Run A/B tests on your site
Stop guessing what works. Test it with real users and let the data decide.
Categories: feature flags, conversion, analytics, growth
Tools that power the Run A/B tests on your site stack
PostHog — Experiments and analytics
Run A/B tests and feature flags with built-in product analytics
Mixpanel — Conversion funnel analysis
Track how different test variants affect funnel completion rates
Vercel — Experiment deployment
Deploy experiment variants and use Global Config or Routing Middleware where needed; measure latency in your setup
Setup guide
- PostHog: Define the experiment hypothesis, primary metric, assignment unit and stopping criteria before launching. Configure the feature flag in PostHog and persist each participant’s assignment. If assignment should follow a person across devices, use a consistent authenticated identity and test sign-in transitions; anonymous device IDs alone cannot do this. (Read official instructions)
- Mixpanel: If measuring results in Mixpanel, explicitly include the experiment ID, assigned variant and exposure event in the events you send there. PostHog’s flag properties are not automatically copied to another analytics SDK. Check both variant exposures and conversions in a test dataset before interpreting results. (Read official instructions)
- Vercel: Deploy the experiment through a Vercel preview first and test both variants with controlled identities. Confirm the selected SDK and assignment code support your server/edge runtime, keep assignments stable across rendering and navigation, and document rollback. Promote only after the exposure and conversion events are verified. (Open Vercel instructions)
Check: Before launch, one controlled identity keeps its assigned variant; exposure and conversion records agree in each analytics tool, and rollback restores the baseline.