A/B Test Sample Size Calculator
Compute visitors per variant from baseline conversion rate, relative or absolute MDE, significance, and statistical power. Optional test duration.
Loading tool...
What is A/B Test Sample Size Calculator?
The A/B Test Sample Size Calculator is a free online tool that estimates visitors per variant for a two-proportion z-test from baseline conversion rate, minimum detectable effect, significance, and power, entirely in your browser.
Key Benefits
- Plan traffic before launch instead of peeking at an underpowered test
- Compare relative and absolute MDE on the same baseline
- Turn total visitors into days when daily traffic is known
- Works offline with no upload of conversion rates
Common Use Cases
- β’Sizing a checkout-button experiment at a 5% baseline and 20% relative MDE
- β’Checking how many extra weeks a 10% MDE would add versus 20%
- β’Briefing stakeholders with visitors per variant and expected conversions
- β’Planning an A/B/C test by raising the variant count to 3
How to Use the A/B Test Sample Size Calculator
- Enter the baseline: Type the current conversion rate as a percent, for example 5 for a 5% checkout rate.
- Choose the MDE: Set the smallest lift you care about. Use relative percent (20 means 5% to 6%) or absolute percentage points.
- Set significance and power: Keep 95% significance and 80% power unless your team uses another standard. Add daily visitors to see days.
- Read the sample size: Use visitors per variant as the stop rule. Copy the summary into the experiment brief.
Key Features
- Two-proportion z-test sample size per variant
- Relative lift or absolute percentage-point MDE
- 90/95/99% significance and 80/90/95% power
- One-sided or two-sided hypothesis
- Duration from daily traffic and 2β10 experiment arms
Frequently Asked Questions
How is sample size calculated?
The tool uses the two-proportion z-test formula with an even split: n is ceil of [z-alpha times sqrt of 2 p-bar (1 minus p-bar) plus z-beta times sqrt of the two binomial variances], all squared, divided by (p2 minus p1) squared.
Relative or absolute MDE?
Relative 20% on a 5% baseline means you want to detect 6%. Absolute 1 percentage point on the same baseline also means 6%. Halving the MDE roughly quadruples the visitors you need.
Does this include peeking or sequential tests?
No. The n is for a fixed-horizon test you evaluate once at the end. Sequential or always-valid methods need a larger sample or a different design.
Are rates sent to a server?
No. The formula runs in JavaScript on your device. You can use it offline after the page loads.