The Quick Guide to Calculating Your Achievable Lift (MDE) with Your Audience

Traditional A/B test planning asks: "How many people do I need to prove a 5% lift?" But what if you have a fixed audience size or a limited budget? You need to flip the question and ask: **"With the audience I have, what is the smallest lift I can reliably prove?"**

This metric is the **Achievable Lift**, and calculating it is crucial for setting realistic Key Performance Indicators (KPIs) and avoiding perpetually inconclusive tests.

Why Achievable Lift Matters

If your campaign is only large enough to detect a 15% lift, but you go into the test expecting to prove a 5% lift, you are setting yourself up for failure. Even if your test is successful and generates an 8% lift, the test will come back as **Inconclusive** because your campaign was not statistically powered to measure an effect that small.

The Achievable Lift calculation allows you to manage executive expectations by telling them, "We can prove a lift of X% or higher, but anything below X% will remain statistically ambiguous."

Using the Calculator to Find Achievable Lift

Our tool allows you to reverse-engineer the power analysis specifically in the Holdout Test mode, which is ideal for this scenario as it isolates the true audience size from budget variables:

1. **Switch to Holdout Test:** Select this mode as it is audience-based, not budget-based.
2. **Enter Total Audience Size:** Input the actual number of users you can feasibly target during the test window.
3. **Check the Box:** Select the **"Calculate Achievable Lift (MDE)"** option.
4. **Run Calculation:** The tool reverses the statistical formula, using your fixed audience size to solve for the smallest possible MDE.

Interpreting Dual Metrics: Absolute vs. Relative

When the tool returns the Achievable Lift, it provides two numbers because both are essential for setting goals:

• **Achievable Absolute Lift:** (e.g., 0.15 percentage points). This is the statistical number you need to be concerned with.
• **Achievable Relative Lift:** (e.g., 15.0% lift). This is the KPI number you need to communicate to your team and leadership.

If the calculated Achievable Relative Lift is $20\%$, but your team only expects a $5\%$ lift, you know immediately that your test is severely underpowered and must be delayed or enlarged.

By calculating your Achievable Lift, you transform your testing process from hopeful guessing to scientific validation.