A/B Test Sample Size & MDE Calculator
Two directions. Forward: given your baseline rate and target lift, how many visitors do you need per variant? Backward: given your available traffic, what's the smallest lift you can reliably detect?
Experimentation Intelligence
Underpowered tests waste quarters. Overpowered tests waste traffic.
The right sample size depends on your baseline conversion rate, the lift you need to see (MDE), your confidence level, and your statistical power target. Most teams either run too short (unreliable calls) or too long (traffic that could've tested the next hypothesis). This tool solves both directions.
Direction
Inputs
E.g. 10 = you want to detect a 10% relative lift (from 3.2% and 3.52%).
More variants require Bonferroni-corrected α to keep false positives in check.
Only used to estimate test duration.
Sample size needed
–
visitors per variant
–
Estimated duration
, days
What this means
Next steps
Once the test is running, check significance here
After you hit your sample size target, use the significance calculator for the full verdict including Bayesian probability.
Frequently asked questions
What does this tool calculate?
The sample size per variant needed to detect a given lift at a chosen significance and power. Also calculates MDE (smallest detectable lift) given traffic.
What is statistical power?
The probability the test will detect a real effect of the target size. 80% power is standard.
What is MDE?
Minimum Detectable Effect. The smallest lift the test can reliably spot. A test with 10% MDE won't spot a 3% real improvement.
Why do I need to plan sample size?
Running a test on too few visitors means you miss real effects. Running on too many wastes traffic that could run other tests.
What baseline conversion rate should I use?
The current conversion rate of the page or flow being tested. Not the industry average, your actual rate.
What significance level is standard?
95% confidence (p < 0.05). Some teams use 90% for lower-stakes tests, 99% for high-stakes.
How long should a test run?
Until sample size is reached, minimum one business cycle (usually a week) to capture seasonality. Not time-based alone.
Can I stop a test early if I hit significance?
Only with sequential-testing adjustments. Naive early stopping inflates false positives massively.
Why does variance affect sample size?
More variable metrics (revenue, time on page) need bigger samples to spot a given lift than less variable ones (conversion rate).
How to size tests for revenue instead of conversion?
Revenue needs a continuous-metric calculation (Welch's t-test). The tool supports this mode.
Does the calculator handle multi-arm tests?
Yes with a Bonferroni-adjusted significance level for each arm.
Free · no account
SEORCE for Chrome, Edge and Firefox
Audits the page you are on: headings, meta tags, schema, indexability, and whether AI crawlers can read it. Export the result as a PDF. Free, with no account.
It audits one page at a time and does not track AI answers. About the extension
Traffic from search and AI
See which visits came from search and AI.
SEORCE Web Analytics runs from one script tag, with Google Search Console connected. It shows visits from every AI source, from search and from campaigns, and it is free on every plan.
3,000+ users · Free forever plan, no card needed