Brand in Performance Max: what brand exclusions really prevent

A brand exclusion solves a blocking problem, and only partly. It does not solve your measurement problem at all. My advice: keep the two apart. Blocking takes ten minutes. Proof of whether it was necessary and what it costs depends on the period you measure against. And the ROAS you see afterwards is a new figure for a different campaign.

The block only applies to search queries, and even there it depends on a tick box

According to Google, brand exclusions in PMax only apply to Search, Shopping and YouTube search. Display and other placements without a search query are not affected. Within search, an option in the campaign settings decides whether Shopping ads may keep running on brand searches.

This table explains one possible false alarm. With the Shopping option on, brand search terms still appear in the report, and ROAS falls less. That is the setting you chose, not a broken exclusion. You can check it with the ad format segment, because the PMax search terms report separates Shopping and text ads for each search term.

Compared with negative keywords, Google calls the brand list often the better option, because it also covers clear misspellings and foreign scripts. You should know two limits of the list in advance:

• Brands made of ordinary words: Google only blocks such terms if they clearly refer to the brand. If your shop is named like a product category, some traffic still gets through. You catch it specifically with a negative keyword.
• Variants without a shared word: parent and child brands have to be added to the list individually. A private label under the umbrella of the shop name therefore needs its own entry.

The brand Search campaign has to be running first

Launch the brand campaign first, check that it serves, and only then exclude. Otherwise you will temporarily have no ad on your own brand, while competitors can keep bidding.

The exact match keyword has a technical reason: a Search keyword that is identical to the search query, or matches it through spelling correction, takes priority over PMax; otherwise the higher Ad Rank decides. After the exclusion, brand plus product also only runs through this campaign and needs keywords there, otherwise a gap opens up. Whether you should bid on your brand at all is a separate calculation: bidding on your own brand.

Evidence only counts from the last setup change

In my view the period distorts the picture more than anything else. The search terms report goes back to March 2023, so a year is easy to export. But that year often contains several setups: before the exclusion, with the Shopping option, after the launch of the brand campaign. The average describes a state that never existed.

A worked example with invented numbers: suppose a PMax campaign costs €10,000 a month. For nine months without an exclusion, €1,000 a month flows into brand search terms, and for the last three months, with the exclusion, it is €50 a month.

The 7.6% is arithmetically correct and says nothing about the current campaign: neither whether the exclusion works nor what it costs. Every change to the brand list, the Shopping option, the brand campaign or negative keywords resets the measurement period.

How to measure the brand share

Brand queries in PMax distort performance measurement. Correcting for this takes four steps:

1. Note the date of the last setup change. That is the earliest permissible start date.
2. Pull the PMax search terms report from that date and segment it by ad format.
3. Mark every term containing your brand, including misspellings and brand plus product.
4. Show the share of cost and the share of conversion value separately, then calculate the return without brand.

The formula:

Non-brand ROAS = (campaign value − brand value) / (campaign cost − brand cost)

This remains an approximation. Low-volume queries are missing because of privacy thresholds, and placements without a search query have no search term anyway. Both end up in the non-brand remainder. Read the brand share as a lower bound. If non-brand ROAS and the blended figure sit on the same side of your break-even ROAS, the split does not change any decision, and you can save yourself the effort.

Why ROAS may fall afterwards

After the exclusion, a falling campaign ROAS is to be expected, and Google also writes that performance may decrease. Again with invented numbers:

Here ten percent of the cost carries a third of the value. After the exclusion, the yardstick is therefore no longer 6 but the break-even ROAS, in simple terms 1 divided by the margin after variable costs. At a 30% margin it is 3.3, so non-brand pays for itself. At 20% it is 5, and the non-brand part was already losing money before, it just could not be seen under the 6. That is the real value of the measure: you learn this figure.

The target then has to follow. My assessment: a target ROAS left over from the blended period can starve non-brand traffic of budget. In the example you demand 6 from traffic that has delivered 4.4 so far. Derive the new target from your margin, not from the old blended figure.

Assess the account as a whole by spend, revenue, orders and new customers, in a window of at least four weeks set in advance. The comparison window before it must also contain only one setup and should not compare a promotion period with normal weeks. My reading of the results:

The last row is the least comfortable. If the budget stays the same, the money that has been freed up finds other traffic. Only a calculation on contribution margin shows whether that traffic pays for itself.

No exclusion without brand Search, Shopping option usually off

Without a running brand Search campaign I would leave the exclusion out, because otherwise you lack visibility on your own brand. The same applies to brands made of everyday words if the report mainly shows category demand. There the block rarely hits cleanly, and the share is hard to calculate. Without a measurement window it is premature too: if you are rebuilding the feed, target or budget in the next four weeks anyway, you will not get a readable result and are better off postponing the exclusion.

As a rule I would switch the Shopping option off, because only then does the calculation above hold. I would only leave it on if your products would otherwise not appear anywhere on brand searches and you want to keep them visible. The price: the brand share in the Shopping format stays in your PMax ROAS and in the bidding signal, you have to calculate it out as described above, and the target must not drop to the pure non-brand figure.

Features as of November 2026.