---
title: Experiments (A/B tests)
description: Test a changed version of your form against the live one and ship the winner with confidence.
icon: flask-conical
sidebar:
  badge: Pro
search:
  keywords: [a/b test, split test, experiment, variant, conversion, optimize]
---

:::note Experiments are available on **Pro** and above. :::

An experiment splits your traffic between the live form (**Control**) and one or more changed versions (**challengers**), then tells you which one performs better.

## Create an experiment

1. **Start a new experiment**

    Open a form, click **Experiments**, then **Create form experiment**.

2. **Write a hypothesis**

    State what changes, for whom, and why. For example: *"Reducing the first
    page to three questions will increase qualified completions from paid
    traffic."*

3. **Choose the primary outcome**

    - **Completion rate**: completions among people who started. - **Start
    rate**: visitors who start answering. - **Qualified completion rate**:
    completions where a chosen field matches, such as *Company size* at least
    50.

4. **Set the expected baseline and worthwhile lift**

    Finerlise uses these to estimate how many responses you need.

5. **Create the challenger**

    Click **Create challenger**, make your changes in the editor, then save it
    as a variant. Variants are locked snapshots, so they can't drift during the
    test.

6. **Set traffic and audience**

    Split **Experiment traffic** between versions (it must total 100%).
    Optionally use **Audience targeting** by UTM source, medium, campaign,
    content or any query parameter.

7. **Launch**

    Check the **Launch checklist** and click **Launch experiment**.

## Read the results

| Measure | Meaning |
| --- | --- |
| **Win probability** | How likely each version is to be the best. |
| **Expected loss** | How much you'd lose, on average, by picking this version if it's actually worse. |
| **Funnel**, **Friction**, **Segments** | Where each version loses people, and how results differ by campaign, referrer or device. |

When the result is clear (**Ready to decide**), end the test and **ship the winner**. The winning version becomes your live form.

You can pause and resume an experiment at any time.

**[A/B test your signup form](/guides/ab-test-your-signup-form)**

A worked example.

**[Drop-off analysis](/responses/drop-off-analysis)**

Find what to test.