> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mention-me.com/llms.txt
> Use this file to discover all available pages before exploring further.
# How to Set Up an A/B Test
> Start an A/B test against a live version, or split a draft section into two variants, then configure how the test runs and concludes.
There are two ways to start an A/B test, depending on whether you are testing against something that is already live. Either way, you configure the test in a short wizard on the section you are testing. On a draft with no test yet, the insights panel also nudges you with a **Run an A/B test** recommendation. For the concepts behind these, see [how A/B tests work](/knowledge-v2/referral/campaigns/ab-testing/how-ab-tests-work).
## Test Against a Live Version (Control Versus Variant)
Use this when you have a live version and want to test a change against it. The live setup becomes the **control**.
On the live version, click **Set up A/B test** (it also appears as a recommendation in the insights panel). Mention Me creates a new draft version based on the running version.
If the live version already has more than one variant, choose the **baseline variant** to test against. Enter a **Version name (internal)**, then click **Create test**. Mention Me creates a new draft version based on the live version, with your chosen baseline as the control.
On the section you want to test (for example Rewards or Sharing), click its **+ A/B test** button. The baseline stays as the control (Variant A); edit **Variant B** to create the change you want to test.
## Split a Section on a Draft (Variant Versus Variant)
Use this when you are building something new and want to compare two fresh ideas, with no live baseline.
On a draft version, click the **+ A/B test** button on the section you want to test (for example Rewards or Sharing). The section splits into **Variant A** and **Variant B**.
Edit both variants so they differ. Neither is a control here; the two are compared as equals.
You can A/B test six sections of a version, each via its **+ A/B test** button: **Branding**, **Images**, **Rewards**, **Sharing**, **Referrer journey** and **Referee journey**.
## Configure the Test
Setting up a test on a section opens a wizard with three steps.
A summary of what will happen: your current setting becomes the control (or Variant A), you edit the other variant, and traffic is split 50/50. Click **Next**.
Choose the **test metric** that decides the winner, **Conversion rate** or **Share rate** (the recommended option is marked). Then set how fast the test should run:
* For a test against a live baseline, pick a **test speed**: **Normal** (highest confidence), **Fast**, or **Fastest**. Each shows its confidence target and estimated duration.
* For a variant-versus-variant test with no baseline, tell Mention Me your weekly order volume (**Less than 200 orders a week** or **More than 200 orders a week**). This sets the confidence target, and the test runs for a fixed 8 weeks.
Decide what happens for each outcome:
* **If a winner is found**. The winning variant is promoted automatically.
* **If a better variant is found (but no winner)**. Choose **Allocate to the best performer** (recommended), keep the control, or pick a specific variant.
* **If there's no meaningful difference**. Choose to keep the control or allocate to the variant.
Click **Save** to finish.
## Edit and Refine
Next, make your variants meaningfully different and refine the test before launch: edit Variant B, use **View differences**, and adjust or remove the test from the section's manage menu. See [manage an A/B test](/knowledge-v2/referral/campaigns/ab-testing/manage-an-ab-test).
## Launch the Test
An A/B test launches like any other version. See [launch a campaign and version](/knowledge-v2/referral/campaigns/manage/launch-a-version). After launch, follow it on the [monitor an A/B test](/knowledge-v2/referral/campaigns/ab-testing/monitor-an-ab-test) page.