Minimum Detectable Effect MDE
The smallest change you want a test to detect reliably; it sets the sample size you need.
In Arabic أصغر أثر قابل للكشف
What is Minimum Detectable Effect?
The minimum detectable effect, or MDE, is the smallest difference between two versions that you want your test to detect with confidence. You set it before the test: if your purchase rate is 4% and you decide what matters is reaching at least 5%, the MDE is one percentage point, or 25% in relative terms.
The link between this number and sample size is strong and inverse: the smaller the effect you want to detect, the more visitors you need, roughly four times the sample when you halve the effect. That makes it the number connecting ambition to reality; it tells you whether your current traffic can answer your question in a reasonable time.
Why does Minimum Detectable Effect matter in marketing?
Before you start, it tells you whether your test is possible with your traffic or will drag on with no result.
- 01
Sets the sample size
From the target effect and your current traffic, you know how many visitors and how much time the test needs.
- 02
Stops pointless tests
If a test needs months to detect a small difference, you save the time and choose a bigger change.
- 03
Ties testing to business
You ask which difference is worth the change, so you do not celebrate a lift that does not cover the cost.
- 04
Not a forecast
It is the smallest effect you can detect, not the expected one. The real effect may be smaller and go undetected.
Worked example
- 1Situation
A store in Makkah wants to test a change to its checkout page and first asks which difference is worth detecting.
- 2Data
The current purchase rate is 4%, and the store believes raising it to at least 5% is what covers the cost of the change.
- 3Calculation
Effect = 5% − 4% = one percentage point, or 1 ÷ 4 = 25% in relative terms.
- 4Decision
The store compares its monthly traffic with what a sample size calculator requires. If it falls short, it raises the target effect or tests a bigger change.
Common mistakes
Picking it by wish, not traffic
Asking to detect a 5% relative lift in a small store means a test that runs for months. Check what your traffic allows first, then decide.
Mixing absolute and relative
One point on 4% is 25% relative. Entering 1% into a calculator expecting a relative value gives a completely wrong sample size. Check which type it asks for.
Changing it mid-test
Moving the target effect after seeing results so the test looks finished breaks its validity. Fix the number before starting and write it in the test plan.
Frequently asked questions
How do I calculate sample size from the MDE?
You need four inputs: current conversion rate, the minimum effect you want to detect, confidence level and statistical power, usually 80%. Enter them into a sample size calculator, such as Evan Miller's or one built into testing tools, to get the visitors needed per version.
What if my traffic is not enough?
You have three options: test a bigger change that should produce a clearer effect, test a metric higher in the funnel like add to cart instead of purchase, or pool traffic from several similar pages into one test.
Takteek services that help here
Before any test, Takteek calculates the smallest effect the client's real traffic can detect; if the test would run too long, we propose a bigger change.
Talk to Takteek
Want to put this to work for your business?
Tell us about your business in four steps and we will come back with a clear plan that starts from your current numbers. The consultation is free.
- A plan built on your numbers
- One team for ads, content and store
- Free consultation, no commitment


