A/B Testing & Version Comparison
Not "I think this is better" — a decision backed by real numbers. Compare two versions and see which one actually wins.
"I think this version is better" isn't a decision — it's a guess with confidence attached. A/B Testing & Version Comparison lets you run two or more real versions of a page or component against real traffic and read back a result backed by actual numbers.
Every result comes with a lift percentage, a confidence interval, and a statistical-significance badge, and you can track a primary metric (your main goal) alongside guardrail metrics (like page speed) at the same time, so a win on one number can't quietly hide a regression somewhere else.
What you get
Multi-variant support
Not just A/B — compare as many variants as you need, at once.
Statistical significance
Is the difference real or just noise — get a clear answer with confidence intervals and a significance badge.
Primary and guardrail metrics
Track your main goal (e.g. clicks) while watching for side effects (e.g. a page-speed regression) at the same time.
Results screen
Lift percentage, confidence interval, and trend over time for every variant, on one screen.
How it works
Define your variants
Set up two or more versions of a page or component — not limited to a strict A/B split.
Traffic splits automatically
Visitors are assigned to a variant and see it consistently across their session.
Read the result, not just a guess
The results screen shows lift, confidence interval, and a significance badge, so you know whether the difference is real.
Frequently asked questions
Am I limited to two variants?+
No — compare as many variants as you need in the same test, at once.
How do you know a result isn't just noise?+
Each result comes with a confidence interval and a statistical-significance badge, calculated from the actual traffic each variant received.
Can I track more than one metric per test?+
Yes — track a primary metric (your main goal, e.g. clicks) alongside guardrail metrics (e.g. page speed), so a win on one number doesn't hide a regression elsewhere.
What happens to a visitor if I end the test?+
Once a test ends you decide which variant to keep permanently; historical results stay available for reference.
Other features