A/B testing in PerfectMatch: AI Shopping Quiz lets you compare two versions of a quiz on your live store and see which one actually performs better — with real shoppers, not guesswork. You'll find it under A/B test in the app's navigation.
How it works
Every test compares two versions of a quiz:
Version A (current live version) — a published quiz your visitors currently see. This is your baseline.
Version B (new version to test) — an unpublished draft. It stays unpublished and is only shown to visitors who are part of the test.
While the test is running, PerfectMatch splits your visitors between the two versions. Each visitor consistently sees the same version, so their experience stays seamless — they never know a test is happening.
What you can test
Anything you can change in the quiz builder: question wording, the number of questions, branching, result pages, product recommendations, design, or placement. A good test changes one meaningful thing so you know what caused the difference.
Choosing a success metric
When you set up a test, you pick a Primary success metric — the number that decides which version wins:
Completion rate — the share of visitors who finish the quiz and reach the results page.
Conversion rate — the share of visitors who go on to place an order.
Revenue per view — how much revenue each quiz view generates on average.
Traffic split
You control how visitors are divided with the Traffic split slider — anywhere from 10/90 to 90/10, in steps of 10. A 50/50 split (the default) collects data fastest. If you'd rather be cautious with a big change, send less traffic to Version B, for example 70/30.
Ready to try it?
See "How do I create and run an A/B test?" for step-by-step setup, and "How do I read A/B test results and declare a winner?" to learn how to interpret your results.