A/B test infrastructure
Stable user-bucket assignment backed by Plausible custom events. The homepage CTA is the first live experiment. Experiments are gated behind a flag so we can add more without a deploy.
Stable user-bucket assignment backed by Plausible custom events. The homepage CTA is the first live experiment. Experiments are gated behind a flag so we can add more without a deploy.
We shipped the A/B testing infrastructure we'll use for ongoing copy and UX experiments. It's a stable bucket-assignment system: each visitor is assigned to a variant once and stays there across sessions, keyed to a hashed visitor ID.
The first live experiment is on the homepage CTA. We're testing two variants to see which drives more free-trial signups.
How it works under the hood: assignments are determined server-side at render time (no client-side flicker) and the variant assignment is passed to Plausible as a custom property so we can segment conversion metrics by variant without a separate analytics tool.
New experiments can be defined in a single file and gated behind a flag, so we can run multiple experiments simultaneously without them interfering.
Commit: f5c1975