PMax and brand search: exclude brand terms or let PMax keep them?

Say most of your conversions come from people searching your brand name. PMax will happily claim those orders and the ROAS looks great, but most of those shoppers would have found you anyway, so you're paying for sales you'd likely get for free. Brand exclusions are the blunt fix, though they also strip out traffic that helps PMax learn. I'd keep brand in for now, split brand revenue from non-brand in your reporting, and watch the share of new customers. If most of the value is brand, the campaign isn't really finding customers, it's collecting them. Has anyone tested a separate brand Search campaign with its own target to see what PMax adds on top?

A separate brand campaign gives you a clean read on brand CPA, but a holdout tells you more. Pause brand for two weeks and compare total orders across all channels, not PMax's reported number, because that number drops while the real total often doesn't. Has anyone actually run that?

A holdout only tells you something if you can see total orders outside Google, which means you need a clean first-party source for them. If everything runs through a single shop platform, a two-week pause is easy to read. If orders are split across marketplaces, phone and in-store, you'll mostly be comparing noise.

Agreed, and a PMax brand exclusion isn't a clean switch. It's set at account level and only covers Search and Shopping inventory, so YouTube, Display and Maps can still serve brand-driven impressions during the pause. Search Console branded clicks make a handy side check. If they fall and total orders hold, brand was mostly collecting. If branded organic clicks rise to fill the gap, some of that revenue is coming back through a different source, which means your "lost" orders weren't really lost.

Brand exclusions don't have to be account-wide. You can apply them per campaign, so you can switch brand off on one PMax campaign and leave the others running. That makes the holdout a lot easier to isolate.

Kai's right that the exclusion can be set per campaign, which is what makes a holdout workable. As far as I know it's a campaign-level setting, so Rowan's account-level point is out of date.

The weak spot is the brand list itself. It only blocks queries that match it, so misspellings, abbreviations and old company names can keep feeding the "paused" campaign. Before you read the results, go through the search terms for that campaign and look for brand-like queries that got through. If a meaningful slice of its revenue is still brand, your holdout is only partly a holdout, and the drop you see will understate how much work brand was actually doing. Fixing the list before the test starts is a lot cheaper than explaining a weak result afterwards.

PMax search term reporting is only partial, which is easy to forget. Google surfaces terms above a volume threshold and lumps the rest into groups, so a clean-looking term list can still be hiding brand leakage. Brand queries also slip through when the brand word sits inside a product search, like "acme trail runner 42", and a broad brand list won't always catch those. For a cross-check that doesn't depend on Google's grouping, compare branded Search Console clicks over the same weeks.

Branded clicks in Search Console only cover organic results, so they can't see paid. If a shopper searches "acme trail runner 42" and clicks a Shopping ad, GSC records nothing. That means a drop in branded organic clicks during a holdout might just be people switching to paid. Compare it with paid brand clicks from the same weeks, or you'll read it backwards.

The regex needs a sanity check too. Something like acme|acme trail catches product queries, but if your brand name is also a common word, it will match unrelated searches. Go through the matched query list before you trust the total. GSC also hides a lot of low-volume queries, so treat the branded number as a floor.

It depends on whether competitors bid on your brand name. If they do, pausing brand coverage doesn't just shift orders to organic. Some shoppers will land on a competitor's ad instead, and that order is simply gone, without showing up as a drop in branded organic clicks. Auction insights on your brand terms will show how often competitors appear next to you and how often they outrank you. Pull that report before the test starts, because it tells you how much of any drop you can actually pin on brand.

Agreed, and the competitor case is easy to miss in the numbers. Before you run the test, look at lost impression share due to rank on your brand terms. If a competitor already outranks you on your own name, pausing brand just hands them the order, and nothing in your reports will show it as a drop.