> ## Knowledge Base Index
> Fetch the complete knowledge base index at: https://help.poool.fr/sitemap.xml
> Use this file to discover available pages before exploring further.
> Pure-Markdown content can be obtained by appending a '.md' suffix to the content URLs listed in the sitemap (without the trailing slash).

# How do I set up an A/B test with Split?

In the Poool Split platform, you can run A/B tests for each of your Engage elements. The goal is to allow you to test different hypotheses in real time and identify optimization opportunities.
With Poool, you can set up these tests in just a few clicks.

||| Today, the tool allows you to test only two versions (Element A and Element B). The ability to test multiple versions is currently under development.


# Testing criteria available

You can test several criteria of elements:

* The overall appearance of your Engage element: which design is more impactful?
* A component within your appearence: what impact does the CTA have on performance?
* Messaging: which wording performs better?

|| Variants are configured within the appearance settings. The display conditions defined for both A/B-tested elements remain identical.


# General operation

Users are randomly assigned to one of the two versions, based on the distribution defined when setting up the A/B test.
For example, if you choose a 70% / 30% split, each impression has a 70% chance of showing version A and a 30% chance of showing version B.

# Setup

Activating an A/B test involves several steps.

First, create the element you want to test (version A) in Engage. Once ready, go to Split and create a new A/B test by giving it a name.

Then select the element you want to test and create a variant (version B). You can then modify its design and content to define the alternative version.

Finally, define the audience distribution (for example 50% / 50% or 70% / 30%) and activate the A/B test.

${youtube}[Split Exemple](gcOBtHS6K6Q)


Once created, the variant can also be recreated in Engage and reused or edited later like any other element.


![](https://storage.crisp.chat/users/helpdesk/website/-/5/f/5/0/5f50f52b22503800/capture-decran-2026-07-03-a-17_98helk.png =400x400)


||| The variants share the same display conditions and targeting rules. Only the appearance or content of the tested element can be modified.

# Best practices

To run your A/B test effectively, we recommend following these steps:

* Define both variants, the duration, and the scope of the test
* Choose the KPIs for analysis
* At the end of the test period, compare the performance of both versions
* Select the version with the best results
* Document the test (versions tested, duration, scope) to iterate on new hypotheses
It is best to test variations on a large sample size to obtain reliable results. To calculate an optimal sample size based on a given factor, several tools exist, including [this one](https://www.evanmiller.org/ab-testing/sample-size.html).

Variant A:
![](https://storage.crisp.chat/users/helpdesk/website/-/5/f/5/0/5f50f52b22503800/capture-decran-2026-07-03-a-17_d4rgtj.png =600x113)


Variant B:

![](https://storage.crisp.chat/users/helpdesk/website/-/5/f/5/0/5f50f52b22503800/capture-decran-2026-07-03-a-17_141czc5.png =600x113)
# Results analysis

In this first version of Split, A/B test statistics are not yet available in the Poool dashboard.
To analyze variant performance, you need to implement tracking in your analytics tool. This will allow you to compare the metrics of your choice (impressions, clicks, conversions, etc.) and identify the best-performing variant.

To learn more about setting up tracking and analyzing results, refer to the dedicated article.


Once your first A/B test is complete, don’t stop there. Form new hypotheses and keep testing!