Skip to main content

How-to

Task-oriented guides for building and analyzing a decision tree. These assume you already know the core concepts and node types.

How to add a node​

  1. Hover over any non-Leaf node to reveal a + button on its right edge.
  2. Click + to open an inline popover, then choose the node type: Decision, Chance, or Leaf.
  3. Enter a Label.
  4. Optionally enter a Payoff (USD).
  5. If the parent is a Chance node, enter the Probability for this branch (required).
  6. Click Add.

The new node is inserted, the layout updates automatically, and the Detail tab opens for the new node.

How to edit a node​

Click any node to select it and open the Detail tab. All properties are editable there.

Double-click a node's label to edit it inline directly on the canvas.

How to delete a node​

Select a node, then press Backspace or Delete, or use the ⋯ context menu.

warning

Deleting a node removes it and all of its descendants. This action cannot be undone. Pocketstats shows a confirmation prompt when the subtree being deleted has 3 or more nodes.

How to navigate the canvas​

The canvas uses an automatic left-to-right hierarchical layout — nodes cannot be dragged to arbitrary positions, since the layout recalculates whenever the tree changes to keep the structure readable. Scroll to zoom, drag to pan, and use the ⤢ Fit button in the toolbar to fit the entire tree in the viewport.

note

The automatic layout is intentional: it keeps even large trees legible without manual arrangement.

How to read a node's Detail tab​

Select any node to open its Detail tab:

  • Decision node: shows the EV of every child branch side by side, with the highest EV highlighted in blue.
  • Chance node: shows the blended EV and the contribution of each child branch.
  • Leaf node: shows the leaf's payoff, its path probability (the product of all branch probabilities from root to this leaf), its weighted EV contribution to the root, and a breadcrumb of node labels from root to leaf.

Each node also displays its EV below its label directly on the canvas:

┌──────────────────┐
│  Enter Market    │
│  EV: $1,240,000  │
└──────────────────┘

See EV formulas for when a node shows EV: — instead of a value.

How to fix invalid branch probabilities​

Every Chance node requires its branch probabilities to sum to exactly 100%. While you're actively editing a probability field, no validation fires — the tool checks the sum as soon as you click away or after you add a new branch. See Probability validation states for what each indicator means. The warning is non-blocking: you can keep editing the tree, but the tool won't display EV for any subtree that depends on an invalid Chance node until it's fixed.

Probabilities over 100% — Happens when you add a new branch and forget to reduce an existing one. Reduce one or more branch values. The ! badge disappears when the sum returns to 100%.

Probabilities under 100% — Happens when you add a branch without entering its probability yet, or when you reduce a value to explore a scenario. Fill in the missing branch or adjust existing values. The ! badge disappears when the sum reaches 100%.

Once the sum reaches 100%, the live EV display is restored immediately for that subtree and any ancestor whose EV depended on it.

How to run a sensitivity analysis​

Sensitivity analysis lets you test how much a single probability or payoff affects the expected value (EV) of your tree — before committing to any changes. While you drag a slider, EVs update live on the canvas. When you release the slider, the tree reverts to its saved values, leaving your model unchanged. See Core concepts for definitions of sensitivity parameter, saved values, EV impact, and crossover indicator.

Enabling the panel

  1. Click the Sensitivity toggle in the canvas toolbar. The sensitivity panel slides in above the canvas.
  2. If you have already added parameters, their sliders appear immediately. Click the toggle again to collapse the panel.

Adding a parameter

  1. Select a node and open the ⋯ context menu, then choose ~ Analyze — or, in the Detail tab for any node, click ~ Analyze next to the payoff or probability field.

The parameter appears in the sensitivity panel with a slider. See Sensitivity panel fields for what each parameter card shows.

Removing a parameter

Click the × button in the top-right corner of a parameter card to remove it from the panel. The slider disappears and the EV calculation reverts to using the saved value for that field.

Reading crossover indicators

If moving a parameter causes the optimal decision branch to change — a different top-level branch now has the highest EV — the panel shows a crossover indicator:

"Branch B" becomes optimal at this value

This tells you the exact threshold at which the decision reverses, which is often the most actionable insight from a sensitivity analysis.

Typical use cases

  • Testing a key probability: if your model hinges on a market adoption rate, add that probability to the panel. Drag from pessimistic to optimistic and observe when (if ever) the decision flips.
  • Testing a cost assumption: add a cost payoff and drag the range to see how much the cost can grow before it changes the optimal path.
  • Comparing two parameters: add both the probability and the cost for a single branch. Because each slider tests its parameter independently, you can compare the EV impact side-by-side and see whether the decision is more sensitive to the probability or to the cost.
note

For multi-variable analysis where you want to test combinations of values rather than each parameter independently, use CSV export and analyze the paths in a spreadsheet. See How to export your model as CSV.

How to edit your tree with AI​

The AI tab in the right panel gives you a plain-English interface for building and refining your decision tree. It reads your current tree structure — all nodes, edges, labels, payoffs, probabilities, current EVs, active sensitivity parameters, and any probability validation warnings — and can make structural changes, update values, rename nodes, and explain outcomes. See AI editing capabilities for the full list of supported tasks.

  1. Click the AI tab in the right panel (it also opens automatically when no node is selected).
  2. Type your request in plain English and press Enter.
  3. Watch the AI's response stream in. If your request involved a structural change, the tree updates automatically once the patch passes validation.
  4. Review the change. If something looks off, use probability validation to check for constraint errors.

If the change isn't what you expected, describe what you want differently or ask the AI to undo it.

note

Edits are applied via a validated patch — the AI does not write free-form text into your tree. If the patch fails validation (for example, if the AI tries to add a branch with a probability that would push a Chance node over 100%), the error is surfaced and no partial change is applied.

tip

The AI works best with specific requests. Instead of "make it better," try "increase the Market Success probability to 70% and reduce the development cost to $400,000."

Conversation history is preserved per project and persists across browser sessions until you clear it.

How to export your model as CSV​

Decision Tree exports your model as a CSV file — one row per unique path from root to leaf. See CSV export format for the column layout and an example.

  1. Click ↓ Export CSV in the canvas toolbar. The file downloads immediately.

The export is useful for:

  • Reporting: paste into a presentation or share with stakeholders who want the full data
  • Further analysis: sort by path_ev to rank outcomes, or sum path_ev across paths to verify the root EV
  • Documentation: archive the full model state at a decision point
tip

Sort rows by path_ev (descending) to instantly rank outcomes by expected value. Sum all path_ev values to cross-check against the root EV shown in the toolbar — they should match.