Performance Max: the six settings that decide whether it works

Performance Max takes away placement control, keyword control and most bidding control. That is not a reason to avoid it — it is a reason to be precise about the inputs it still accepts, because they are now doing all of the work.

1. Conversion values, not conversion counts

PMax optimises towards whatever you tell it is valuable. If every lead form submission is worth the same, the algorithm will happily buy you the cheapest, worst leads in the market. Pass real values — different values for a quote request versus a demo booking versus a purchase — and the same budget starts buying different traffic within days.

2. Feed quality, if you have a feed

• Titles that lead with what the product is, not with your brand name.
• Every optional attribute filled — category, GTIN, brand, size, colour. Each one is a targeting signal you are otherwise withholding.
• Custom labels for margin tier and stock depth, so you can split campaigns by profitability rather than by product type.

3. Asset group structure

One asset group per theme, not one per campaign. Each group should have its own headlines, images and landing page, because the asset group is the closest thing PMax has to an ad group. A single catch-all group means one creative pool competing against itself across unrelated queries.

4. Audience signals are a hint, not a target

Audience signals seed the algorithm; they do not constrain it. Give it your customer list and your high-intent site visitors, then expect it to move beyond them. If you need hard constraints, they have to come from exclusions and from campaign structure.

5. Exclusions you must set manually

• Brand exclusions, so PMax stops billing you for traffic your organic listing already wins.
• Account-level negative keyword lists applied to the campaign.
• Placement exclusions for mobile app inventory, which is where a lot of low-quality volume comes from.

6. Enough budget to exit the learning phase

PMax needs conversion volume to learn. Splitting a modest budget across four campaigns leaves all four in permanent learning. One campaign with enough daily budget to hit a meaningful number of conversions a week will outperform four starved ones every time.

Set these six deliberately and PMax is a strong channel. Leave them on defaults and it will spend the budget, report a respectable-looking cost per conversion, and quietly buy your own brand traffic back to you.