Skip to content
Experiments

Experiments

Run A/B tests on your flows to find what works best.

Experiments

Experiments let you A/B test your flows to find the version that drives the best results. Instead of guessing which copy, layout, or structure works best, you can run a controlled test and let the data decide.

How experiments work

An experiment splits eligible users into groups. Each group sees a different flow that you have already built, and Setgreet tracks a chosen metric for each one. You decide which flow performed best based on the results.

The basic structure:

  • Control -- a flow you have already created (typically the current version).
  • Variant(s) -- other flows you have already created that represent alternative approaches.
  • Traffic split -- the percentage of eligible users assigned to each variant.

The SDK handles variant assignment automatically. Each user is consistently assigned to the same variant for the duration of the experiment, so their experience is stable.

Variants are existing flows, not copies of the control. Build each variant as a separate flow in the flow builder before creating the experiment, then select them when configuring the experiment.

Experiment lifecycle

StatusDescription
DraftThe experiment is configured but not running. You can still edit variants and traffic splits.
RunningThe experiment is live. Users are being assigned to variants and metrics are being collected.
PausedThe experiment is temporarily stopped. No new users are assigned, but existing data is preserved.
CompletedThe experiment has ended. Results are final and you can mark a winning variant.

What you can test

Experiments work at the flow level. You can test any difference between flow variants:

  • Different copy or messaging
  • Different screen counts or ordering
  • Different component layouts
  • Different images or media
  • Different call-to-action buttons

Available metrics

When creating an experiment, you choose one primary metric to measure success:

MetricDescription
Completion ratePercentage of users who reached the final screen of the flow.
Conversion ratePercentage of users who triggered a defined conversion goal.
Dismiss ratePercentage of users who closed the flow before completing it.

You can only pick one primary metric per experiment. It is the basis for calculating lift over control and for the winner recommendation.

Experiments require the Growth plan or higher. The Starter plan does not include A/B testing.

Next steps

On this page