How to Conduct an Incrementality Holdout Study with Google Ads’ Performance Max Campaigns

Introduction

Performance Max (PMax) campaigns in Google Ads present unique challenges for conducting incrementality holdout studies. PMax automatically optimizes across multiple placements (Search, Shopping, Display, YouTube, and Discover), making it difficult to isolate the impact of specific ad types. The biggest issue is that PMax tends to over-index on branded terms, as branded searches are low-hanging fruit for conversions. If advertisers attempt to run non-branded holdout tests without proper controls, their results can become invalid due to PMax absorbing traffic in unexpected ways.

In this guide, we’ll break down how to properly structure an incrementality holdout test with PMax campaigns, avoiding common pitfalls and ensuring clean, actionable results.

‍

Why PMax Complicates Incrementality Testing

1. PMax Absorbs Non-Branded Traffic When Other Campaigns Are Paused

Many advertisers separate their PMax campaigns into:

• Branded PMax Campaign (which includes branded and some non-branded traffic)
• Non-Branded PMax Campaign (explicitly targeting non-branded queries)

If you attempt to test the incrementality of non-branded traffic by pausing the Non-Branded PMax Campaign in certain locations, the Branded PMax Campaign will start absorbing non-branded queries in those areas. This contamination skews results, making it seem like branded traffic is driving more conversions than it actually is.

2. PMax Takes Over When Individual Channels Are Paused

When running holdout tests for specific ad types (Display, Video, Shopping, or Search), PMax dynamically reallocates budget to those channels in response to pauses, nullifying the test.

For example:

• If you pause Display campaigns in test regions but keep PMax running, PMax will start allocating more budget to Display placements.
• If you pause Search campaigns, PMax will fill the gap with Search placements.

This prevents a true measurement of incremental impact because PMax compensates for paused channels.

3. The Right Way to Structure a PMax Incrementality Test

To ensure a clean and reliable incrementality holdout study, advertisers must structure their experiments to fully isolate the channel or audience being tested. This requires turning off all PMax campaigns in test regions and launching controlled interim campaigns.Step-by-Step Process for Structuring a PMax Holdout Test

‍

1. Testing Display Incrementality

Goal: Measure the impact of Display campaigns while preventing PMax from interfering.

✅ Correct Setup:

• Pause ALL Display campaigns in test regions.
• Pause ALL PMax campaigns in test regions.
• Launch a new PMax campaign without Display assets (only Text and Shopping assets).
• Optionally, create a new Display campaign and run it in control regions for additional insights.

❌ What to Avoid:

• Leaving PMax active, as it will allocate spend to Display placements, invalidating the test.

2. Testing Search Incrementality

Goal: Determine the true impact of Search campaigns.

✅ Correct Setup:

• Pause ALL Search campaigns in test regions.
• Pause ALL PMax campaigns in test regions.
• Launch a new PMax campaign without Text assets (only Shopping assets).
• Optionally, create a separate Search campaign for control regions.

❌ What to Avoid:

• Keeping PMax active, as it will reallocate budget to Search placements.

3. Testing Shopping Incrementality

Goal: Measure the effectiveness of Shopping campaigns separately from PMax.

✅ Correct Setup:

• Pause ALL Shopping campaigns in test regions.
• Pause ALL PMax campaigns in test regions.
• Launch a new performance max campaign without Shopping assets.
• Optionally, create a Shopping campaign for control regions.

❌ What to Avoid:

• Allowing PMax to continue running, as it will shift budget to Shopping.

4. Testing Video Incrementality

Goal: Evaluate the impact of YouTube campaigns independently.

✅ Correct Setup:

• Pause ALL Video campaigns in test regions.
• Pause ALL PMax campaigns in test regions.
• Launch a new performance max campaign without Video assets.
• Optionally, create a separate Video campaign for control regions.

❌ What to Avoid:

• Keeping PMax active, as it will distribute spend to YouTube placements.
  ‍

Using Mike Rhodes' PMax Script for Deeper Insights

‍

Google does not provide granular insights into PMax’s spend allocation across channels. However, Mike Rhodes has developed a PMax script that helps advertisers understand where their budget is actually going.

How This Script Helps:

• Breaks down PMax spend by Search, Shopping, Display, and Video.
• Reveals how much budget is going to branded vs. non-branded queries.
• Provides channel-specific visibility for better decision-making before and after a test.

🔗 Download the script here: Mike Rhodes’ PMax Script
‍

‍Conclusion: Conducting a Clean Incrementality Test with PMax

Key Takeaways:

1. Turn off ALL PMax campaigns in test regions to prevent cross-channel budget reallocation.
2. Create controlled interim campaigns to maintain spend in non-tested areas.
3. Use channel-specific exclusions to limit how new PMax campaigns distribute spend.
4. Leverage Mike Rhodes' script to analyze budget allocation before and after the test.

‍

By following these best practices, advertisers can ensure accurate incrementality results and make data-driven decisions about their Google Ads investments.

Ready to take the guesswork out of incrementality testing? Try Stella and unlock the full potential of your Google Ads campaigns today.
‍

Or try our virtual demo below: