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

# A/B test your checkout

> Show shoppers two or more checkout designs and see which one earns more

An A/B test splits your shoppers between your live checkout and one or
more other designs, then shows you how each one performs — so you can
make the winner live with confidence instead of guessing.

## Before you start

* You need at least two **published** checkout templates: the one that is
  live, plus another one to test against it. Drafts can't join a test.
* Tests run on the checkout page. The thank-you page can't be tested yet.
* One test runs at a time.

## Starting a test

1. Open the checkout designer and click **A/B test**, next to the template
   picker.
2. On the **Setup** tab, give the test a name. **Live now** shows the
   template shoppers see today.
3. Under **Also show shoppers**, tick up to three more published templates.
   Use each slider to set the share of shoppers who see that template; your
   live template keeps whatever is left.
4. Each template starts as **All shoppers**. Change that to **Mobile
   shoppers**, **Desktop shoppers**, or a country to show the template to
   that audience only.
5. Choose a **Goal** — the number the test is judged on: **Conversion
   rate**, **Revenue per shopper** (the default, since it combines
   conversion and order size), or **Average order**.
6. Choose **Run for**: **Until I stop it**, a set number of days, or until
   each template has seen a set number of shoppers.
7. Choose **When it ends**: **Tell me**, or **Make the winner live**.
8. Click **Start test**.

## While a test runs

* The **A/B test** button shows **Test running** and how many days it has
  been going. The template picker shows each template's share of traffic.
* Each shopper sees one template for their whole visit, and again if they
  come back, so nobody gets switched mid-checkout.
* Your own previews are never counted.
* **Set live** is unavailable on the toolbar while a test runs. To change
  the live checkout, end the test from **A/B test**.
* You can publish changes to a template during a test; its row in the
  results is marked **changed** so you know the numbers span two versions.
  You can't change the shares or the templates in a running test — stop it
  and start a new one.

## Reading the results

Open **A/B test** and switch to the **Results** tab. The strip at the top
shows how far the test has run, total shoppers, the goal, and — when the
winner is set to go live automatically — the date that happens.

The table shows, for every template in the test:

* **Shoppers**, with the number who arrived from Meta, TikTok, or Google
  ads in brackets.
* **Conv. rate** — the share of shoppers who completed a purchase.
* **Avg order** — the average order value.
* **Rev / shopper** — revenue per shopper, with each template's difference
  from your live checkout shown beside it.
* **Status** — **Ahead** when a template is clearly beating your live
  checkout on the goal with enough shoppers to trust, **Too early** until
  then.

Numbers refresh about once an hour, so a new test shows zero for its first
hour.

## Ending a test

* Click **Make live** on any template to make it your live checkout and end
  the test. Shoppers already mid-checkout finish on the page they started
  with.
* Click **Stop test** to end it and keep your current live checkout.
* If you chose **Make the winner live**, the winner goes live at the end of
  the run only when it is clearly ahead. If no template is clearly ahead,
  your live checkout stays and you get an email so you can decide.

Ended tests stay listed under **Past tests**. Open one to see its final
numbers and winner; you can still make any of its templates live from
there.

## Good to know

* Refunds don't change a test's results.
* The ad counts in brackets cover shoppers who clicked a Meta, TikTok, or
  Google ad on the same device they bought on. Everyone else is still
  counted in the total.
* Starting, stopping, and making a winner live need the same permission as
  publishing checkout designs.