Decision Tree

The decision tree allows you to use branching logic to segment decisions based on multiple conditions. It organizes rules into a clear, hierarchical structure, where conditions form the branching points, outcomes are usually defined at the leaf nodes, and actions run on every node the path visits. This makes it ideal for situations where more than two attributes or conditions need to work together to determine a result. The tree structure provides a straightforward way to visualize complex decision-making processes, and it clarifies how different conditions lead to specific outcomes.

A run starts at the root node, which has no parent, and follows one path only. At each visited node, actions run in the order listed before the conditions are checked in the order listed. The first true condition selects the branch, and the remaining conditions are skipped.

Actions write a fixed value exactly as typed. A value set by an earlier node is overwritten when a later node sets the same attribute path. An action without an attribute path is skipped. The run stops at a leaf, which has no attribute to branch on, or when no condition is true. In the second case, nothing is raised after that node's actions run.

A finished tree reads left to right as a map of the whole decision: each branching box branches on one attribute.

decision-tree-canvas-overview.png

🎯 Node Conditions: Each node in a decision tree uses the same evaluation operators as other Decisimo components. This includes numeric comparisons, text matching, interval operations, and more. For complete operator reference and examples, see Evaluation Logic Reference.

To add and test a Decision Tree in Decisimo, follow these steps:

  1. Access Decision Trees Section:

    • Click the "Decision Trees" button after logging into the Decisimo Dashboard.

    • Create a new decision tree by clicking the "+ Create a decision tree" button.

  2. Create Decision Trees:

    • Add Title and Identifier:

      • Enter the title and a unique identifier for the decision tree.

        tree-create-form.png

    • Define the Root Node:

      • Specify the initial condition or attribute that will split the root node.

  • Set node conditions: Click the "⊕ split" button and define the branching condition using any available operator (greater_than, equal_to, closed_interval, etc.).

    tree-node-split.png

  • Configure condition values: Set the comparison values for each branch (e.g., change value from 0 to 1, define ranges, set text matches).

    đź“‹ Available Operators: Decision trees support all standard evaluation operators including intervals, text matching, and special conditions. See Evaluation Logic Reference for the complete list.

  • Categorical shortcut: splitting on a text attribute with several known values (a status, a category, a country code) one condition at a time is tedious. Click "Categorical" under Presets, type the values comma-separated, and hit "Generate", so the portal creates one branch per value plus a "Default" branch for anything else, all in one step.

    decision-tree-categorical-split.png

  • The system should accept and save these changes correctly.

  • Use the UI to add new nodes.

  • Observe the placement and connections of the new nodes.

    • Expected Result: New nodes should be placed correctly and connected to the rest of the tree.

  • Use the delete function to delete the node, whether it's the parent node or a child node. Deleting the parent node will delete all child nodes associate with it.

    tree-delete-node.png

  • Add the nodes, the required actions, and the attribute path if required. A leaf node's outcome is one or more actions: click "Add action", pick the attribute path to write to, and set its value (e.g. $.loan_decision = APPROVED).

    decision-tree-node-action.png

  • Save the Decision Tree.

Attribute Path

For Node Conditions: Attribute paths specify which data field to evaluate at each decision point (e.g., $.credit_score, $.customer.age).

For Node Actions: Attribute paths define where to store results when a path through the tree is completed (e.g., $.loan_decision, $.risk_category).

🎯 JSONPath Examples:

$.credit_score >= 700     // Node condition
$.customer.age < 25       // Nested attribute condition
$.loan_decision = "APPROVED"  // Action result

đź“‹ Complete JSONPath Guide: For comprehensive attribute path syntax, data type handling, and troubleshooting, see JSONPath & Attribute Resolution Guide.

Testing Decision Trees

Decision trees support comprehensive testing to validate your branching logic:

Unit Testing

Test with specific values to trace the path through your decision tree:

tree-unit-test.png

Batch Testing

When you want to check many records at once, the Batch test tab opens a spreadsheet-like grid, with one column per attribute the tree branches on, one row per test record. Paste records straight from Excel with CTRL+V, or type into the blank row that always waits at the bottom. Add an “Expected” column to get PASS or FAIL per row, hit “Run all”, and export the results to CSV. Blank cells are sent as NULL. It works like batch testing for functions.

Value suggestions: the tree already knows the values it branches on, so each cell offers them as you type, including every threshold used in a condition on that attribute, plus a value that satisfies each one. Pick from the list instead of guessing which values actually send a record down a given branch.

The result column shows the decision the tree reached: the value written by the last node that acted. When a leaf sets several attributes, the result lists each path and its value. Click the result to unfold the full execution path, including every node visited, the condition that matched, and the actions fired.

decision-tree-batch-test.png

Decision trees have no “Suggest rows” button. Scorecard bins are independent, so the portal can derive one record per bin. A tree record instead has to satisfy every condition along a whole root-to-leaf path at once, which the portal cannot work out for you. The per-cell value suggestions cover the same ground.

Object Testing

Test with complete data objects to validate real-world decision flows and see which path is taken through the tree.

Verbose Output

Enable verbose output to see detailed execution traces:

  • Which nodes were evaluated
  • What conditions matched at each branch point
  • Which actions were executed
  • The complete path taken through the tree

🧪 Testing Best Practices: For comprehensive testing strategies, debugging techniques, and understanding execution traces, see Testing Decision Logic and Evaluation Logic Reference.

Revisions

Revision history can tell you what a decision tree looked in the historical version. You can view what was the definition, and who made the changes.

You can see the revision history of a decision tree by clicking “Revisions” from the drop-down menu of “Settings” button icon when viewing a decision tree.

tree-settings-menu.png

A new modal window will appear after you clicked “Revisions”, with a list of all historical changes and who made them. The window is titled “Revision” and contains a table with two radio buttons under “Compare”, followed by “Revision”, “Time”, “User” and “View” columns. The bottom of the window has “Close” and “Compare” buttons. By clicking the eye icon, you can see the definition of the decision tree at that time.

To compare revisions, select one revision in the left radio column and another in the right radio column, then choose “Compare” to see their definitions side by side.

tree-revisions.png