A/B test significance calculator

Q: What is statistical significance?

The probability that the observed difference between variants is not due to chance. If p-value < 0.05, we are 95% confident the difference is real, not random fluctuation.

Q: How many conversions are needed for a test?

At least 100–200 conversions per variant for reliable results. The smaller the expected difference, the larger the sample. To detect a 5% lift you need ~1500 conversions per variant.

Q: What if the result is not significant?

Either run the test longer and collect more data, or accept that the difference is too small to detect. Don't stop the test early — that increases false positives.

Q: What is test power?

The probability of detecting a real effect if it exists. Standard is 80%. With low power you may miss a real improvement. Increase sample size to raise power.

Q: Can I stop the test early if I see a difference?

No — that's "peeking" and leads to false conclusions. Set sample size in advance and wait. Or use sequential analysis methods that account for interim checks.

Q: What does p-value = 0.05 mean?

If the variants were equal, the probability of seeing this or a larger difference would be 5%. It is NOT the probability that B is better than A. p-value is about the data given H0, not about hypotheses given the data.

Q: What test does the calculator use?

Z-test for two proportions (two-sample). Suitable for comparing A and B conversion when each group has enough conversions. For small samples, Fisher's exact test is preferred.

Q: Where can I run an A/B survey test?

In SurveyNinja you can set up branching by random rule and show different question or creative variants to different respondents, then export data and check significance here or in analytics.

Is the conversion difference significant? Z-test for two proportions

A/B test significance — z-test for two proportions. Enter conversion counts and sample sizes — get p-value and whether the difference is significant.

Enter survey data

Variant A (control)

Visitors

Conversions

Variant B (test)

Visitors

Conversions

Significance level (α)

90% 95% 99%

Assumptions and limitations

Two-tailed z-test for two proportions
Samples assumed independent
n × p > 5 for both variants

Significance

—

Enter data

In SurveyNinja, NPS, CSAT and more are calculated automatically. Create a survey in 5 minutes and get real-time analytics.

Start free

A/B test examples

1 Conversion: variant A vs B

Variant A: 100 conversions from 1000 (10%). Variant B: 130 conversions from 1000 (13%).

Difference 3 pp. Calculator gives z-score and p-value. If p < 0.05 the difference is significant at α = 0.05 — variant B is better from a statistical standpoint.

2 CTA clicks

Variant A: 45 clicks from 500 (9%). Variant B: 52 clicks from 500 (10.4%).

Small 1.4 pp difference — the calculator shows whether the sample is enough for significance.

3 Sign-ups (different volumes)

Variant A: 80 from 2000 (4%). Variant B: 120 from 2000 (6%).

2 pp difference. With this n the two-sample z-test shows if the difference is significant at α = 0.05.

4 Cart abandonment

Control: 200 abandonments from 1000 (20%). New funnel: 150 from 1000 (15%).

5 pp drop in abandonment — the calculator helps check if the improvement is significant.

5 Newsletter sign-up

Variant A: 30 sign-ups from 600 (5%). Variant B: 42 from 600 (7%).

2 pp increase — with 600 visits per variant power may be low for a significant result.

6 Purchase conversion

Old landing: 25 purchases from 500 (5%). New landing: 35 from 500 (7%).

+2 pp — enter the data and check p-value before deciding.

What to avoid

Stopping the test early Wait for the planned sample size. Early stopping (peeking) distorts p-value and increases the risk of false positives.
Comparing many groups without multiple comparison correction The more pairwise comparisons, the higher the risk of false positives. Fix one primary metric and sample size before launch.
Confusing statistical and practical significance A difference can be significant (p < 0.05) but too small for business. Look at the difference in % and at minimum detectable effect.
Not planning sample size before launch Calculate minimum size with the minimum sample for comparison calculator. Otherwise the test may miss a real effect.

FAQ about A/B test significance

What is statistical significance?