Once your A/B test is running, open it from the A/B test page to see how the two versions compare. Results can take a few minutes to appear after visitors interact with your quizzes.
The statistical significance card
At the top of the results you'll see a badge that reads either Significant or Not yet significant, along with a confidence percentage. "Significant" means we're at least 95% confident the difference between the versions is real — for example, "We're 96.2% confident that Version B performs better."
Below the badge, a sample progress bar shows how many quiz views you've collected against the recommended sample size. Small differences need lots of data, so early numbers can swing a lot.
Our advice: wait for the "Significant" badge before making a decision. Ending a test too early risks picking a version that only looked better by chance. Use Refresh data to pull the latest numbers (available once per hour).
Comparing the versions
Metric cards show each key metric with the lift — how much better or worse Version B is doing compared to Version A.
The comparison table lists Views, Completions, Completion rate, Orders, Revenue, Revenue per visitor, and Average order value side by side, with your primary metric at the top.
The trend chart shows your primary metric for both versions over time.
Question drop-off compares where shoppers abandon each version, question by question.
You can also filter all results by device (desktop or mobile).
Declaring a winner
When the test reaches significance, a Declare winner button appears for the leading version. You can also click End test at any time and choose:
Version A wins — Version A stays published; nothing changes on your storefront.
Version B wins — Version B is published and replaces Version A on your storefront.
End Without Winner — the test ends and your currently published quiz stays in place.
Ending a test can't be undone, so you'll be asked to confirm.
Automatic winners
If you set an end date and ticked Automatically publish a winner when the test ends, PerfectMatch handles this for you: the version with more conversions wins and is published automatically. If it's a tie, Version A is kept.