How to Draw a Decision Tree for Two Options

Two options with a later decision point

A studio can buy equipment now or rent for one project. Buying creates later branches for repeat demand and resale. Renting creates a later decision after real demand is observed. The tree shows that renting purchases information and preserves a later choice.

Map only uncertain events and later choices that change the action. Extra branches do not automatically create better analysis.

Use a tree only for sequences

A tree adds value when choosing A or B leads to different uncertain events or later choices. If every consequence happens immediately, a checklist or matrix is easier to read.

Use a tree when acting now changes what can happen or what can be chosen later. If the options only differ across several criteria at one time, a table is clearer. Write the first decision, then list only events that lead to a different later action or consequence. This keeps the tree focused on sequence rather than decorating a static comparison with branches.

Separate choices from events

Draw decision points as choices you control and event points as outcomes you do not control. Mixing them makes luck look like a decision and can hide where flexibility exists.

Label each branch as a choice you control or an event you do not control. A later choice needs an owner and available options. An uncertain event needs a defined outcome and evidence basis. Mixing the two can make a future decision look inevitable or make a controllable action look like luck. Keep the distinction visible in the branch text.

When branch outcomes have defensible probabilities, map them in the interactive decision tree builder and keep each probability and outcome visible beside the expected result.

The branching rule for a useful tree

Prefer the branch that remains acceptable after realistic event ranges and preserves valuable later choices where uncertainty is high.

List the branches first

Keep branches comparable

Use the same time horizon and consequence categories at each endpoint. Include the option to stop, delay, or change course when it is genuinely available.

Use the same time horizon and outcome unit at every endpoint. Include the work and cost required to reach that point, not only the final payoff. If two branches end in the same action with similar consequences, combine them. The tree should make a decision easier to inspect, not reward the option with the most detailed branch list.

Use ranges when probability is weak

Do not invent a 63 percent chance because a tree asks for a number. Label confidence, use a plausible range, and check whether the preferred branch changes across that range.

When probabilities are weak, show a range or threshold instead of a confident decimal. Ask what probability would make the preferred first action change. Then compare that switching point with the available evidence. If no defensible estimate exists, keep the branches as scenarios and choose using downside capacity, reversibility, or a small information-gathering step.

Prune the tree without losing the decision

Begin with every material branch, then remove events that lead to the same action and similar consequence. Combine outcomes only when the response would be the same. A smaller tree is easier to audit and less likely to create the appearance of sophistication through detail.

If one branch contains many dependent events, replace the chain with a scenario and state the combined assumption. The tree should show where the decision changes, not simulate every operational event that might occur. Label any pruned branch and the reason it was combined so a reviewer can restore it if the consequence becomes material.

Where a branch allows learning before commitment, stage the choice around reversible steps and delay only the permanent component.

Branches that add detail without changing action

Treating an event as controllable
Adding trivial branches
Inventing precise probabilities

Draw and prune the decision tree

First decision
Name A and B without outcomes attached.
Events
List only events that materially change consequences.
Later choices
Add points where you can adapt after learning.
Endpoints
Use one shared time horizon and consequence set.
Range test
Vary the weakest probability or estimate.

When a scenario is clearer than another branch

Complex legal, medical, engineering, or investment trees require domain evidence and qualified review. A simple tree is an organizing aid.

Sources used for this guide

HM Treasury, The Green Book. Official appraisal guidance discussing decision trees and uncertainty.