Skip to main content
Meta Conversion Lift Experiment

How Conversion Lift studies work, how to create an experiment, and how to interpret the incrementality results

K
Written by Kevin Wolf
Updated yesterday

The Basic Mechanism

Conversion Lift uses a randomized controlled trial methodology to measure the true impact of your advertising. Here's the process:

  1. When you start a test, Meta will automatically divide your target audience into two groups:

    • Test Group: People who can see your ads

    • Control Group: People who won't see your ads (also called the holdout group)

  2. The groups are created using randomization to ensure they are statistically similar in terms of:

    • Demographics

    • Geographic location

    • Past behaviors

    • Other relevant characteristics

  3. Throughout the test period, Meta tracks conversions using your selected Meta Pixel.

  4. The difference in conversion rates between these groups represents the "lift" - the true incremental impact of your advertising. This shows you which conversions wouldn't have happened without exposure to your ads.

Creating a Conversion Lift Test

Prerequisites

You must have:

  • An active Meta ad account

  • A campaign with $5,000 USD minimum spend in the past year

  • At least 500 optimized conversions

  • Proper Pixel event tracking setup

Step-by-Step Setup Process

1. Navigate to the Meta Attribution Page

2. Select the Experiment Column(s)

Navigate to Select Columns and either:

  • Select Experiments preset columns

  • Or find Experiment inside Customize Columns

3. Configure your Conversion Lift Test

  • Select the campaign to test

  • Name the Ad Study

  • Select the Pixel that aligns with the primary conversion event you want to measure

  • Select your preferred holdout group and test duration:

    • Holdout group: The percentage of audience that won't see your campaign

    • Test duration: Time period to run the test with reduced campaign reach

    • Choose either:

      • Optimize for Speed: Faster results, with potential for higher negative impact on revenue

      • Balanced Approach: Recommended for new users

      • Minimize Revenue Impact: Slower results, with least impact on revenue, but results might be less statistically significant

      • Manual Configuration: Customize your holdout and duration

4. Launch & Monitor

  • Review your configuration settings

  • Click "Run Test" to begin

  • Monitor results in the Experiments columns

  • Wait for the test to gather sufficient data for statistical significance. It is recommend to let the test finish running, but you can stop the test if you are satisfied with the results.

Did this answer your question?