playbooks

home / skills / gtmagents / gtm-agents / decision-trees

decision-trees skill

/plugins/personalization-engine/skills/decision-trees

This skill helps design and audit branching logic using decision trees to optimize eligibility rules and fallback paths.

npx playbooks add skill gtmagents/gtm-agents --skill decision-trees

Review the files below or copy the command above to add this skill to your agents.

Files (1)
SKILL.md
1.4 KB
---
name: decision-trees
description: Use when designing branching logic, eligibility rules, and fallback paths.
---

# Personalization Decision Trees Skill

## When to Use
- Planning logic for dynamic experiences across web, in-app, email, or sales plays.
- Auditing existing decision flows for complexity, coverage, or compliance gaps.
- Simulating new branches before deploying rule or model updates.

## Framework
1. **Objective Mapping** – tie each node to business KPIs and user intents.
2. **Signal Hierarchy** – prioritize deterministic signals (consent, account tier, lifecycle) before behavioral or predictive ones.
3. **Fallback Design** – ensure every branch has a safe default when data is missing or risk flags appear.
4. **Experiment Hooks** – embed test slots at key decision points with guardrail metrics.
5. **Monitoring** – log path selections, success rates, and anomaly alerts for continuous tuning.

## Templates
- Decision tree canvas (node, condition, action, fallback, owner).
- Signal priority matrix (signal → freshness → reliability → privacy risk).
- Simulation checklist (scenarios, expected path, validation steps).

## Tips
- Keep trees shallow where possible; offload complexity to scoring models or external services.
- Version control decision logic alongside content assets for traceability.
- Pair with `governance` skill to log approvals for high-impact branches.

---

Overview

This skill helps design and audit decision trees for branching logic, eligibility rules, and safe fallback paths. It provides a practical framework, templates, and simulation checklists to translate business objectives into deterministic and testable decision flows. Use it to reduce complexity, ensure coverage, and prepare rules for production.

How this skill works

The skill guides you through objective mapping, signal hierarchy, fallback design, experiment hooks, and monitoring. It supplies a decision canvas and signal priority matrix to structure nodes, conditions, actions, owners, and safe defaults. Simulations and checklists let you validate expected paths and surface gaps before deployment.

When to use it

  • Designing personalized experiences across web, in-app, email, or sales plays
  • Auditing existing decision flows for complexity, coverage, or compliance gaps
  • Simulating new branches before rule or model updates go live
  • Defining eligibility rules for feature gating, promotions, or support routing
  • Preparing decision logic for A/B tests or staged rollouts

Best practices

  • Map each node to a clear business KPI and user intent to avoid orphaned branches
  • Prioritize deterministic signals (consent, account tier, lifecycle) before behavioral or predictive inputs
  • Design safe fallbacks for missing data and for risk-flagged paths
  • Keep trees shallow; move complex evaluations to scoring models or external services
  • Version control decision logic alongside content and pair with governance for approvals
  • Embed experiment hooks and guardrail metrics at key decision points

Example use cases

  • Eligibility flow for a tiered trial: check consent → account tier → usage threshold → fallback messaging
  • Customer support routing: priority signals (SLA, region) → behavioral signals (recent activity) → fallback to human queue
  • Marketing personalization: signal matrix selects content variant, with experiment slot and rollback path
  • Fraud screening: deterministic risk checks first, predictive model only when signals are ambiguous
  • Feature rollout: staged decision nodes that enable feature for cohorts and record path metadata for monitoring

FAQ

How do I choose which signals to prioritize?

Prioritize deterministic, reliable, and privacy-safe signals first (consent, account tier, explicit settings), then use behavioral or predictive signals when deterministic data is insufficient.

What should a fallback include?

A safe default action, clear owner, logging for monitoring, and a path for escalation or manual review if risk is present.