// a/b test significance

Is your winner real,
or just noise?

Most teams call an A/B test the second one variant pulls ahead, and half the time they're shipping randomness. Enter your numbers and get the honest answer: the uplift, the confidence, and whether you've actually got a winner yet, or need more data.

Variant A · control

Variant B · challenger

reset try sample numbers
// what should you test next?
Not sure what to test, or what your numbers are telling you?

Send me your funnel and current numbers and I'll tell you the highest-leverage experiment to run next, free. It's part of how I help founders and marketers turn traffic into revenue instead of guessing.

// what this is

The A/B Test Significance Calculator tells you whether the difference between two variants is a real effect or random chance. Enter visitors and conversions for each, and it returns the conversion rates, the relative uplift, the statistical confidence (from a two-proportion z-test), and a plain-English verdict, plus how much more data you need if the result isn't conclusive yet. Runs in your browser, nothing stored.

Who this is for

Marketers, founders, and growth teams running experiments on landing pages, emails, ads, or checkout flows. If you've ever eyeballed two conversion rates and wondered "is B actually better?", this gives you the statistically honest answer instead of a hunch.

How significance is calculated

The tool runs a two-proportion z-test, the standard method for comparing two conversion rates. It computes each rate, the pooled standard error, and a z-score, then converts that to a confidence level. At 95% confidence, there's roughly a 1-in-20 chance the difference is random. It's the same math behind commercial testing platforms, shown transparently.

What the verdict means

A significant result means the difference is unlikely to be luck at your chosen threshold, so you can act on it. Not yet significant means the data can't rule out chance, keep the test running. The calculator also estimates the additional sample size per variant you'd likely need to detect the current effect, so you know whether to wait or move on.

Common A/B testing mistakes

Calling tests early the moment one variant leads, stopping the instant you hit 95% (peeking inflates false positives), testing tiny changes that need huge samples, and ignoring the size of the uplift. A 0.1% "win" at 95% confidence on a low-traffic page is rarely worth shipping. This tool shows both the confidence and the practical size of the effect so you can judge both.

Frequently asked questions

What confidence level should I use?

95% is the standard for most marketing tests. Use 99% for high-stakes or hard-to-reverse changes. Below 90%, treat the result as inconclusive.

How many visitors does an A/B test need?

It depends on your baseline rate and the size of the change you want to detect. Smaller expected lifts need much larger samples. When your result isn't significant, this tool estimates how many more visitors per variant you likely need.

Can I test more than two variants?

This calculator compares two at a time (A vs B). For multi-variant tests, compare each challenger against the control separately, and note that testing many variants at once raises the chance of a false positive.

Does it store my data?

No. Everything runs in your browser and nothing is sent anywhere or saved.

Can you help me build a testing program?

Yes. Send me your funnel and I'll point you to the highest-leverage experiment to run next, free, as part of how I help teams turn traffic into revenue.

Uses a two-proportion z-test (two-tailed). Confidence and required-sample estimates are standard approximations, not a substitute for a full power analysis on high-stakes decisions. Everything runs in your browser; nothing is stored or sent.
Built by Vishesh Kulshrestha · part of vishkul/tools.