How it works
p1 = baseline; p2 = p1 × (1 + MDE). n = ceiling([zα√(2p̄(1−p̄)) + zβ√(p1(1−p1)+p2(1−p2))]² ÷ (p2−p1)²).
Worked example
A 5% baseline, 20% relative MDE, 95% confidence and 80% power plan about 8.2k observations per variant.
Assumptions and limits
Estimate without continuity correction. Requires independent observations, two variants, and equal allocation; not for sequential tests or multiple comparisons.
Method reference: NIST: difference of proportions
See it in real data
ASO
Keywords and search rank in the app stores
Frequently asked questions
How do I use the a/b test sample size calculator?
Plan a per-variant sample to compare two conversion rates with equal allocation, a two-sided test, and a normal approximation.
Does a 20% MDE mean 20 percentage points?
No. With a 5% baseline, 20% relative lift means reaching 6%, a difference of 1 percentage point.
Do I need an account and are my inputs uploaded?
No account is required. Calculations run in your browser. Entered numbers and text are not included in tool analytics events; only the tool ID, language, and action are recorded.