Decision tree

A decision tree models a decision the way people often explain it: ask a question, and depending on the answer, ask the next question or give the answer. Use a tree when different cases need different questions. For example, enterprise customers need a risk check before approval, while small orders do not. When every case asks the same questions, a decision table is usually easier to read.

Parts of a decision tree

Part	Where you define it	What it is
Decisions	Decisions tab	The questions the tree can ask. Each has a Name, a Type and an Expression that works out the value, for example Quote.totalamount.
Outputs	Output tab	The values the tree returns, each with a name and a type. An output can be a list.
Decision cards	Decision Tree tab	A point in the tree where a decision is asked.
Branches	Decision Tree tab	The paths out of a decision card. Each branch shows its condition as a chip, such as > 50000. A branch with no condition is the default.
Outcome nodes	Decision Tree tab	The end of a path. Each sets the values of the outputs.

Build a decision tree

1. Create a logic block and choose the Decision tree card. See Create a logic block.
2. Add the inputs on the Input tab, and rows or queries on the Data tab.
3. Add a decision for each question the tree can ask.
4. Add an outcome for each value the tree returns.
5. Start the tree: select Choose first decision.
6. Under each answer, add another decision or an outcome. When you add a decision to the tree you can reuse an Existing Decision or create a New Decision.
7. Give every decision card a default branch (no condition), or make sure its branches cover every value.
8. Save, then select Run to test each path.

Work with the canvas

• Decision cards are white, outcome nodes are light orange, and open branches that still need a decision or outcome show as dashed placeholders.
• Hover over a card, branch or node to see its actions.
• Drag the background to pan. Use Zoom in, Zoom out and Reset zoom to fit large trees. Expand gives the canvas the whole window.
• Deleting a decision from the Decisions tab removes its nodes from the tree, and deleting an output removes it from every outcome node. The editor asks you to confirm first.

How it behaves

• Start at the first decision. The tree works out that decision’s value and checks its branches in order.
• The first branch whose condition holds is followed. A branch with no condition is the default and is taken when no other branch matches.
• Continue until an outcome. The outcome node’s output values are returned.
• No matching branch is an error. If a decision has no matching branch and no default, the run fails with “Decision tree evaluation failed to reach a decision.” Give each decision a default branch.
• Comparisons follow the expression rules, including empty numbers counting as 0. See Empty values.

Example: quote approval routing

Inputs: QuoteId (Lookup to Quote) Data: the quote and its customer account Decisions: Segment = Account.customertypecode, Amount = Quote.totalamount, Discount = Quote.discountpercentage Outputs: Approver (Text), ApprovalDays (Whole Number)

flowchart TD
  S{Segment} -->|Enterprise| A{Amount}
  S -->|default| D{Discount}
  A -->|> 50000| O1[Sales Director<br/>3 days]
  A -->|default| O2[Account Manager<br/>2 days]
  D -->|> 15| O3[Sales Manager<br/>2 days]
  D -->|default| O4[No approval<br/>0 days]

An enterprise quote of 80,000 goes Segment → Amount → Sales Director, 3 days. A partner quote with a 20% discount goes Segment → Discount → Sales Manager, 2 days. Different segments ask different questions, which a decision table could not express as compactly.

Related

• Decision table
• Inputs and outputs
• Run and test