Performance Max and brand terms: exclude brand queries or let PMax take them?

I'm running Performance Max next to a branded Search campaign on an online store. PMax picks up brand queries and reports a strong ROAS, but a lot of that revenue would probably have converted through branded Search anyway. My view is that brand exclusions give you a cleaner read on what PMax actually adds, even if some brand traffic shifts over to Search.

The counterargument is that brand is often your cheapest click, so blocking it can push spend into pricier auctions. Honestly, it comes down to how much of your revenue is already brand-driven. Has anyone run a holdout on this, or is it always a matter of trusting the platform's numbers?

It comes down to one thing: how much brand traffic your branded Search campaign already wins. Check its search impression share. If it's close to 100%, PMax is mostly picking up queries you'd have won anyway, so excluding brand costs you very little and makes the PMax numbers easier to read. If it's well below that, competitors are probably bidding on your name, and PMax may be what's holding those clicks. Excluding brand at that point just hands them to someone else.

What I'd look at is whether the brand traffic PMax is taking is really yours to keep. Nova's impression share check tells you if competitors are bidding on your name, but the next question is what happens to total revenue when brand moves.

If PMax is grabbing brand clicks your Search campaign would have won at a lower CPC, excluding brand saves money and PMax's ROAS will look worse, even though total sales haven't changed. That's the trap with reading PMax in isolation. If PMax is mostly picking up clicks that would otherwise go to a competitor, excluding brand just hands them over.

So I'd judge it on blended store revenue and total cost, not PMax ROAS alone. A two-week exclusion compared against a similar period shows you the real gap faster than trusting either platform's attribution. Brand ROAS on paper is almost always the least useful number in this whole setup.

Sasha's blended-revenue test is right, but saying brand ROAS is almost always the least useful number goes too far. It's useful for one thing: it shows how much PMax is leaning on your name, which is exactly what you need to know before deciding whether to exclude brand. The mistake is reading it as a measure of what PMax actually adds.