The most common mistake in Google Ads accounts right now is treating Performance Max and Search as separate campaigns doing separate jobs. They are not. They share auction inventory, they compete for the same queries, and without deliberate structural decisions, they will cannibalize each other in ways that are difficult to see in platform reporting.
What Performance Max Actually Is
Performance Max is a single campaign type that runs across every Google network simultaneously: Search, Shopping, Display, YouTube, Gmail, and Maps. It uses your asset groups, audience signals, and conversion data to determine where, when, and to whom to show your ads.
The critical architectural point is that PMax does not target keywords. It targets intent signals. These signals include your landing pages, your creative assets, your audience signals, and the conversion history of your account. The auction system then decides which network and format to use for each impression.
This means PMax can and does appear on Search results pages. When it does, it is not using keyword targeting in the traditional sense. It is matching on inferred intent, which is broader, less predictable, and largely invisible in the search terms report.
Where PMax Wins and Where Search Beats It
PMax performs best when your goal is volume across a broad intent range, your creative assets are strong, your conversion tracking is solid, and you have enough conversion history for the algorithm to learn effectively. For ecommerce accounts with product feeds, it is often the highest-volume campaign type in the account.
Search performs best where intent is explicit and specific. Brand queries, competitor queries, and high-intent non-brand queries with clear purchase signals are Search territory. The user has typed exactly what they want; you want to respond with exactly what you offer. PMax's broader matching is a liability here, not an asset.
The mistake is treating these as competing frameworks. They are complementary, but only if the account structure enforces the boundary between them.
The Cannibalization Problem
When PMax and Search run simultaneously without guardrails, PMax will absorb brand queries that Search was handling efficiently. You will see Search impression share drop, PMax volume increase, and aggregate CPA either hold or appear to improve. The reality is that PMax is claiming credit for conversions that Search would have generated at lower cost.
Google's position is that the system prioritizes the best ad for each query. In practice, this means PMax often wins the auction for queries where a Search ad would have performed better, because the system does not always have visibility into the full performance difference.
The structural solution is not complicated: brand keyword exclusions in PMax prevent it from showing on brand queries. Competitor keyword exclusions prevent overlap with competitor campaigns. These exclusions need to be set up correctly and reviewed regularly as brand terms evolve.
Feed-Based Optimization for Ecommerce
For ecommerce accounts, PMax is heavily feed-dependent. The quality of your product feed is more important than any campaign setting. Titles, descriptions, images, and categorization determine how well the algorithm can match your products to relevant queries.
Feed segmentation is the primary lever for maintaining some control within PMax. Separating products by margin, performance tier, or category into different asset groups allows differentiated bidding and clearer performance reporting. Without segmentation, a small number of high-converting products will absorb most of the budget while the rest of the catalogue is barely shown.
The best-practice structure for ecommerce is separate asset groups for top-performing products, mid-range products, and new or seasonal products, each with appropriate ROAS targets and creative assets specific to that category.
The Role of Search in a PMax Account
Search campaigns serve two functions in an account that runs PMax: brand protection and high-intent capture.
Brand protection means ensuring that searches for your brand name, brand variants, and branded product terms return exactly the ad you want, with the messaging you control, at the position you choose. PMax cannot guarantee this. A dedicated brand Search campaign can.
High-intent capture means identifying the specific non-brand queries where intent is explicit enough that keyword-level control produces better results than PMax's intent matching. Category-specific queries, comparison queries, and queries with strong purchase signals often belong here. The test is simple: does a tightly structured Search campaign on these queries outperform PMax on the same terms? If yes, keep them in Search.
Reporting Limitations and How to Work Around Them
PMax's search terms report shows a fraction of the actual queries it matched. This is not a reporting error; it is by design. The queries that appear are those with enough volume to meet Google's threshold. The rest are aggregated or hidden.
The practical workaround is to use Search campaigns as a signal layer. Run Search campaigns broadly enough that the search terms report captures what users are actually searching. Then use that data to build exclusions for PMax, structure asset groups, and identify the high-intent queries worth isolating in Search.
Incrementality testing is the only reliable way to measure PMax's true contribution. Run a geo-based holdout test: pause PMax in a set of comparable markets for four weeks and measure the difference in conversion volume and revenue. The results are consistently more informative than any attribution comparison between PMax and Search reporting.
The Account Structure That Actually Works
The structure I have found most durable across accounts of different sizes is this: one dedicated brand Search campaign with tight keyword control and its own budget; PMax with brand and competitor exclusions, segmented asset groups, and conversion-based bidding; and selective non-brand Search campaigns for the highest-intent queries where keyword control produces measurably better results than PMax.
This is not the simplest structure, but it is the one that produces reliable, defensible performance without the attribution confusion that comes from letting the two campaign types overlap without boundaries.