Choose the sample size
your test can defend.
I no longer maintain a separate calculator on Atticusli. Use the canonical GrowthLayer planner for sample size, duration, significance, SRM checks, and an analyst-ready handoff.
Free to use · no duplicate formulas · built for real experiment decisions
One planning workflow, from question to experiment.
The calculator is useful only when its assumptions survive into the test brief. GrowthLayer keeps the planning result connected to what happens before and after launch.
Set the smallest useful effect
Start with the lift that would justify the engineering, design, and maintenance cost—not a convenient number that makes the test look fast.
Check traffic and duration
Compare the required sample with eligible daily traffic, allocation, exposure, and a full business cycle before you commit the test.
Carry the decision forward
Share the assumptions, export the analyst summary, and move the test into a repository so the result is not stranded in a browser tab.
Four choices determine the answer.
Sample size is not a property of a page or a traffic number alone. It is the result of a decision about what matters, how certain you need to be, and how the test will run.
This page explains the decision. GrowthLayer runs the workflow.
Maintaining two calculators would split search ownership, analytics, and trust. GrowthLayer is the single source for the statistical engine and the next operational step: connect the result to an experiment record your team can revisit.
Use the unified A/B test plannerCommon sample-size questions
How do I calculate sample size for an A/B test?
Start with your baseline conversion rate, the smallest effect worth detecting (your MDE), your significance level, and your desired power. GrowthLayer’s planner turns those choices into a sample target and estimated duration, then keeps the planning assumptions attached to the experiment.
What is minimum detectable effect (MDE) in A/B testing?
MDE is the smallest change that would justify acting on the result. If your baseline conversion rate is 5% and your relative MDE is 10%, you are planning to detect a move to 5.5%—an absolute difference of 0.5 percentage points. Smaller MDEs require more traffic.
How long should I run my A/B test?
Estimate duration by dividing the total required sample by the traffic that can enter the experiment each day. Also run through at least one complete business cycle, usually seven days or more, and do not stop early just because a p-value crosses a threshold.
Why does this page link to GrowthLayer instead of running a calculator here?
A single maintained calculator is safer than two copies of statistical logic. GrowthLayer owns the current planning, duration, significance, and sample-ratio-mismatch tools, plus the analyst handoff into an experiment repository. This page stays focused on explaining the decision and connecting you to that workflow.
Ready to size the test?
Put your assumptions into the GrowthLayer planner, then carry the result into the experiment record your team will use.
Open GrowthLayer’s A/B test plannerMake Growth Decisions
You Can Defend
Know what to test, when to trust the result, and what to do next. Practical decision guides for analysts, growth teams, and founders.
Opens Substack to confirm · Free · Unsubscribe anytime
Read Better Decisions
Practical guides on experimentation, analytics, commercial judgment, and using AI without outsourcing the decision.
Open Substack (opens in new tab)