A/B Testing Skill

Purpose

Enables experimentation on lead generation sites to optimize conversion rates through simple, privacy-friendly client-side testing with statistical rigor.

Core Rules

1. One test at a time — Per page, avoid interaction effects
2. Statistical significance — Don't call winners too early (p < 0.05)
3. Track conversions — Not just clicks or exposures
4. Minimize flicker — Apply variants before paint using visibility hidden
5. Respect privacy — No PII in experiments, localStorage only
6. Minimum duration — Never less than 1 week (day-of-week effects)
7. Sample size first — Calculate required n before starting
8. Persist assignments — User sees same variant across sessions
9. Track exposures — Log every variant view to GA4
10. Conversion-focused — Optimize for business metrics, not vanity metrics