Free tool

How many visitors does your A/B test need?

Enter your current conversion rate and the smallest lift worth detecting. You get the visitors needed for each version and how many weeks the test will take.

The share of visitors who convert today. Use the last 4 to 8 weeks of data for the page you’re testing.%
The smallest change you’d want the test to catch. Smaller lifts need far more visitors.
%

A relative lift of 10% takes 3.00% to 3.30%.

(optional)Unique visitors who reach the page or flow you are testing, not whole-site traffic.
Confidence levelHow sure you want to be before calling a winner.
Statistical powerThe chance the test spots a real lift of at least that size. 80% is the usual choice.
Test typeTwo-sided can detect a variant that is worse. One-sided only checks whether it is better, so it needs fewer visitors.
Versions including controlEach extra variant needs its own sample, and the significance threshold is split between them.
Advanced
Use less than 100% if only part of your traffic sees the test.%

53,211 visitors per versionVisitors each version needs before the test can reliably detect the lift. Stopping early on a good result makes a false winner more likely.

Long for an A/B test. A bigger change or a busier page would shorten it.

Total visitorsPer-version number times the number of versions.106,422across 2 versions
Weeks to runTotal divided by daily visitors in the test, rounded up to whole weeks so every weekday is counted equally.1171 days
Rate to detectThe control rate and the variant rate the test is built to catch.3.00% → 3.30%
Get started →Check a finished test

The calculation runs in your browser. Nothing you enter is sent or stored.

An A/B test sample size calculator tells you how many visitors each version needs before a test can reliably detect a real difference. It depends on your current conversion rate, the smallest lift you care about, and how sure you want to be. This one also turns the total into weeks of traffic, so you know before you launch.

Reading it

How to read your result.

Visitors per version: the number each version needs. Stopping well before it makes a “winner” much less trustworthy.

Weeks to run: total visitors divided by the visitors you send to the test each day, rounded up to whole weeks. Whole weeks matter because weekday and weekend visitors behave differently.

Smallest lift worth detecting: a smaller lift needs a lot more traffic. Halving the lift roughly quadruples the visitors.

Relative or absolute: a 10% relative lift on a 3% baseline is 0.3 percentage points (3.0% to 3.3%). A 10-point absolute lift would be 3% to 13%. Check which one you’ve set.

Minimum detectable effect: what it is and how to set it
One test, worked through
+10% relative lift3.00% → 3.30%53,211 visitors per version · 11 weeksLong test
+20% relative lift3.00% → 3.60%13,914 visitors per version · 3 weeksRealistic

Your page converts at 3% and you’d like to catch a 10% relative lift (3.0% to 3.3%) at 95% confidence and 80% power. Each version needs about 53,200 visitors, so a two-version test needs about 106,400. At 1,500 visitors a day that is 71 days, or 11 weeks. Raise the target to a 20% lift (3.0% to 3.6%) and it falls to about 13,900 per version, or 3 weeks.

How it works

How this calculator works.

It uses the standard sample size formula for comparing two conversion rates, and runs entirely in your browser.

Step 1

Rates

Your baseline rate is the control rate. The variant rate is the baseline plus the lift you entered.

Step 2

Visitors per version

The visitors per version come from a two-proportion z-test with the confidence level and power you chose. A higher confidence level or higher power needs more visitors.

Step 3

Test type

Two-sided is the default. One-sided needs fewer visitors but cannot detect a variant that does worse.

Step 4

More than two versions

With three or four versions, the significance threshold is divided by the number of variants being compared with control (Bonferroni correction), so each version needs more visitors.

Step 5

Total and weeks

Total visitors is the per-version number times the number of versions. Weeks is the total divided by the daily visitors in the test, rounded up to whole weeks.

Limits: This is a fixed-horizon plan for a yes/no outcome such as a purchase, sign-up or click. It does not apply to averages such as revenue per visitor. Tests analysed with Bayesian or sequential methods set their stopping rules differently, so treat the number as a planning guide. What is sequential testing?

Check the finished test with the significance calculator
Low traffic

What to do when the test would take too long.

A small lift on a 3% page can need over 100,000 visitors. These changes bring the number down without lowering the standard:

Test a bigger change

A new offer or layout can move a rate more than a button colour.

Test on a busier page

Or a step further up the funnel, where more visitors reach the goal.

Pick a goal with a higher rate

Such as “added to cart” rather than “purchased”.

Cut the number of versions

Two versions need far fewer visitors than four.

Skip the A/B test

If nothing works, compare before and after.

A/B testing with low trafficHow long to run an A/B testBefore-and-after analysis
Web Experiments

Get the plan before every launch.

Cuttl’s Web Experiments plans each test for you, so there’s no calculator step to remember.

  • A pre-test planner that estimates how many weeks a test needs
  • The minimum detectable effect is set before launch and recorded
  • A warning when a test would take too long, with a before-and-after check suggested instead
  • Peeking warnings, false-positive risk and sample ratio mismatch checks once it’s live
See the test planner in Web Experiments
Questions

Sample size calculator questions

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

It depends on your conversion rate and the smallest lift you want to detect. At a 3% conversion rate, detecting a 10% relative lift takes about 53,200 visitors per version at 95% confidence and 80% power.

What is the minimum detectable effect?

It’s the smallest change the test is built to catch. A smaller minimum detectable effect needs many more visitors, so pick the smallest lift that would be worth acting on.

What confidence level and power should I use?

95% confidence and 80% power are the usual defaults. Raising either one makes the test more reliable and needs more visitors.

Should I enter a relative or an absolute lift?

Relative is how most teams talk about lift (“10% more conversions”). Absolute is in percentage points. The calculator shows both so you can check you’ve entered what you meant.

How long should I run my A/B test?

Until it has reached the planned number of visitors, in whole weeks. This calculator divides the total by your daily visitors and rounds up.

Why does the calculator say my test would take so long?

Small lifts on low-traffic pages need a lot of visitors. Test a bigger change, a busier page or a higher-rate goal, or compare before and after.

Does this work for revenue or average order value?

No. It is built for yes/no outcomes like purchases and sign-ups. Averages need a different calculation.

Plan every test before it starts.

Cuttl estimates how many weeks a test needs, flags tests that would take too long, and checks the traffic split once it’s live, alongside heatmaps and recordings for each variant.

Get startedTry the live demo