PMax's biggest structural flaw is not that it is a black box. Every ad platform is a black box now. The flaw is that PMax bundles brand traffic, non-brand search, shopping, display, YouTube, and Discovery into a single blended reporting number — then optimizes bids based on that blended average. Which means the campaign gets rewarded for winning traffic it was going to win anyway, and punished for the parts that were doing the actual acquisition work.
The most common pattern we see: PMax reports a 6.2× ROAS. The team is thrilled. We split brand from non-brand and discover brand searches alone account for 68% of the reported conversion value — traffic that would have converted through organic branded search anyway. The 'non-brand' portion of PMax is actually running at 1.9× ROAS, which is below break-even for most e-commerce margins.
The Two Ways PMax Cannibalizes
1. Direct Brand Query Overlap
Someone Googles your brand name. PMax serves your Shopping or Search ad. They click the ad instead of the organic listing right below it. You pay $2-$8 for a click you would have gotten for free. PMax counts the conversion. Organic traffic drops. Google reports campaign ROAS higher than it deserves.
This is the easiest cannibalization to detect. If your organic branded search click-through rate dropped materially in the last 90 days but organic branded impressions did not, PMax is likely absorbing clicks that were previously going to organic.
2. Search Query Absorption from Existing Non-Brand Campaigns
You have a dedicated Search campaign bidding on non-brand keywords. PMax launches. It starts winning some of the same queries because its bid signals are stronger — it has more data, more conversion signals, and Google internally prioritizes it. The Search campaign impression share drops. PMax impression share (which you cannot see for individual queries) goes up. Same queries, same customers, but attribution shifts to PMax.
The Detection Query
You cannot get per-query performance out of PMax directly. What you can get is the Insights tab (Tools → Insights → Search Terms Insights) and the Categories report, which groups similar queries. Both are less granular than a Search campaign but usually enough to spot brand overlap.
- 1Pull the Search Terms Insights report for your PMax campaign, filtered to the last 30 days.
- 2Filter categories that contain your brand name, variations of your brand, or trademarked product names.
- 3Sum the conversion value and cost across those branded categories.
- 4Divide branded-category conversion value by total PMax conversion value. This is your brand contribution ratio.
If that ratio is above 40%, your PMax campaign is primarily a brand campaign wearing a Performance Max costume. The reported blended ROAS is inflated by traffic that had nothing to do with the algorithm — it was going to convert regardless.
How to Fix It
1. Apply Brand Exclusions at the Campaign Level
PMax now supports brand exclusion lists (they rolled this out in mid-2023, though the UI is buried in campaign settings). Add every variation of your brand name — with common misspellings, trademarked product lines, and your domain name — to the exclusion list. This forces PMax to compete for non-brand traffic only.
The first week after applying brand exclusions is uncomfortable. Reported ROAS drops. Conversion volume drops. It looks like the campaign got worse. It did not. It stopped counting the free traffic. What you see now is the actual acquisition performance.
2. Run a Parallel Brand Search Campaign
Brand search should be its own dedicated Search campaign with its own budget and its own bidding strategy — usually manual CPC or Target Impression Share. That way brand queries are still captured (you probably do want to defend against competitors bidding on your brand), but the cost and performance are visible separately from your non-brand acquisition efforts.
3. Measure PMax on Non-Brand Only
Once brand is excluded from PMax, the reported ROAS you see is the actual acquisition ROAS. It will be much lower than what PMax previously reported. This is not the campaign getting worse — it is the campaign finally being measurable. From there you can make real decisions: scale it up, cut it, restructure it, or supplement it with more targeted Search campaigns.
Why Most Accounts Never Do This
Applying brand exclusions makes the number look worse. Nobody wants to walk into a meeting and say 'the ROAS I reported last quarter was inflated by 40% and now we need to talk about what to do differently.' It is much easier to leave PMax as-is, keep reporting the blended number, and hope nobody asks hard questions about incrementality.
The problem is that decisions compound. If PMax gets scaled based on its inflated ROAS, actual acquisition efficiency drops — because the additional budget goes to more brand cannibalization, not more net-new customers. Six months later the account is spending double and revenue is up 20%. Nobody understands why the "high-ROAS" campaign is not scaling to expectations.