For Google Ads managers · Updated Nov 9, 2026
Run a Performance Max test
Performance Max doesn't support standard custom experiments, so a Scowtt test on Performance Max runs as a paired-campaign split instead. You build a second Performance Max campaign that mirrors your existing one, point it at Scowtt Custom Goal, and split the budget evenly between the two.
How it works: You duplicate your Performance Max campaign, set the copy to bid on Scowtt Custom Goal, and launch both on the same day with even budgets. Scowtt reviews the setup and reads the result with you against your real funnel outcomes.
Why not a Performance Max experiment?
Performance Max has its own experiment types, but none of them compares two bidding signals:
| Experiment type | What it compares | Fit for a Scowtt test |
|---|---|---|
| Uplift | Performance Max added alongside other campaign types | No. It tests whether Performance Max helps at all, not which signal it bids on |
| Upgrade | Moving budget from Shopping or DSA/GDA into Performance Max | No. It's a migration test |
| Optimization | Variations of settings or assets within one campaign | Partly. It covers settings changes, not a separate bidding signal |
Before you start
- Scowtt Custom Goal created in your account. See Set up your Google Ads test campaign.
- The existing Performance Max campaign you'll test against, agreed with your Scowtt team.
Steps
-
You: Duplicate the campaign
Duplicate your existing Performance Max campaign, so the assets, asset groups, audience signals, locations and schedule all match.
-
You: Point the duplicate at Scowtt
On the duplicate, set the conversion goals to Scowtt Custom Goal only, and the bid strategy to Maximize conversion value.
-
You: Split the budget evenly
Give each campaign its own daily budget, roughly 50/50 of the combined spend. Don't use a shared budget: one campaign can starve the other and ruin the comparison.
-
You: Match exclusions on both
Apply brand exclusions and account-level negatives identically to both campaigns, so they compete with each other as little as possible.
-
Scowtt: Review the setup
Your Scowtt team checks both campaigns before launch.
-
You: Launch both on the same day
Start both campaigns together, then leave them alone.
-
Scowtt: Read the results with you
Your Scowtt team compares the two on cost per funnel stage, using the ground truth reporting columns.
Also available: Campaign Mix experiments (early access)
Google is introducing Campaign Mix experiments, which test up to five groups of campaigns against each other in one experiment, including across campaign types such as Search, Performance Max and Demand Gen. For Performance Max, this could give a cleaner traffic split than two paired campaigns.
They're currently in early access, so they aren't available in every account yet. Scowtt's recommended approach is still the setup on this page, while Campaign Mix is proven at scale. If it's available in your account and you'd like to use it, talk to your Scowtt team first so the test still compares Scowtt's signal fairly. Google's guide: About Campaign Mix experiments.
Good to know
- Expect more noise than Search. Performance Max campaigns share inventory and can take traffic from each other, so a paired split is directionally sound rather than a clean randomized trial. Even budgets, the same start day, and comparing on cost per funnel stage rather than raw volume keep it fair.
- If you'd prefer a native Performance Max experiment, the Uplift experiment is set up at Campaigns > Experiments, plus button, Campaign types > Performance Max Uplift. It defaults to a 50/50 split and compares against similar campaigns Google selects about a day after launch. Turn on comparable-campaign reporting to see them. Be clear about what it answers: the added value of Performance Max, not the value of Scowtt's signal.
Who does what
| Step | Who | What changes hands |
|---|---|---|
| Duplicate the campaign | You | — |
| Point the duplicate at Scowtt | You | — |
| Split the budget evenly | You | — |
| Match exclusions on both | You | — |
| Review the setup | Scowtt | Setup review |
| Launch both on the same day | You | — |
| Read the results with you | Scowtt | Results, by cost per funnel stage |