PMax reports a spectacular ROAS. Some of it is customers who were already yours.

The symptom pattern

You are probably looking at cannibalization if several of these are true at once:

• Performance Max reports the highest ROAS in the account by a wide margin
• Total store revenue has not risen in proportion to the PMax spend increase
• Your branded Search campaign's impression share has fallen since PMax launched
• Branded Search volume dropped while PMax conversions rose by a similar amount
• Blended MER is flat or declining while platform-reported ROAS improves

The last one is the tell. Attribution can be moved around between campaigns; total revenue divided by total spend cannot. If platform-reported performance is improving and blended efficiency is not, credit is being reallocated rather than value created.

The diagnosis, in order

1. Pull branded Search campaign impression share for the six weeks before PMax launched and the six weeks after. A meaningful drop with no bid or budget change is your first piece of evidence.
2. Open the Performance Max insights and look at the search categories driving conversions. If your own brand name is prominent, you have your answer.
3. Compare total store revenue across the same two periods. If PMax conversions rose by 400 and total orders rose by 90, you know where the other 310 came from.
4. Check blended MER across both periods. Flat MER with improved platform ROAS is the signature.

What to do about it

1. Apply brand exclusions

Google provides brand exclusions for Performance Max, which prevent the campaign from serving against your own brand terms. This is the direct fix and it should be the first thing you do. Expect reported PMax ROAS to fall afterward — that fall is the measurement getting more honest, not the campaign getting worse.

2. Run a dedicated branded Search campaign

Keep your brand terms in a tightly controlled Search campaign with its own budget, where you can see exactly what you are paying and what it returns. Branded search is usually cheap and usually worth defending against competitors bidding on your name — but it should be a visible, deliberate line item rather than something absorbed into an opaque campaign.

3. Add account-level negative keywords

Account-level negatives apply to Performance Max and are the coarser instrument for keeping it out of territory you want controlled elsewhere.

4. Judge PMax on new-customer acquisition

If your store reports new versus returning customers, judge Performance Max on the new-customer share of its conversions rather than on blended ROAS. A campaign whose conversions are overwhelmingly returning customers is a retention channel priced as an acquisition channel.

When Performance Max genuinely earns its place

This article is a caution, not a condemnation. Performance Max works well in specific circumstances, and dismissing it wholesale is as expensive as running it blind.

• Large catalogs where manually structuring campaigns per product is impractical
• Accounts with strong conversion tracking and enough volume for the system to learn from
• Brands with little existing branded search volume to cannibalize — where the cheapest available conversions genuinely are new customers
• Discovery of demand across placements you would not have bid on manually

The rule is straightforward: run it with brand exclusions on, a separate branded Search campaign, and judge it on incremental new customers rather than on the ROAS it reports about itself.