Decision tree for business decisions: branches, probabilities and payoffs
The essentials in 30 seconds
- A decision tree maps choices, uncertain events, and payoffs, then rolls back from right to left to the best first move.
- Squares mark choices, circles mark chance, and branch probabilities sum to 1.
- In the example, launching now is worth 107.5 and a pilot first is worth 115 (US$ thousands).
- The pilot stays ahead while it costs less than 27.5.
- The EQOS Asia Prediction pillar projects 3 trajectories (EQOS Asia method page, as of October 2026).
A decision tree draws a sequence of choices and uncertain events as branches, puts a probability on each uncertain branch and a payoff at each end, and then works back to the best first move. It makes timing visible, for example whether to test before committing. This page builds one tree with illustrative figures, as of October 2026. It applies the guide on expected value to make a decision.
On this page
Compare the EQOS Asia offers and request a 30-minute pre-diagnostic.
What is a decision tree for business decisions made of?
A decision tree has three kinds of element. A decision node, drawn as a square, is a point where you choose. A chance node, drawn as a circle, is a point where an uncertain event decides the path, and each branch from it carries a probability. An end point, drawn as a triangle, carries the payoff of the full path in one currency.
Read the tree from left to right as a story: first you choose, then the world responds, then you may choose again. The branches from a decision node are options under your control. The branches from a chance node are outcomes you do not control, and their probabilities sum to 1. This split keeps the tree honest about what is a choice and what is luck.
How do you draw a decision tree step by step?
Start with the decision on the table and write its options as branches. For each option, ask what uncertain event follows, and add a chance node with two to four outcomes. Where a second choice follows an outcome, add another decision node. Continue until the paths end, then write the net payoff at every end point.
Keep the tree small. Two or three levels carry most business decisions, and a tree with more than 20 end points is hard to defend. Give each probability a source, such as a base rate or a pilot result, and write the payoffs net of all costs on the path, including the cost of any test along the way.
What does a decision tree show for a pilot before a launch?
Take a product launch, all figures illustrative and in US$ thousands. Launching now succeeds with a 55 percent chance and nets 400, and fails with a 45 percent chance and nets minus 250. A pilot costs 20. With a 50 percent chance the pilot is positive, and the launch then succeeds with a chance of 80 percent. Otherwise the team stops.
Two paths come out of the first decision node: launch now, or run the pilot first. The pilot path adds a chance node for the pilot result, and a second decision node after a positive result. Writing the paths this way shows the value of waiting: the pilot buys information at a known price before the large commitment.
How do you roll back a decision tree to find the best choice?
Work from the right. At each chance node, replace the branches with their expected value. At each decision node, keep the best option. Launching now gives 0.55 × 400 + 0.45 × (−250), which equals 220 − 112.5, or 107.5. After a positive pilot, the launch is worth 0.80 × 400 + 0.20 × (−250), or 270, and after a negative pilot the team stops at 0.
Subtract the pilot cost of 20 on every pilot path: 250 after a positive result and minus 20 after a negative one. The pilot node is then worth 0.50 × 250 + 0.50 × (−20), or 115. The pilot path at 115 beats launching now at 107.5, so the tree recommends the pilot as the first move.
How much can the pilot cost before the decision tree changes its answer?
The pilot path is worth 135 before its cost, since 0.50 × 270 equals 135. Subtracting a cost c gives 135 − c. Setting 135 − c equal to 107.5 gives c = 27.5. A pilot priced at 20 stays attractive, and a pilot priced at 30 gives 105, which falls below launching now.
This break-even tells the team how hard to negotiate on the price and scope of the pilot. It also shows how thin the advantage is: 7.5 on a base of 107.5. A recommendation with a thin margin deserves a second look at the probabilities before the team commits.
How do you keep the probabilities of a decision tree consistent?
Two rules keep a tree consistent. The branches of each chance node sum to 1, and the probabilities after a test agree with the probability before it. In the example, a positive pilot with a chance of 0.50 and an 80 percent launch success gives 0.40. A negative pilot with a 30 percent launch success gives 0.15. The total is 0.55, which matches the direct launch.
Write the conditional probabilities in a small table beside the tree. If the numbers do not reconcile, one of them is wrong, and the mismatch is useful: it reveals an estimate that nobody had cross-checked. A tree whose probabilities reconcile is far easier to defend in front of a board.
When does a decision tree for business decisions reach its limit?
A tree grows fast. Each added event doubles the number of paths, and most real decisions have continuous ranges that a few branches only approximate. A tree also treats the probabilities as given. When many variables interact, a simulation covers the ranges better, and the tree remains the clearest way to show the sequence of choices.
The EQOS Asia decision engine adds a structural view: its Prediction pillar projects 3 trajectories with estimated tipping dates (EQOS Asia method page, as of October 2026). A reader can try a first score in the simulator on the same decision and compare the result with the tree’s recommended first move.
See also in this guide
Frequently asked questions
What is a decision tree in business?
A decision tree is a diagram of choices, uncertain events, and payoffs. Squares mark choices, circles mark uncertain events with probabilities, and end points carry payoffs. Rolling the tree back from right to left gives the expected value of each first move and shows the best one.
How do you calculate a decision tree?
Start at the end points. At each chance node, replace the branches with their probability weighted average. At each decision node, keep the option with the highest value. Continue to the left until the first decision node shows the value of each option.
What is the advantage of a decision tree over a simple expected value?
A tree handles sequences. It shows whether to test before committing, when to stop, and what information is worth, while a single expected value treats the choice as one step. The tree keeps the order of events visible to every reader.
Can the EQOS Asia pre-diagnostic check a decision tree?
Yes. The free 30-minute pre-diagnostic applies the EQOS Asia equation to your decision across 9 dimensions and shows which carry the most risk. It runs by video in Singapore time (SGT) under NDA, and it gives a second view beside your tree (EQOS Asia contact page, as of October 2026).
Test your knowledge
1. Which shape marks a decision node?
- A circle
- A triangle
- A square
Show the answer
A square.
2. What do the branches of one chance node sum to?
- 1
- 100
- 0.5
Show the answer
1.
3. What is the value of launching now in the example?
- 115
- 107.5
- 220
Show the answer
107.5.
4. What is the highest pilot cost at which the pilot path still wins?
- 20
- 27.5
- 35
Show the answer
27.5.
5. In which direction do you roll back a tree?
- From right to left
- From left to right
- From the middle
Show the answer
From right to left.
About the author
Stéphane Jambu
President, EQOS Forecast Asia
Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile
Sources: EQOS Asia method page (eqos-asia.com/method.html), simulator page (eqos-asia.com/simulator.html), and contact page (eqos-asia.com/contact.html), consulted on 22 October 2026. Page updated on 22 October 2026.