Skip to main content
Experiments enable A/B testing of paywalls, offerings, and remote configurations. Each experiment has one or more groups (variants), and users are attached to a specific group either automatically by audience rules or manually via the API.

​
Key concepts

  • Status — draft, active, paused, or finished. Drive transitions via POST /v4/experiments/{experiment_id}/status; finishing releases all assigned users.
  • Groups — Variants within an experiment. Each group carries a payload (offering, screen, remote config value, etc.).
  • User attachment — POST /v4/experiments/{experiment_id}/users/{user_id} pins a user to a group, overriding automatic evaluation; the assignment is never re-evaluated and persists until detached or the experiment finishes. Group membership is sticky: raising the traffic allocation never re-admits previously rejected users, and finishing the experiment releases everyone. Both attach and detach invalidate the SDK’s in-memory config cache. See Changing experiment traffic mid-flight.
  • Summary — /v4/experiments/summary returns aggregated performance metrics. Required: filter[experiment_uid][] (1–50 UIDs) — discover via GET /v4/experiments, then iterate the summary in batches.

​
Available endpoints

​
Authentication

All v4 endpoints require a Secret Key. See Authentication.