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

# A/B testing

> Test collection variants against each other with real statistics.

Run A/B tests per collection to find out whether a merchandising change
actually performs better — different content blocks, imagery or ordering.

## How a test works

1. Pick a collection and set up variant **A (control)** and variant **B**.
2. Give the experiment a name, a start and end date, and optionally an email
   for notifications.
3. While the test runs, traffic is split between the variants.

## Reading the results

For each variant you get daily time series of **page views, clicks,
add-to-carts and purchases**, plus the per-view rates (click-through rate,
add-to-cart rate, purchase rate). Add-to-carts and purchases only count
products that are actually in the collection.

Statistical significance is computed with a **two-tailed z-test** for each
rate metric, so you can tell a real effect from noise. When the test
concludes, an **AI-written summary** states which variant won, with what
confidence, and suggests next steps.

<Note>
  Give tests enough time and traffic. Low-traffic collections need longer test
  windows before the p-values become meaningful.
</Note>