A/B Test Sample Size Calculator

Find how many visitors each variant needs, and how many days that takes at your own traffic, before you launch a test.

Your numbers

Baseline and effect

Current (baseline) conversion rate

Your control's conversion rate over the last few weeks.

Minimum lift worth detecting

Relative lift to detect

20% means detecting a move from 4% to 4.8%.

Test parameters

Confidence level

Statistical power

Power is your chance of detecting a real effect if one exists. 80% is the common default.

Variants, including control

Traffic

Daily visitors to the test, all variants combined

visitors/day

Your result · updates as you type

Visitors needed per variant

10,314

20,628 visitors in total across 2 variants, to reliably tell 4% apart from 4.8% at 95% confidence and 80% power.

Total visitors needed: 20,628
Days to collect that: 26
Weeks: 3.7

Halving the lift you want to detect roughly quadruples the sample size needed. Chasing tiny lifts on low traffic rarely finishes.

Visitors per variant needed for other lifts

0.5× lift

39,473

1× lift

10,314

1.5× lift

4,780

2× lift

2,799

Breakdown

Baseline rate: 4%
Target rate to detect: 4.8%
Confidence / power: 95% / 80%
Visitors per variant: 10,314
Total (2 variants): 20,628
At your daily traffic: 26 days

How the sample size is worked out

The calculator uses the standard two-proportion sample size formula, based on your baseline rate, the smallest lift worth detecting (the minimum detectable effect, or MDE), your confidence level and your statistical power (the chance of detecting a real effect if one exists).

Visitors per variant = (z_confidence + z_power)² × (p₁(1 − p₁) + p₂(1 − p₂)) ÷ (p₂ − p₁)²

Example: the default numbers

Baseline conversion rate: 4%
Lift to detect: 20% relative (4% → 4.8%)
Confidence / power: 95% / 80%
Daily visitors, both variants: 800

You need 10,314 visitors per variant, 20,628 in total. At 800 visitors a day, that's about 26 days.

Using the result

Halving the lift you want to detect roughly quadruples the sample size. Chasing a 2% lift on a low-traffic store often means a test that never finishes: consider testing a bolder change instead.
More variants split the same traffic further. A 3-way test needs the same visitors per variant, but 50% more traffic in total, and takes proportionally longer.
Power of 80% is the common default. Raise it to 90% for a decision you really don't want to get wrong, at the cost of a bigger sample.

Questions people ask

How many visitors do I need for an A/B test?

It depends on your baseline conversion rate and the smallest lift you actually care about detecting. A 4% baseline store testing for a 20% relative lift needs a bit over 10,000 visitors per variant at typical settings.

What's statistical power?

It's your chance of detecting a real effect, if one truly exists, given your sample size. 80% power means you'd catch a real effect of that size 4 times out of 5; the other time, a real winner would look inconclusive.

What if I don't have enough traffic to reach this sample size?

Either test a bigger change (bigger lifts need smaller samples), lower your power to 80% if you're at 90%, or accept a longer test and plan around it rather than stopping early.

Does this account for multiple variants?

Yes. Set the variant count and the total visitors needed scales accordingly, though visitors needed per individual variant stays the same.

Last reviewed 30 September 2026.