Your best Performance Max campaign is just your brand name

Most advice about PMax and brand search stops at 'watch out for cannibalization' without explaining why it happens structurally. The reason is simple once it's stated plainly: PMax's bidding algorithm has a single objective — the lowest cost per conversion inside whatever target you've set — and it has access to every inventory type Google sells, including the exact-match branded search a customer types when they already intend to buy from you.

Branded search converts at somewhere between 10 and 20 times the rate of a cold prospecting impression, at a fraction of the CPC, because the person searching your brand name has already made most of the decision. An algorithm with no explicit instruction to avoid that traffic will find it, bid on it, and report it back as a top-performing conversion — because by its own objective function, it is.

This isn't a bug or an edge case. It's the predictable output of optimizing purely for cost-per-conversion without a constraint on incrementality, and PMax's opacity — no keyword-level reporting, no placement-level bid control — makes it unusually hard to notice happening until someone goes looking specifically.

The most common piece of advice — add your brand terms to your account-level negative keyword list — solves less of the problem than most advertisers assume.

The convincing version of this isn't a warning, it's a before-and-after. Apply a properly configured brand list to every PMax campaign — attached at the campaign level, covering brand-plus-modifier variants, not just the bare brand term — and watch what happens to the campaign's headline numbers.

In accounts where this hadn't been done, the pattern is consistent: the 'best performing' campaign's conversion volume and ROAS both drop sharply, sometimes by more than half, because a large share of what it was reporting as performance was branded demand that would have converted through organic search or a direct visit at zero incremental media cost.

This isn't an argument against PMax — the remaining, non-branded performance after a proper exclusion is the number that actually reflects whether the campaign is finding incremental demand, and for many advertisers it's still a reasonable result once measured honestly.

It is an argument for treating any PMax report with brand traffic still mixed in as unreliable for budget decisions. A campaign can look excellent and be almost entirely relabeled organic demand, or it can look mediocre once brand is excluded but represent genuinely incremental customers — you cannot tell which without doing the exclusion and re-measuring.

The broader pattern connects to a theme worth stating plainly: automated bidding systems optimize exactly what you tell them to, with no judgment about whether the traffic they find is actually new. Brand exclusion isn't a minor hygiene setting — for most advertisers running PMax, it's the difference between a real performance number and an expensive way to measure demand you already had.