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

# A/B tests

> Show two discounts to your visitors and keep the one that sells more

An A/B test splits the visitors of one bundle between its current offer (variant A) and a new one (variant B), then shows you which converts better. When you end the test, the offer you choose becomes the bundle's offer.

## Start a test

<Steps>
  <Step title="Open the setup">
    In the bundle editor, click **Set up A/B test** above the preview. You can also go to **A/B tests** in the navigation and click **New A/B test**.
  </Step>

  <Step title="Set variant B">
    Variant A is your bundle as it is today. For variant B, change the discount type or value. Bundles with tiers let you change every tier. If a tier's title or subtitle mentions the discount, update the text too.
  </Step>

  <Step title="Choose the traffic split">
    The slider sets the share of visitors who see variant B. It starts at 50%.
  </Step>

  <Step title="Start">
    Click **Start test**. **Save as draft** keeps the test without showing it to anyone.
  </Step>
</Steps>

A bundle must be published to start a test, and it can run one test at a time.

## What visitors see

* Each visitor is assigned a variant on their first visit and sees the same one on every later visit from that browser.
* The discount a visitor was shown is the discount they get in the cart and at checkout.
* Visitors whose browser blocks both local storage and cookies see variant A and are left out of the results.

## Read the results

Open a test from **A/B tests** to compare the variants.

| Column | Meaning |
| - | - |
| Visitors | People who saw the variant. Each person counts once. |
| Add-to-cart rate | Visitors who added the bundle to their cart. |
| Orders | Bundle orders placed by visitors of the variant. |
| Conversion rate | Orders divided by visitors. |
| Revenue | Revenue of the bundle's products in those orders. |
| AOV | Revenue divided by orders. |
| Revenue / visitor | Revenue divided by visitors. Use it when variant B has a deeper discount: more orders can still mean less money. |
| Lift vs. A | Change in conversion rate against variant A, with the confidence that the difference is real. |

The banner above the table tells you where the test stands:

* **Collecting data**: a variant has fewer than 100 visitors, or the test has no order yet.
* **No clear winner yet**: the difference could still be chance.
* **Variant is winning**: the confidence is 95% or more.

<Tip>Let a test run for at least one full week, so that weekdays and weekends are both covered.</Tip>

## End a test

Click **End test** and choose the offer to keep. Choosing variant B replaces the bundle's discount with variant B's. Choosing variant A leaves the bundle unchanged. From then on, all visitors see the offer you kept.

Finished tests stay in the **A/B tests** list with their results.

## Limits

* A test changes the discount and the tier text. It can't change the bundle's products, tier quantities or design.
* Translated tier text applies to variant A only. Variant B shows the text you entered in the test.
* Avoid editing the bundle's offers while its test runs: variant B keeps the tiers it was created with.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.