What is Performance Max incrementality?
Performance Max incrementality measures whether PMax campaigns generate conversions that would not have occurred without the campaign running. A PMax campaign with strong reported ROAS may still have low incrementality if it is primarily capturing brand searches, retargeting existing customers, or taking credit for demand that would have converted through organic or other paid channels.
Why is ROAS not enough to evaluate PMax performance?
ROAS measures revenue attributed to the campaign by Google's attribution model. It does not measure whether that revenue was caused by the campaign. PMax can show high ROAS while producing few genuinely incremental conversions if it is primarily serving users who would have converted anyway through brand search, direct traffic, or retargeting. Attribution gives you credit, not causation.
How do you run a brand exclusion holdout test?
Add brand exclusions to PMax for a defined test period (minimum 4 weeks). A dedicated brand Search campaign must be running to maintain brand coverage. Compare total account conversions, revenue, and total spend before and after, not just PMax metrics. If total account performance holds or improves with brand exclusions in place, PMax was likely claiming brand credit rather than generating new demand. This is the simplest incrementality test to run.
How do you run a geo or time-based holdout?
Split your geographic coverage or time schedule to create a test and control group. Pause PMax in one region or reduce PMax budget in one time period while holding all other variables constant. Compare conversion rates between test and control. This approach requires volume in both groups to produce meaningful results. Use this method when brand exclusion tests are inconclusive or you want to test incremental value in specific markets.
What should you measure before and after?
Before: document total account conversion volume, total account ROAS or CPA, brand Search impression share, organic branded search clicks (from Search Console), and direct traffic conversions. After: compare all of these, not just PMax metrics. Incremental demand should show up in total account growth, not just reallocation of credit between campaigns. The key is measuring the total, not the parts.
| Test Type | Best For | Weakness | Setup Difficulty | What To Measure |
|---|---|---|---|---|
| Brand exclusion holdout | Accounts with active brand Search campaign | Only measures brand query incrementality | Low | Total account ROAS and conversion volume |
| Geographic holdout | Accounts with regional coverage and sufficient volume | Requires volume in both test and control regions | Medium | Conversion rate difference between regions |
| Time-based pause test | Small accounts without geo coverage | Seasonality can confound results | Low | Pre/post total account conversions |
| Google Ads experiment (campaign draft) | Accounts with access to campaign experiments | PMax experiments have limited configuration options | Medium | Incremental conversions and ROAS |
| Incrementality study (Google partnership) | Large accounts with dedicated Google rep | Not available to most advertisers | High | Modeled lift measurement |
Before starting the test
- –Baseline period established: minimum 4 weeks of stable performance
- –Total account conversions, ROAS, and CPA documented for baseline period
- –Brand Search campaign is live and has adequate budget to capture brand demand
- –Organic branded click baseline from Search Console exported
- –Test type and duration agreed before starting
During the test
- –No other structural changes to the account (budget, bidding, campaigns)
- –Brand Search impression share monitored weekly
- –Any external events (promotions, seasonality) documented
After the test
- –Compare total account metrics, not PMax metrics
- –Compare organic brand clicks (Search Console) for displacement signs
- –Calculate conversion volume difference between test and control or test and baseline
- –Document results before making permanent structural changes
For more on what to look for during a test, see our guide on brand search cannibalization. To understand what strong PMax performance actually looks like, read about profitable but not incremental campaigns.