Developer docs

Experiments

Run a published treatment against control. Allocation, uplift, and confidence when there is traffic. A person publishes the winner. We do not silently pick a variant while you sleep.

How assignment works

decide() returns an experiment arm for both control and treatment so measurement has a real control group. The SDK sends Experiment Assigned, deduped per tab session. Preview traffic never assigns.

js Emitted by the SDK — you do not fire this yourself

executen.track("Experiment Assigned", {
  experiment_id: "exp_…",
  experiment_arm: "treatment",
  experiment_version: 3,
  page: "/"
});

Conversion is still orders from your order events. Pause or complete when you are ready. Action Center reminds you a winner is waiting.