Before every A/B test
Check that the test can reach significance within a realistic timeframe.
← All free tools
Free tool · A/B Test
MDE Calculator | sample size
Estimate how many visitors you need to detect a given relative difference (Minimum Detectable Effect) between control and variant. Two-sided z-test on two proportions.
Current control rate. E.g. 4% = 4.0.
Smallest relative difference you want to detect. 10% = going from 4% to 4.4%.
Sum of both groups. Used to estimate duration.
Recommended: alpha 5%, power 80%. Industry standard for B2B A/B tests.
Result
Per variant
39,475visitors
Total (2 groups)
78,950visitors
Absolute difference detected
0.400%
Estimated duration
79days
Formula: per-variant sample size for a two-sided z-test on two proportions, with power 1−β and significance α. The normal approximation is valid when np ≥ 5 and n(1−p) ≥ 5.
How it works
4 parameters, one reliable estimate
The calculator applies the classic sample size formula for two independent proportions, used by every serious statistics tool (Optimizely, AB Tasty, Evan Miller, etc.).
01
Enter the baseline
The control’s current conversion rate (e.g. 4%).
02
Pick the relative MDE
The smallest relative difference you want to detect (e.g. 10%).
03
Set alpha & power
Standard: alpha 5% (95% confidence), power 80%.
04
Estimate the duration
Enter daily traffic to get the number of test days.
Formula
n = (Zα/2 · √(2 · p̄ · (1−p̄)) + Zβ · √(p₁(1−p₁) + p₂(1−p₂)))² / (p₂ − p₁)²
Where p₁ = baseline, p₂ = baseline · (1 + relative MDE), p̄ = (p₁ + p₂) / 2. Zα/2 and Zβ come from the standard normal distribution.
When to use it
4 key moments to reach for the calculator
CRO roadmap
Prioritise feasible tests vs. tests that would need months of traffic.
Product team brief
Align Product, Growth and Data on duration and uplift expectations.
Leadership pitch
Justify that a given uplift requires a given volume — or the other way around.
FAQ
Frequently asked questions
What is the MDE (Minimum Detectable Effect)?
The MDE is the smallest difference between your control and your variant that your test can statistically detect, given your baseline, alpha and power. The larger your sample, the smaller the detectable MDE.
Absolute or relative MDE: what is the difference?
Relative MDE is expressed as a percentage of the baseline (e.g. +10% on a 4% baseline = going to 4.4%). Absolute MDE is expressed in points (e.g. +0.4pt = going from 4% to 4.4%). Our calculator uses relative MDE — more intuitive for ROI planning.
What statistical power should I choose?
80% is the standard for most growth A/B tests. It means that if the effect truly exists, you have an 80% chance of detecting it. For critical decisions (redesign, pricing), prefer 90%.
What confidence level (alpha) should I choose?
Alpha 5% (= 95% confidence) is the standard. It is the accepted risk of concluding there is an effect when there is none. For iterative growth tests with low stakes, alpha 10% can be acceptable.
The estimated duration is very long — what can I do?
Three levers: (1) allocate more traffic to the test (fewer variants), (2) target a larger MDE (bolder effects), (3) accept a looser power or alpha. If none of that is possible, the test is probably not feasible and the hypothesis needs reformulating.
Is the calculation valid for multivariate tests?
This tool assumes 2 groups (A vs B). For 3 or more variants, split the available traffic across variants and apply a correction (Bonferroni: alpha / number of comparisons).
Need a senior opinion on your CRO roadmap?
We audit your A/B stack, your tracking and your roadmap. 5 days, €0 for B2B scale-ups.