the branches where it goes well. The tree is only as useful as the honesty of the enumeration.

Another common error is treating the tree as if it produces a definitive answer. It does not. It produces a more structured version of your judgment, which is still a judgment. The numbers are estimates; the conclusion is therefore an estimate. The tree should inform your decision, not replace it.

When to use it

Decision trees are most useful for decisions that meet three conditions: the stakes are high enough to justify the analysis, the structure is complex enough that intuition is unreliable, and the relevant probabilities are at least roughly estimable.

Major financial decisions, career decisions, medical treatment decisions, and significant business decisions usually meet these conditions. Routine choices — what to have for dinner, which route to drive home — do not. The analysis takes more time than the decision is worth.

The practitioner who reaches for decision trees too often becomes paralyzed. The practitioner who never reaches for them makes major decisions on autopilot. The skilled decision-maker reserves the tool for cases where it earns its keep.