Performance Max is no longer the black box it was when Google retired Smart Shopping. The reporting surfaces have grown, search themes are finally a first-class input, and brand exclusion lists are usable instead of a beta toy. None of that matters if you're still running the campaign with the defaults Google nudged you into years ago.
This post covers what we're doing differently on Performance Max accounts right now. No fabricated ROAS screenshots - just the structural decisions, the audience signal patterns, and the brand exclusion playbook we use on accounts running through our Performance Max management practice.
What changed in PMax over the past two years?
Three things matter most:
- Search themes went from a suggestion to a primary lever. You can now tell Google what queries you want to compete on, and the system actually weights those signals against its own automated query expansion.
- Brand exclusions moved out of the support-ticket-only era. Account-level negative brand lists, campaign-level brand exclusion lists, and the ability to exclude your own brand from PMax to protect a separate brand search campaign are all now self-serve.
- Reporting finally exposes asset-group-level spend, conversions, and asset performance ratings - not as fully as a Search campaign, but enough that you can make structural decisions instead of guessing.
That doesn't mean PMax is "fixed." It means the parts you have to manage manually are different than they were.
How should I structure asset groups today?
The honest answer: one asset group per intent cluster, not per product feed segment. Earlier guidance to mirror your Shopping campaign structure inside PMax was a reasonable starting point, but it produces asset groups full of products with nothing in common except a category label.
What works better:
- Intent cluster, not category cluster. Group products that share a buyer mindset - replacement-cycle, gifting, new-installation, accessory-attach - rather than products that share a taxonomy node.
- One asset group per landing page family. If your accessories live on a different landing page experience than your core SKUs, those are two asset groups. Message-match the creative to the page, not to the product.
- Cap at three to five asset groups per campaign. Beyond that, Google's optimization struggles to allocate budget cleanly. If you genuinely have more than five intent clusters, split into multiple PMax campaigns, not more asset groups in one.
For the underlying Google Shopping feed, the same principle applies: structure your custom labels to mirror intent clusters, not just margin tiers. Custom_label_0 for intent, custom_label_1 for margin, and let your asset group rules pull from both.
Are audience signals worth setting?
Yes - but not the way the in-platform UI implies.
Audience signals are hints, not targeting. Google's documentation has been clear about this for some time. What that means in practice:
- Customer Match lists are your strongest signal. Upload your purchaser list, your high-LTV cohort, and your repeat-buyer cohort separately. Don't merge them.
- Your own first-party visitor segments (via GA4 audiences imported into Google Ads) outperform demographic interest signals 9 times out of 10.
- Demographic and interest signals are the weakest tier. We set them, but we set them last and we don't agonize over them.
- Search themes carry more weight than the audience signal section in current PMax behavior. If you have to choose where to spend your setup time, put it into search themes.
For accounts running on a strong first-party data foundation - meaning Enhanced Conversions and Customer Match are both wired up - the audience signal section becomes meaningfully more useful. Garbage in, garbage out applies here as forcefully as it does anywhere else in paid media.
What's the brand exclusion list playbook?
Brand exclusions in PMax serve two distinct jobs:
Job 1: Stop PMax from cannibalizing your own brand search campaign. If you run a dedicated brand Search campaign - and you should, for impression share and message control - your PMax campaign needs to exclude your own brand. Otherwise PMax will eat brand traffic at PMax CPCs (often lower) and claim the conversions, leaving your brand Search campaign starved and looking underfunded.
Job 2: Stop PMax from spending on competitor brand terms. Google's automated query expansion will, left unchecked, bid on competitor brand searches. Sometimes you want this. Often you don't - competitor brand traffic converts poorly for many ecommerce categories and burns budget that should be flowing to high-intent non-brand queries.
Tactical:
- Build an account-level negative brand list with your own brand plus every variant (misspellings, brand + product, brand + reseller terms).
- Build campaign-level brand exclusion lists for your top 10 competitor brands. Add to it monthly based on the brand query report.
- Pull the brand query report (now exposed in the Insights tab) every two weeks. New competitor brands show up faster than you'd expect.
For a deeper structural walkthrough, see our PMax brand exclusion strategies post.