The most effective way to use GA4 audiences with Google Performance Max is not to control who the campaign targets, but to provide the machine with rich, value-based signals about what a truly good customer looks like. Your goal is to inform the algorithm, not to fence it in. The common impulse to use audiences for precise targeting is a holdover from an older era of digital advertising, and it fundamentally misunderstands how PMax is designed to work.
The real problem underneath this question is a desire for control in a system that’s been deliberately turned into a "black box," as Mati Ram explains on Ecommerce Coffee Break. For years, savvy advertisers built careers on granular control of keywords, bids, and placements. PMax changes the game. As Sam Piliero notes, it combines Google Shopping, YouTube, Display, Search, and Demand Gen into a single campaign. Google’s proposition, much like Meta’s with Advantage+, is "give us your budgets and your assets and we will advertise for you." Trying to force old-world tactics onto this new paradigm is the core of the frustration many advertisers feel. The job is no longer micromanaging bids; it's curating the highest quality data inputs.
Over the last 18 months, PMax has aggressively grown to represent nearly 40% of all Google ad spend, a huge jump that Tony Chopp mentioned on Ecommerce Playbook. This isn’t a niche tool anymore; it's a central pillar of Google Ads For Ecommerce. The platform is learning at an incredible rate, but its evolution has made trusting the data harder than ever. This is where the consensus view on PMax gets things both right and wrong. The consensus is correct that PMax simplifies management, as Brett Curry points out. It is also dead right about the absolute necessity of splitting PMax into separate brand and non-brand campaigns. Without this, as Tony Chopp warns, PMax will simply bid on your own brand name, take credit for those high-intent conversions, and report a beautifully inflated ROAS that tells you nothing about its actual ability to find new customers.
Where the consensus often goes wrong is in treating PMax audience signals as restrictive targeting. Feeding PMax a "Past Purchasers" audience doesn’t mean it will only show ads to them. It uses that list as a seed to find millions of other users who look and act similar. This is a prospecting tool, which Shri Kanase makes clear on eCommerce MasterPlan. Therefore, feeding it a generic list of all past customers is a low-value activity. You are telling it to find more people who look like everyone, which it was already going to do. The real leverage is in the specificity of the signal.
Giving the Machine a Better Brain
Instead of generic audiences, your job is to feed the machine signals about your absolute best customers. This is where GA4 becomes incredibly powerful. You should build and send tiered, value-based audiences to Google Ads.
Here are some specific examples:
- High LTV Customers: Create an audience of the top 20% of your customers by lifetime value. Name it "VIP Customers (High LTV)."
- Frequent Purchasers: Build an audience of anyone who has purchased three or more times. Name it "Repeat Purchasers (3+)."
- High AOV Customers: Isolate users whose average order value is 50% or more above your site average.
When you add these audiences as signals to your non-brand PMax campaign, you aren’t just giving it a map. You’re giving it a destination with a specific treasure marked on it. You are teaching the machine to look for buying power and loyalty, not just a single conversion. This is the quality data input that the black box needs to generate quality outputs.
Another critical tactic is using negative audiences. You should absolutely exclude your existing customer list from your non-brand PMax campaign. This forces it to be a pure new customer acquisition engine and gives you a much cleaner read on your customer acquisition costs.
The Hidden Costs of a Black Box
Ignoring these nuances has significant second-order effects. First is the cannibalization that the Ecommerce Playbook hosts often discuss. Without a clean brand vs. non-brand split, PMax obfuscates true performance and makes it impossible to decide if it's working, as Ginny Marvin noted. You end up paying more for your own branded traffic.
Second, and more importantly, is the danger of trusting in-platform metrics. On the "Meet PAM: The Profit Allocation Model Revolutionizing Ad Budgets" episode, Luke Austin breaks this down perfectly. A PMax campaign might report a fantastic 7x ROAS. But if its actual incrementality is only 35% (meaning only 35% of its reported sales were truly caused by the ads), your real, incremental ROAS is a much lower 2.45x. They’ve found non-brand PMax campaigns can have an incrementality of around 75%, while brand-heavy PMax could be as low as 30%. Without doing the hard work of incrementality testing and blended analysis, you could be pouring money into a campaign that feels great on the Google Ads dashboard but is providing diminishing returns to the business as a whole.
My 30-60-90 Day Plan
If I were taking over an account today, here is what I would do.
First 30 Days: Foundation & Structure
- Audit the GA4 and Google Ads conversion tracking setup to ensure revenue and product data are flawless.
- Create the value-based-audiences in GA4 (High LTV, Repeat Purchasers, High AOV) and the corresponding customer lists.
- Restructure the Google Ads account to have separate Brand and Non-Brand PMax campaigns. The Non-Brand PMax must have brand terms added as negative keywords.
Next 30 Days: Implementation & Learning
- Apply the new High LTV, Repeat Purchaser, and VIP customer lists as audience signals to the Non-Brand PMax campaign. Exclude all past purchasers from this campaign.
- Following Sam Piliero’s advice, I’d regularly check the PMax insights tab to see how spend is being allocated across channels and what kinds of search themes are emerging. This isn't for optimization, but for understanding the machine’s behavior.
Next 30 Days: Optimization & Scaling
- Begin analyzing performance based on blended ROAS (Total Revenue / Total Ad Spend) and New Customer Cost of Acquisition (NC-CAC). I would stop making decisions based purely on PMax’s in-platform ROAS.
- Using the incrementality benchmarks from Ecommerce Playbook as a starting point, I would build a model to estimate the true incremental return of each channel.
- Budget decisions would now be based on incremental lift. If PMax has a higher platform ROAS but Meta is more incremental, I would shift spend to Meta. This is how you move from managing ad campaigns to managing a profitable growth portfolio.