55 of 56
A/B testing
Banner, block, photo or button copy — test two versions on real shoppers and keep the one that performs better.
Why it matters
A fashion store is full of judgment calls. Does a dress sell better on a model or as a clean flat lay? Which banner should open the season on the homepage? Should the button say "Add to cart" or "Add to outfit"? Usually these questions are settled by taste or by the loudest voice in the meeting. An A/B test replaces the debate with a check: some shoppers see one version, others see the other, and the store measures which one performs better.
How it drives sales
- Decisions based on data, not hunches. You see which version leads to more add-to-carts or orders — on your own traffic and your own assortment.
- Small wins add up. A photo, a banner headline, button copy, the order of blocks — each test is a small step, and together they reshape the funnel.
- Less risk in a redesign. A new homepage or a new product page can go to half of your visitors first, and to everyone only after it proves itself.
- Faster iteration. The winner goes live in one click — no developer ticket, no waiting for a release.
How it works
- A manager opens A/B tests in the CMS and picks what to test: a homepage banner or slider, a block from the page builder, block order, a product photo, button copy or a popup variant.
- Sets up variants A and B — for example, "Add to cart" vs "Add to outfit".
- Uses a slider to set the traffic split (50/50, or a more cautious 80/20), chooses a goal — add to cart, order or click — and the test duration.
- The storefront splits visitors: each shopper is assigned to a group and sees the same variant on every return visit.
- Results update in real time: conversion per variant, visitor counts, a daily chart and a confidence indicator that shows whether there's enough data to draw a conclusion.
- Once there's enough data, the manager clicks "Apply winner" — variant B becomes the default for everyone.
The numbers in the video are demo data: every test's outcome depends on your traffic, your assortment and what exactly you're comparing.
What you can test
| Element | Example test |
|---|---|
| Homepage banner | On-model shoot vs still life |
| Page builder block | Lookbook vs product grid |
| Block order | New arrivals above sale, or the other way round |
| Product photo | First photo on a model or a flat lay |
| Button | "Add to cart" vs "Add to outfit" |
| Popup | Signup discount vs a gift |
Built-in safeguards
- One shopper, one variant. The group sticks to the visitor, so nobody sees the store "flicker" between versions.
- Stop a test at any time — everyone goes back to the original version, and the data is kept.
- The confidence indicator tells you when it's too early to call it, protecting you from decisions based on random early spikes.
- Test results stay in the history and complement your analytics reports.