eigent-ai/ab-test-setup

A/B Test Setup

Overview

Use this skill to guide the full experiment lifecycle: hypothesis, design, sample size, implementation, analysis, and playbook documentation. Keep tests focused, measurable, and resistant to common errors like peeking early or testing too many changes at once.

Workflow

1. Define the business goal, user segment, current baseline, target metric, and guardrail metrics.
2. Write a specific hypothesis:
   • If we change X for audience Y, metric Z will improve because...
3. Design the test:
   • Control and variant.
   • Primary metric.
   • Secondary and guardrail metrics.
   • Traffic split, eligibility, exclusions, and duration.
4. Estimate sample size or minimum detectable effect when baseline traffic and conversion rates are available.
5. Create an implementation checklist:
   • Tracking, randomization, QA, exposure logging, analytics events, and rollback.
6. Define decision rules before launch:
   • Ship, revert, iterate, or continue testing.
7. Analyze results after the test reaches the agreed sample size.
8. Document what changed, what was learned, and follow-up experiments.

Test Backlog Pattern

When building a backlog, score each idea with ICE:

• Impact: expected business or user benefit.
• Confidence: evidence quality.
• Effort: complexity and implementation cost.

Prioritize tests that combine high impact, credible evidence, and low operational risk.

Example Prompts

• I want to A/B test our signup CTA button. Current conversion rate is 3.2%, 8,000 visitors/month. Help me design the test, calculate the required sample size, and define what success looks like.
• Our A/B test just hit sample size. Here are the results [paste metrics]. Is this statistically significant? Should we ship the variant, revert, or keep testing?
• Build a prioritized A/B test backlog for our onboarding flow. Use ICE scoring. Sources to mine: our drop-off analytics, last month's support tickets, and these 3 heatmap observations.