A/B Testing
Test every idea.
Ship the winner.
Build variants, launch the experiment, and see the results in one place. No code, no dev queue.
Run the whole test in one place
Create variants in minutes. Ask Maker to generate them, or make your own. Every version stays on brand.
Set it up in a prompt or by hand. Both get you a live test with the split and goal in place.
See what wins, then keep going. Track every variant in real time. Publish the winner and start the next test.
Everything a real experiment needs
Edit any element by describing it. Click what you want to change and say how. Maker applies it to that variant only.
Know when a result is real. Maker calculates the probability each variant wins and flags one when the data is conclusive.
Measure in revenue, not just clicks. Connect GA4 once. Every test reports conversions, transactions, and revenue per variant.
Decide how many people see it. Start at 10% of traffic and expand when you’re confident. Target by device, location, URL, or time of day.
“The results of constant A/B testing have shown, our conversion rate has improved 4X in 12 months!”
FAQs
No. You point Maker at a page, describe the change you want, and it builds the variant. Splitting traffic, collecting the data and publishing the winner all happen in Maker, so nothing waits on a release cycle.
Yes. Maker sits in front of the site you already have, so your CMS carries on doing its job. Shopify, Magento, BigCommerce, WooCommerce, Salesforce Commerce Cloud and custom builds are all supported, and if you have a product feed it can be connected. Products, prices and inventory stay in sync.
Nothing is migrated and no template is rebuilt. Variants are served on top of the pages you already publish.
As many as you like. The limit worth respecting is statistical rather than technical: every extra variant splits the same traffic further, so a four-way test takes far longer to reach a conclusive result than a two-way one.
Two or three is the usual starting point. Maker flags when a split is too thin to call a winner in a sensible timeframe, so you find out before you have spent a month waiting.
Anything on the page: layout, imagery, copy, the order of sections, the buy block, the whole template. If you can describe the change, it can be a variant. You are not limited to swapping a button colour.
No. Variants are served from the same infrastructure as the rest of your Maker pages, so a visitor in a test loads the page the same way a visitor outside one does. There is no flicker step where the original paints first and then gets rewritten.
Until the data says something, not until a date passes. Maker tracks the probability that each variant is the winner and tells you when the result is conclusive, rather than leaving you to eyeball a chart.
As a rule of thumb, give a test at least one full business cycle, usually two weeks, so weekday and weekend behaviour are both represented.
Traffic is split by the percentage you set, and a visitor stays on the variant they were first assigned, so nobody sees the page change between visits. You can also narrow a test to a device, location, URL pattern or time of day.
Nothing, until you say so. Publishing the winner makes it the live page. The other variants stay in the project as a record of what you tried, so the next test starts from what you already learned.