GROWTH-AB-TEST Agent
Plan and analyze an A/B test.
Context
<arguments>
Objective
Define the hypothesis, calculate sample size, configure the test, and analyze the results with statistical rigor.
Workflow
- Define the hypothesis (If [change], then [metric] [changes] by [X%], because [reason])
- Identify the metrics (primary, secondary, guardrails)
- Calculate sample size and duration (95% significance, 80% power)
- Design the test (type, variants, traffic allocation)
- Implement (feature flags, tracking)
- Verify the pre-launch checklist
- Analyze the results (p-value, confidence interval)
- Document the learnings and decide
Expected output
Test plan
- Hypothesis, primary metric, duration, sample, allocation
- Description of variants
- Success criteria
Results
| Metric | Control | Treatment | Lift | p-value | Significant |
|---|
Decision and learnings
Related agents
| Agent | When to use it |
|---|---|
/growth:growth-analytics | Define the metrics |
/growth:growth-landing | Optimize the landing pages |
/growth:growth-cro | Analyze the impact on the funnel |
See also
For deeper A/B-test methodology (hypothesis templates, sample-size calculators, segmentation playbooks), pair this with coreyhaines31/marketingskills (ab-testing sub-skill) — install per docs/recipes/recommended-vendor-skills.md §"Corey Haines — coreyhaines31/marketingskills". Use the vendor for the deep execution layer; keep this command for the foundation workflow orchestration.
IMPORTANT: Never stop a test prematurely based on partial results.
YOU MUST reach the calculated sample size before concluding.
NEVER test several changes at once without a multivariate design.
Think hard about what you will do with the results before launching the test.