Updated 20 October 2026 · Page Mix
1. Decide what you're testing and why
Start with a hypothesis tied to a real problem. For example: "Visitors from ads don't scroll past the first screen. A page that puts the key benefit and reviews above the fold will raise add-to-cart rate." Your analytics, session recordings and customer questions are the best sources of hypotheses.
Test meaningful differences. A new layout or a different order of persuasion is more likely to show a measurable effect than a new button colour, especially on a store without much traffic.
2. Choose one primary metric
Pick it before you start: conversion rate for the product, add-to-cart rate, or revenue per visitor. Watch others too, but don't switch the goal after seeing results.
3. Build the variation
For a product page, the variation is usually an alternate product template (see how to create one). Options:
- Duplicate your current template and rework it in the theme editor (quickest for small changes).
- Have a developer build new sections (best for new functionality).
- Use Page Mix to create variations from reference pages (quickest for a new layout or style).
Before testing, check the new page properly: buy box, variants, mobile layout, speed, and any app blocks. Our pre-publish checklist helps.
4. Pick a way to split traffic
- Shopify Rollouts. Rollouts (Markets → Rollouts) includes an Experiment type that compares a treatment with a control, 50/50 by default. It works on themes, so it fits theme-level changes. See Shopify's documentation for plan requirements and what it can change.
- A testing app. Several Shopify apps split traffic between templates or URLs. Template assignment is stored on the product in Shopify, so confirm how your app serves an alternate product template.
- Sequential test. Assign the new template for a period and compare with a similar period before. It's cheap, but confounded by seasonality, promotions and ad changes.
5. Run it long enough
- Run in whole weeks to cover weekday and weekend behaviour.
- Don't stop early because one side is ahead. Decide the duration or sample size up front.
- Avoid launching big promotions or changing ads mid-test, or note them if you must.
- If traffic is low, expect wide uncertainty. Test bolder changes, or accept a directional answer.
6. Read the result honestly
Look at the primary metric and its uncertainty, not just which number is higher. Check results by device: a page can win on desktop and lose on mobile. If the result is unclear, that's useful too: the change probably isn't large, so keep the page that's easier to maintain and test something bolder next.
7. Roll out and record
Assign the winning template to the products it suits, and keep a simple log of each test: hypothesis, dates, result, decision. Over time that log is worth more than any single win.