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

# Experiments (A/B tests)

Experiments let you **compare two or more versions of a workflow path** and see which one performs better for your business goal (e.g. reduce cancellations, increase renewals, recover failed payments).

An experiment is a **split step**: it routes each customer into exactly one branch (**Variant A**, **Variant B**, …) based on your chosen traffic split.

![Experiment split example](https://cdn.juo.io/content-uploads/experiments_1_15b28fb826.png)

## How to create an A/B test

In **Admin Panel → Workflows**:

1. Create or open a workflow (start from a trigger).
2. Add an **Experiment (A/B test)** step where you want the split to happen.
3. Create **2+ variants** (A/B/C…) and name them clearly (e.g. “Discount 10%”, “Offer pause”).
4. Set a **traffic split** (typically 50/50 for A/B).
5. Connect each variant branch to the actions/conditions you want that variant to run.
6. Save and publish.

## Variants persistence

* **Within a single customer session / run**: once a customer is assigned a variant, they stay in that variant while they move through the workflow (even if the run pauses and resumes later).
* **Across separate runs** (e.g. the same customer enters the workflow on another day): variant assignment may not be guaranteed to stay the same.

## Reading results

In the Experiment panel you will see:

* **Enrollments**: how many customers entered the experiment step per variant
* **Conversion rate** (if shown for your workflow): success rate per variant

## Choosing a good traffic split

* **50/50** is the default for A/B tests and gives the fastest learning.
* Use **90/10** only when you’re rolling out a risky change and want a smaller exposure.
* The editor will automatically adjust variant traffic so it sums to **100%**.

![Traffic split example](https://cdn.juo.io/content-uploads/experiments_2_671544e97b.png)

## When to use an experiment

Use experiments when you have a clear question like:

* Does offering **10% off** retain more subscribers than offering **skip**?
* Is it better to show **free product** vs **discount** at step 1?
* Does changing the copy or button labels increase acceptance?

Avoid experiments when:

* You’re still unsure what problem you’re solving (start with a single workflow and iterate).
* You can’t measure success (no clear outcome / metric).
* You’re planning many changes at once (hard to interpret results).

<Tip>
  Make sure each variant branch ends up doing something meaningfully different (offer, copy, timing, channel, etc.).
</Tip>