> ## Documentation Index
> Fetch the complete documentation index at: https://help.experro.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Variants

Variants sets the shape of the test — how many versions you are comparing and what share of traffic each one receives. What each variant actually shows is configured later, on the **Experience** step.

## Define Variant Traffic Allocation

Every A/B test includes a **Control Group**. The control is your existing experience and the baseline every variant is measured against. It cannot be removed.

Add variants with **Add Variant**. The experiment supports four in total, including the control group.

<Frame>
  <img src="https://mintcdn.com/experro/vL9u48aLzyWUbwlI/images/experiments/a_b_testing/imagesource_variants.png?fit=max&auto=format&n=vL9u48aLzyWUbwlI&q=85&s=71a184a0f1c1b70c498cf0271e511061" alt="" width="1440" height="900" data-path="images/experiments/a_b_testing/imagesource_variants.png" />
</Frame>

| Element                | Behavior                                              |
| ---------------------- | ----------------------------------------------------- |
| **Variant Name**       | Control Group, Variant A, Variant B, Variant C.       |
| **Traffic Allocation** | The share of entering traffic this variant receives.  |
| **Delete**             | Removes a variant. Unavailable for the control group. |
| **Add Variant**        | Disabled once four variants exist.                    |

## How the Split Works

On first load the experiment starts at 50% control and 50% Variant A. Adding a variant redistributes traffic equally, and you can then set each share by hand.

The total must equal 100%. **Total Allocation** turns red when it does not, and the experiment cannot start.

## Worked Example

Carrying on from [Target Audience](/experiments/a_b_testing/target_audience), 100 visitors enter the experiment. With four groups at 25% each:

* Control Group — 25 visitors
* Variant A — 25 visitors
* Variant B — 25 visitors
* Variant C — 25 visitors

## Key Considerations

* At least one variant besides the control group is required.
* The allocation you set here applies to every experience type in the experiment.

<Tip>
  More variants split your traffic thinner, so each one takes longer to reach a trustworthy result. Two or three groups usually answer the question faster than four.
</Tip>

<Note>
  Changing allocation after an experiment starts does not move visitors already assigned. New visitors are distributed until the split matches the new configuration.
</Note>

## What's Next

* [Experience](/experiments/a_b_testing/experience)