MDE Calculator
Find out the smallest change your A/B test can detect given your traffic and conversion rate.
What Is Minimum Detectable Effect?
MDE is the smallest change in your metric that a test can reliably detect. If your MDE is 2% but you're expecting a 0.5% lift, your test doesn't have enough power to see the difference — you'll get an inconclusive result. More traffic and higher baseline rates let you detect smaller effects.
Reading the chart
Each bar shows the smallest effect you can detect at that test duration. Green means detectable, red means you need more time or traffic.
Confidence level
Controls your false positive rate. At 95% confidence, there's a 5% chance of declaring a winner when there's no real difference.
Statistical power
Your ability to detect a real effect when one exists. At 80% power, you have a 20% chance of missing a real improvement.
How MDE Is Calculated
MDE is derived from standard statistical power analysis. The formula connects your sample size, baseline conversion rate, and chosen confidence/power levels to find the smallest effect you can reliably detect.
MDE = (Zα + Zβ) × √(p(1−p) × (1/ncontrol + 1/nvariant))
Zα — Significance threshold
Set by your confidence level. At 95% confidence (two-tailed), Zα = 1.96. Higher confidence means a larger Z value, which increases MDE — you need a bigger effect to be “sure” it's real.
Zβ — Statistical power
Set by your power level. At 80% power, Zβ = 0.84. Higher power (90%) increases Zβ to 1.28, making MDE larger — you need more data to catch smaller effects.
p — Baseline conversion rate
Your current conversion rate before the test. Higher baselines produce more variance, which paradoxically makes smaller relative effects harder to detect. A 50% rate has maximum variance.
n — Sample size per group
The number of visitors in each group. This is the biggest lever — doubling your sample size reduces MDE by ~30%. More traffic means you can detect smaller effects.
What Is a Good MDE?
“Good” depends on what you're testing and how big an impact you expect. Here are practical benchmarks based on common subscription business tests:
Excellent
You can detect subtle changes. Typical for high-traffic landing pages, sign-up flows, and mature optimization programs.
Workable
Good enough for most tests. You'll catch meaningful changes but miss micro-optimizations. Typical for mid-funnel tests like trial-to-paid.
Only big bets
You can only detect dramatic changes. Fine for bold redesigns or pricing overhauls, but don't bother running iterative tests at this MDE.
How to Reduce Your MDE
Run tests longer
Each additional week of data reduces MDE. Doubling test duration cuts MDE by ~30%. The calculator above shows exactly how MDE shrinks over time.
Increase traffic to the test
More visitors per week means more statistical power. Redirect more traffic to the page being tested, or combine traffic sources.
Use a one-tailed test
If you only care about improvement (not detecting harm), switch to one-tailed. This reduces MDE by ~20% but means you won't catch negative effects.
Lower your confidence level
Dropping from 99% to 95% or 95% to 90% reduces MDE. Acceptable for early-stage exploration where speed matters more than certainty.
MDE vs Sample Size
MDE and sample size are two sides of the same coin. Most A/B test calculators ask “how many visitors do I need?” — but that assumes you already know the effect size you're targeting. The MDE approach flips the question: “given my actual traffic, what can I detect?”
Sample size calculators
You input your desired effect size and it tells you how many visitors you need. Problem: most teams pick an optimistic effect size, get a small sample requirement, run a short test, and declare “no significant result.” They never had the power to detect a realistic effect.
MDE calculators (this one)
You input your actual traffic and it shows the smallest effect you can detect at each duration. This grounds your test design in reality — you see whether the test is worth running before you start it.
MDE for Subscription Businesses
Different stages of the funnel have very different traffic volumes and baseline rates. Use the metric presets above to match your test.
High-traffic metrics
Sign-up rate, landing page conversion — typically enough volume to detect 1–5% relative changes within 2–4 weeks.
Mid-traffic metrics
Free-to-paid, trial conversion, checkout completion — may need 4–8 weeks to detect meaningful changes.
Low-traffic metrics
Cancellation rate, payment recovery — often too low-volume for traditional A/B testing. Consider before/after analysis instead.
Common mistake
Running a test for a fixed time then calling it. If your MDE is larger than the expected effect, the test was doomed from the start.
Frequently Asked Questions
What is Minimum Detectable Effect (MDE)?
Minimum Detectable Effect is the smallest change in a metric that an A/B test can reliably detect at a given confidence level. If your MDE is 2% but you're expecting a 0.5% lift, your test doesn't have enough statistical power to see the difference.
How do I calculate MDE for my A/B test?
MDE depends on three factors: your sample size (traffic), your baseline conversion rate, and your chosen statistical power and significance level. Enter your weekly visitors and baseline conversion rate into the calculator above, and it will show you the smallest effect you can detect at each test duration.
What sample size do I need for an A/B test?
The sample size you need depends on your baseline conversion rate, the size of the effect you want to detect, and your chosen confidence level and statistical power. Generally, detecting smaller effects requires larger sample sizes. Use this calculator to see what's detectable at your current traffic levels.
What confidence level should I use for A/B testing?
The standard confidence level for A/B testing is 95%, which means a 5% false positive rate. For high-stakes tests (pricing changes, major redesigns), consider 99%. For early-stage exploration, 90% may be acceptable to detect effects faster.
What is a good MDE for an A/B test?
For most subscription businesses, an MDE under 10% relative is workable. Under 5% is excellent — you can detect subtle optimizations. Above 15%, only dramatic changes will show as significant. The right target depends on whether you're running bold experiments or iterative optimization.
How do I reduce my MDE?
Four levers: run the test longer (each additional week helps), increase traffic to the test, switch from two-tailed to one-tailed (if you only care about improvement), or lower your confidence level from 99% to 95%. Increasing traffic has the biggest impact.
What is the difference between MDE and sample size?
They're inverse views of the same relationship. A sample size calculator asks “how many visitors do I need to detect a 5% change?” An MDE calculator asks “given my 10,000 weekly visitors, what's the smallest change I can detect?” MDE is more practical because it starts with your actual traffic rather than an assumed effect size.
Can I use MDE for revenue or retention metrics?
Yes, but with caveats. Revenue metrics have higher variance than conversion rates, so you typically need larger samples. For retention, the time horizon matters — you may need months of data before you can even measure the outcome. Use the metric presets in the calculator above to model different funnel stages.
Get the playbook every Thursday
Weekly strategies for subscription businesses. Real companies, real numbers, tactics you can steal. From the operator who grew Codecademy from $10M to $50M ARR.