MDE Explained: How to Calculate the Minimum Detectable Effect for Marketing

Stop wasting budget. Learn the one number that guarantees a conclusive A/B test result.

Authored by Lalit Jain · lalit.7.jain@gmail.com · LinkedIn

The **Minimum Detectable Effect (MDE)** is the most critical metric in any A/B testing framework, yet it’s often overlooked. Simply put, the MDE is the smallest percentage change (lift or drop) in a key metric—like your conversion rate—that you design your experiment to reliably detect. If the true effect of your test is smaller than your MDE, your test is destined to fail, leading to inconclusive results that waste both time and budget.

Why MDE is the Foundation of Statistical Power

MDE directly dictates your required sample size. If you want to detect a tiny change, you need a massive amount of data to prove that the result isn't random noise. Conversely, if you only care about large, double-digit changes, you can run a much shorter test. The decision on your MDE is a **business decision** first and a statistical one second.

Statisticians refer to this relationship as **Statistical Power** (often set at 80% or 90%). A test with 80% power means that if a lift as big as your MDE truly exists, you have an 80% chance of correctly identifying it. If your test is *underpowered* (meaning your budget/duration is too small for your MDE), you might miss a successful change, suffering a costly false negative.

Statistical chart illustrating the Minimum Detectable Effect (MDE) as the distance between two conversion rate distributions.
The distance between the peaks (MDE) determines how easily you can detect a difference.

Calculating MDE: The Four Key Drivers

Calculating the MDE involves balancing four variables:

  1. Baseline Conversion Rate (CVR): The historical average CVR for your control group. Lower baselines require larger samples because the data is "noisier."
  2. Confidence Level (Alpha): Typically 95%. This minimizes false positives (seeing a lift that isn't real).
  3. Statistical Power (Beta): Typically 80%. This minimizes false negatives (missing a lift that *is* real).
  4. Sample Size: The number of conversions (or audience members) required.

Marketers often struggle with the complex formulas required to solve for the necessary sample size. This is why tools are essential. If you know your baseline CVR and the MDE you *must* prove, you can instantly find the corresponding sample size. The required sample size is the output that dictates your necessary budget and time.

Absolute vs. Relative Lift

When thinking about MDE, it's vital to clarify whether you are using an **Absolute** or **Relative** measure:

  • **Absolute Lift:** The raw percentage point difference. (e.g., a **5.0%** CVR to a **5.5%** CVR is a **0.5 percentage point** absolute lift).
  • **Relative Lift:** The percentage change relative to the baseline. (e.g., a **5.0%** CVR to a **5.5%** CVR is a **10%** relative lift, calculated as 0.5 / 5.0).

Always align your MDE with your marketing goals. If a 10% relative lift is the minimum change that justifies rolling out a new product, then that should be your MDE target for the test.

Don't Calculate MDE By Hand!

Instead of manually working through complex formulas to determine your required sample size, use our free, built-in tool. Just input your Baseline CVR and your desired MDE, and our **Statistical Significance Calculator** will tell you exactly how many conversions you need for a conclusive result.

Optimizing Your MDE for Limited Budget

If you plug in your numbers and the calculator tells you that you need 20,000 conversions but your budget can only deliver 5,000, your test is **underpowered**. You have two options:

  1. Increase your budget or test duration significantly.
  2. Increase your MDE. Adjust your expectations to only measure a larger, more impactful change (e.g., target a $15\%$ relative lift instead of $5\%$).

Understanding the **Minimum Detectable Effect** is the single most important skill for any growth marketer. It is the gatekeeper to running tests that actually matter.

Recent Updates & Change Log

[Sept, 2025]

  • Added ability to calculate **Achievable Lift (MDE)** in both Absolute and Relative terms for Holdout Tests.
  • Implemented **Holdout Test Mode** to determine required audience size for incremental lift.
  • Added **Dynamic Split Slider** for A/B test budget distribution.
  • Implemented a clear **Calculation Breakdown** and actionable **Recommendations**.

This tool is actively maintained and improved.