Build consistent campaign links, check the GA4 channel, and export in bulk.
- Runs in your browser
A/B test calculator
Find out whether version B really beat version A, how sure you can be, and how big the lift could plausibly be. Or plan your next test: how many visitors it needs and how long it will run.
- Free
- No sign-up
- Runs in your browser
A/B test calculator
Loading the tool…
Runs in your browser. Nothing you add leaves this page.
Three steps
How it works
Enter visitors and conversions for each version, or switch to Plan a test.
Pick a confidence level, and whether to test for a change in either direction.
Read the result, the p-value, and the lift’s likely range, with the math shown.
GuideReviewed
How to read an A/B test without fooling yourself
Contents
What the significance check does
The calculator runs a two-proportion z-test, the standard test for comparing two conversion rates. It asks how often a gap as large as yours would show up by chance if the change made no real difference. That’s the p-value. A low p-value means your gap would be unusual if nothing had changed; it isn’t the chance that B is better.
When the p-value falls below your significance level, 5% at 95% confidence, the result is called statistically significant. The calculator also shows the likely range for the lift, so you can see how small or large the real effect could be.
Plan the sample size before you start
A test needs enough visitors to detect the smallest lift worth acting on. The planner uses the standard sample size formula for two proportions: tell it your current conversion rate, the lift you care about, your confidence level, and power (the chance of catching the lift if it’s real), and it returns the visitors each version needs.
Smaller lifts need far more visitors. At a 3% conversion rate, detecting a 10% relative lift (3% to 3.3%) with 95% confidence and 80% power takes about 53,000 visitors per version.
Mistakes that create false winners
Most bad test calls come from how the test was run, not from the math.
- Peeking: checking results every day and stopping the moment they look significant makes false winners far more likely. Decide the sample size up front and stop there.
- Many variants: each extra version compared with the original is another chance of a fluke. The planner splits the false-winner risk across comparisons when you test more than two versions.
- Small samples: with fewer than about 10 conversions or 10 non-conversions in a version, the test’s math is unreliable. The calculator warns you.
- Partial weeks: visitors behave differently on different days, so run whole weeks.
Significant isn’t the same as important
Statistical significance says a gap would be unusual if there were no real difference. It doesn’t measure how big or valuable the lift is. Look at the low end of the likely range: if even that would be worth shipping, you have a strong result. If the range runs from barely positive to large, you may need more data before you plan around it.
Questions, answered
What people ask before they use the A/B test calculator.
Enter the visitors and conversions for both versions. The calculator runs a two-proportion z-test and tells you whether the p-value is below your significance level, such as 5% at 95% confidence, along with the likely range for the lift.
More free tools
All free toolsCheck headlines and descriptions against Google, Meta, and LinkedIn limits, with a live ad preview.
- Runs in your browser
ROAS, break-even ROAS, profit after ad spend, and the target to aim for.
- Runs in your browser
Related guides
- What is market research? Types, methods, and B2B examplesLearn what market research is, how primary, secondary, qualitative, and quantitative research fit together, and how to choose a method for a B2B decision.
- How to write a creative brief for a marketing campaignLearn to write a campaign creative brief, compare concepts, and review creative work with a worked B2B SaaS example and copyable template.
- How to measure content marketing ROI, including what attribution can't seeMeasure content marketing ROI with the full cost, three tiers of return, attribution blind spots, holdout tests, a worked example, and a report template.
- Drip campaigns: how to plan lifecycle email sequences that don't read like automationPlan drip campaigns that work: types, triggers and exits, timing, writing, lead scoring, compliance, measurement, and three full example sequences.
From Ampere
A good test needs versions worth testing.
Ampere is an AI marketing agent that drafts the landing page copy, ad variants, and emails you test against each other, all from your saved Brand memory and ready for your review.
No card required