Aller au contenu principal
GROWTH

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

MetricControlTreatmentLiftp-valueSignificant

Decision and learnings

AgentWhen to use it
/growth:growth-analyticsDefine the metrics
/growth:growth-landingOptimize the landing pages
/growth:growth-funnelAnalyze the impact on the funnel

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.


See also