> ## Documentation Index
> Fetch the complete documentation index at: https://docs.flow-board.co/llms.txt
> Use this file to discover all available pages before exploring further.

# A/B Testing

> Run A/B tests on flow variants and promote the winning experience.

Labs is where you run controlled experiments. It helps you compare two or more variants on the same audience instead of guessing which onboarding performs best.

## What Labs is made for

Use Labs when you want to:

* Test different onboarding messages or sequences
* Compare conversion between two flow variants
* Validate a redesign before rolling it out to everyone
* Promote a winner after the results are clear

## How Labs works

Labs sits on top of audiences. The audience defines who is eligible, and the experiment compares how different variants perform for that same group.

In practice, that means:

* You choose the audience rule the test should use
* You assign the competing flow variants
* Flowboard tracks traffic and results while the test is running
* You stop the test and promote the winner when you are ready

## What you can do in Labs

<AccordionGroup>
  <Accordion title="Monitor active experiments">
    Review running tests and compare how each variant is performing against the others.
  </Accordion>

  <Accordion title="Review archived experiments">
    Keep a history of past tests so your team can understand what was tried and what won.
  </Accordion>

  <Accordion title="Sync tests">
    Refresh experiment data when you need the latest test state reflected in the dashboard.
  </Accordion>

  <Accordion title="Promote a winner">
    End the experiment and send live traffic to the best-performing variant.
  </Accordion>
</AccordionGroup>

## When to use Labs vs a normal rollout

<CardGroup cols={2}>
  <Card icon="flask" title="Use Labs">
    Choose Labs when your goal is to learn which variant performs better with the same audience.
  </Card>

  <Card icon="rocket-launch" title="Use a normal rollout">
    Skip Labs when you already know which flow should go live and only need controlled delivery through audiences.
  </Card>
</CardGroup>

## Best practices

* Test one major idea at a time.
* Use clear experiment names that describe the hypothesis.
* Let the experiment run long enough to gather meaningful traffic.
* Avoid changing audience rules in the middle of a test unless necessary.
* Record why a winner was chosen so future teams can learn from the result.

<Tip>
  The cleanest experiments compare variants that serve the same user segment and aim at the same success metric.
</Tip>

## Related guides

* [Target flows with audiences](/features/audiences)
* [Measure performance in Pulse](/features/pulse)