How to Keep Control of Performance Max
Why it feels uncontrollable
Performance Max removed the levers a generation of practitioners learned on, and the common reaction is to conclude that nothing can be controlled. What actually happened is that control moved: from choosing which query to bid on, to choosing which signals, assets, feeds and exclusions the system learns from.
An account that feels like a black box is usually an account where those four things were inherited rather than decided. The system is not being opaque so much as inferring your intentions from whatever the import happened to contain, and it will keep doing that until somebody tells it otherwise.
Structure, feed rules and brand separation turn a black box into something you can steer.
The reported performance problem
Performance Max will happily serve your own brand searches and report them as conversions. The campaign looks excellent and the business sees no new demand.
Separating brand traffic is the first thing to do, before judging anything else.
Structure that creates visibility
Split asset groups by product margin and intent rather than dumping the catalogue into one. Use feed rules to exclude products you do not want promoted. Give each group its own creative so performance is attributable.
Then compare incremental results against the period before it launched, not against the campaign’s own reporting.
When it genuinely works
Large catalogues with clean feeds and strong conversion data do well. Lead generation accounts with weak conversion definitions usually do not.
Where control still exists
Performance Max removes the levers everybody learned on and leaves four that matter more: the conversion signal it optimises towards, the feed it draws products from, the creative assets it assembles, and the exclusions you set. An account that feels uncontrollable is almost always one where nobody has deliberately set all four, so the system has been left to infer what you want from whatever was imported.
Asset groups are the structural lever. One group covering everything gives the system no way to distinguish your high-margin lines from your loss leaders, so it will find volume wherever it is cheapest. Splitting by margin, or by product category where margin varies with it, lets you set different targets for things with different worth.
Setting it up so it is steerable
- Feed the campaign one primary conversion with real values, not a list of equal-weighted actions.
- Split asset groups by margin or category, not by whatever the import produced.
- Exclude brand terms into a separate search campaign so you can see what you are paying for demand you already had.
- Supply every asset slot properly, including the long headlines and the video, or the system generates its own.
- Add placement and content exclusions, then review the placement report monthly.
Letting brand traffic hide inside it
Left unexcluded, brand searches get absorbed and reported as cheap conversions, which makes the campaign look excellent and makes it impossible to tell what it contributed. Pull brand into its own campaign first; every other measurement question about Performance Max becomes answerable afterwards.
Common questions
Can we see where ads appeared?
Partially, and more than at launch. The placement report is worth a monthly read even though it is incomplete.
How long before it stabilises?
Expect two to three weeks of learning after any significant change to signals, assets or budget. Judging it earlier means judging the learning phase.
Should we run search campaigns alongside it?
Usually yes, for brand and for the terms where you want explicit control. They serve different purposes.
Is it worth it for lead generation?
It can be, where the conversion signal reflects lead quality. Where it does not, it will find a great many worthless leads very efficiently.
