Experiments
Prove what works before you roll it out
Test recommendations, overlays, and AI chat against a control group, protect key metrics with guardrails, and keep the winner.
Free-returns overlay
What it does
Test without risking what already works
Every experiment runs on the same intent and audiences as the experiences it tests.
Test what customers see
Run experiments on recommendations, overlays, and AI chat.
- Recommendations
- Overlays
- AI Chat
Pick the metric that matters
Use conversion rate or click-through rate as the primary metric.
- Conversion rate
- Click-through rate
Protect what matters
Guardrails watch the metrics you can’t afford to hurt while a test runs.
- Guardrail metrics
- Control group
Resolve and roll out
Decide yourself or let experiments resolve automatically, then keep the winner.
- Manual
- Auto-resolve
- Keep the winner
Lifecycle
From draft to decision
Every experiment follows the same path, so everyone knows where a test stands.
- 1
Draft
Choose the experience, variants, primary metric, and guardrails.
- 2
Running
Customers are split between the control and each variant.
- 3
Collecting
Results build until they’re conclusive or guardrails step in.
- 4
Winner eligible
A variant clearly beats the control on the primary metric.
- 5
Resolved
Keep the winner or return to the control, manually or automatically.
Results
Results you can act on
See how each variant compares with the control, with confidence intervals, guardrail checks, and a clear next step.
- Confidence intervals
- Guardrail checks
- A clear decision when a winner emerges
- Control
- Variant