A/B Testing Sample Size (Monthly) Calculator
Estimate the visitors needed for an A/B test and how many months of traffic you may need to detect a meaningful conversion-rate change.
Overview
Use this A/B testing sample size calculator to estimate how many visitors each version needs and how long the experiment may take based on your baseline conversion rate, target uplift, confidence setting, statistical power, and monthly traffic.
How it works
The calculator converts your baseline rate and desired relative uplift into expected control and variant conversion rates. It then applies a standard two-proportion sample-size approximation. Higher confidence and power require more visitors. Smaller conversion-rate differences also require larger samples. Finally, the combined sample requirement is divided by the number of visitors expected to enter the test each month.
How to use this calculator
- 1Enter the current conversion rate for your control version.
- 2Choose the smallest relative uplift that would be meaningful for your decision.
- 3Enter z-scores for your preferred confidence level and statistical power.
- 4Add your average monthly visitor volume and the share eligible for the test.
- 5Review the required sample per variant and estimated duration in months.
Example Calculation
Baseline conversion rate
3%
Minimum detectable uplift
10%
Confidence level z-score
1.96
Statistical power z-score
0.84
Monthly visitors
100000
Traffic included in test
100%
Visitors needed per variant
53,151 visitors
With a 3% baseline rate, a target uplift of 10%, 95% confidence, 80% power, and 100,000 eligible monthly visitors, the test needs roughly 53,000 visitors per variant, or about 1.1 months of traffic.
Frequently asked questions
What is sample size in A/B testing?
Sample size is the number of visitors or users each test version needs before you can reliably compare their conversion rates at the selected confidence and power settings.
What is a minimum detectable uplift?
It is the smallest relative improvement you want the test to detect. For example, a 10% uplift changes a 3% conversion rate to 3.3%.
Why does a smaller uplift need more traffic?
Small differences are harder to distinguish from normal random variation, so more observations are needed to detect them with the same confidence and power.
Which z-score should I use for 95% confidence?
For a standard two-sided 95% confidence level, use 1.96. For a two-sided 99% confidence level, use 2.576.
Which z-score should I use for 80% power?
Use 0.84 for 80% power. If you want 90% power, use 1.282, which will increase the required sample size.
Does this calculator support more than two variants?
No. This estimate is designed for one control and one variant with equal traffic allocation. Tests with multiple variants generally need more total traffic and additional statistical adjustments.
Why might my actual A/B test take longer than estimated?
It may take longer if eligible traffic falls, conversion rates change, visitor allocation is uneven, tracking has gaps, or you need to account for business cycles and experiment quality checks.
Explore Related Calculators
Assumptions and warnings
Assumptions
- The test uses two variants with an even 50/50 traffic split.
- The calculation uses a two-sided comparison of two conversion rates.
- Visitor behavior, traffic quality, and the baseline conversion rate are assumed to remain reasonably stable during the test.
- The estimated duration is based on average eligible monthly visitors and does not account for implementation delays or data-quality checks.
- Results are planning estimates rather than a guarantee of a valid experiment outcome.
Warnings
- This calculator provides a statistical planning estimate only and is not a substitute for a full experiment design review.
- Avoid ending a test early solely because an interim result looks favorable; repeated checking can increase false-positive risk.
- Very low baseline conversion rates or very small target uplifts can require substantially more traffic than this simplified estimate suggests.