PMax Brand Cannibalisation: How to Audit Performance Max
Performance Max can capture searches for your brand and reach people who already buy from you. That is worth investigating. It is not, on its own, evidence that those sales would have happened without ads.
Direct answer
Performance Max can serve against searches for your own brand and can reach existing customers. Where that happens at scale, the campaign's reported efficiency partly reflects demand you already had. That warrants investigation. It does not establish what would have happened without the spend.
Keep three things separate: attributed conversions (what the platform credits), observed brand-query exposure (where ads actually appeared), and causal lift (what changed because the ads ran). Only the third answers the budget question, and only a designed test estimates it.
The audit checklist: five signals to read together
These checks are read-only; none requires pausing a campaign. Separate your own brand from brands you resell, match attribution settings and comparison periods, and keep undisclosed search terms unknown.
| Signal | Alternative explanation | Next check |
|---|---|---|
| Brand share | A high classified share may reflect existing demand or a short-term promotion, not spend that could simply be removed. | Separate own-brand queries from resold brands; record undisclosed terms as unknown and compare like-for-like periods. |
| Brand search demand | Search interest can move with organic visibility, PR or other media rather than PMax activity. | Compare Search Console brand impressions, paid Search and total orders around campaign changes. |
| Channel mix | A search-heavy period may reflect inventory eligibility or creative changes, not a lasting campaign bias. | Read channel reporting alongside asset groups, exclusions and product availability. |
| New customers | Platform classification may miss existing customers or be affected by customer-list coverage. | Reconcile with first-party customer records and note unmatched orders. |
| Contribution | Higher attributed revenue can coexist with lower retained margin after returns, discounts or stock constraints. | Compare contribution after variable costs and ad spend in finance data, allowing for return lag. |
Reporting alone does not prove incrementality. Agree the commercial outcome and decision rule before a feasible controlled test, then account for stock, promotions, price, other media and conversion lag as described below.
Search themes are not a list of the queries you served
Search themes are signals an advertiser supplies to the campaign. They tell the system what to consider; they are not a report of the searches that actually triggered ads. Google: search themes.
For what was served, use the search terms reporting and the channel-level reporting Google documents for Performance Max, which distinguish search terms from other inventory. Google: Performance Max reporting. Terms that are not disclosed stay unknown in your analysis. Rolling them into a non-brand bucket manufactures a number you do not have.
What brand exclusions actually control
Google's current instructions describe brand exclusions as applying to specified search inventory, listed as Search, Shopping and YouTube search, with an option to allow Shopping ads for excluded brands. Google: brand exclusions and brand lists. Interfaces and available options differ between accounts, so read the current instructions against your own account rather than a screenshot from elsewhere.
Two consequences follow. Applying an exclusion does not prove the remaining sales are incremental, and it does not remove every route by which an existing customer can see an ad. Whether you should exclude at all is a commercial question about margin, competitor activity on your brand terms and what your Search campaigns already cover, not a default setting.
Designing a test worth trusting
A controlled experiment can estimate lift, subject to its design and its uncertainty. A sequential before-and-after, or simply pausing a campaign for a while, is directional evidence and is confounded by everything else that moved in that window. New-customer rate is a useful diagnostic, not a measure of incrementality.
- Write the hypothesis and the decision it will inform before you start.
- Choose a primary outcome at total-business level, ideally contribution rather than platform-reported revenue.
- Assign units to test and control where the account and geography allow it.
- Check baseline balance between groups before the test period.
- Allow for conversion lag in both the run length and the read.
- Work out sample size, power and the minimum detectable lift. If the effect you care about is smaller than that, the test cannot answer the question.
- List confounders in advance: stock, promotions, price changes, other media.
- Consider contamination and spillovers between groups.
- Set guardrails and a predetermined decision rule, then follow it.
Platform experiment options for Performance Max vary by account and change over time; not all of them are brand holdouts. Check what is currently available to you: Google: Performance Max experiments. There is no universal duration and no universal threshold.
Hypothetical diagnostic example
Illustrative only, not client results
| Bucket | Spend |
|---|---|
| Total Performance Max spend | £10,000 |
| Classified own-brand search spend | £4,000 |
| Classified non-brand search spend | £3,000 |
| Unclassified or other inventory | £3,000 |
Known own-brand share of total spend: £4,000 / £10,000 = 40%. Known own-brand share of classified search spend: £4,000 / £7,000 = 57.1%.
Neither figure estimates wasted spend or incremental revenue. They describe how much of the spend you have been able to classify, and how the classified part splits. These are illustrative buckets; campaign reporting may not expose a complete partition like this in your account.
FAQs
Does brand traffic in Performance Max mean wasted spend?
Not by itself. Brand exposure shows where ads appeared, not what would have happened without them. Attributed conversions, observed brand-query exposure and causal lift are three different things. Treat a high brand share as a reason to investigate, using search terms, channel reporting and your own customer records, rather than as a measure of waste.
Do brand exclusions stop all remarketing to existing customers?
No. Google's current instructions describe brand exclusions as applying to specified search inventory, listed as Search, Shopping and YouTube search, with an option to allow Shopping ads for excluded brands. They control where ads can appear against those brand terms. They do not remove every route by which an existing customer can see an ad, and they do not demonstrate that remaining sales are incremental.
Is two weeks long enough to prove Performance Max is not working?
There is no universal duration. The period you need depends on your conversion lag, volume, the size of the effect you want to detect and the confounders in that window, such as promotions, stock and other media. A short sequential pause is directional evidence, not causal proof.
Sources
Related reading
Want a second read on your Performance Max spend?
We audit brand exposure, channel split and test design as part of retained Google Ads work. We cannot tell you your incrementality from the outside, but we can tell you what your reporting does and does not currently show.
Stay informed
Get our latest insights on Google Ads strategy delivered to your inbox. No fluff, no spam, just honest thinking.
Unsubscribe anytime. No spam, ever.
Next step
Where does Performance Max actually belong in your account?
We look at what PMax is being asked to do, which products it should carry, and which demand it is simply re-buying.