Three ways Performance Max quietly wastes your budget

Performance Max is the campaign type clients most often ask for by name and most often run badly. It’s not a bad product. It’s a product that punishes vague inputs, and most accounts feed it vague inputs.

Three failure modes account for nearly every disappointing PMax account I’ve inherited.

Failure one: letting it cannibalise brand search

Left alone, Performance Max will happily spend your budget buying clicks from people searching your own company name — traffic you were going to get for free, or for pennies, from an existing brand campaign.

The reported ROAS looks spectacular, because you’re paying to intercept people who had already decided. Then you look at total new customer volume and nothing has moved.

The fix: apply brand exclusions at the campaign level, and keep a dedicated exact-match brand search campaign running alongside. Then judge PMax on incremental conversions, not blended ROAS.

Failure two: one asset group for everything

An asset group is the unit of relevance in PMax. Cramming every product line into one means the system averages your creative, your audience signals and your landing pages into mush.

Split asset groups the way a customer would split them — by the problem they’re trying to solve, not by your internal org chart. Each group gets its own headlines, its own images, its own audience signal and its own destination URL.

Failure three: treating conversions as one bucket

If a newsletter signup and a $12,000 contract both count as “1 conversion”, the algorithm will optimise enthusiastically toward newsletter signups. They’re cheaper and there are more of them. It’s doing exactly what you asked.

The fix is conversion values. Assign a real number to each action, even a rough one derived from close rate and average deal size. A lead worth $400 in expected value and a signup worth $5 should not be interchangeable in the bidding.

The reporting problem nobody solves for you

Performance Max reporting is deliberately thin. You can’t see placement-level detail the way you can in a standard display campaign, and search term visibility is limited to category-level insight.

Build the accountability outside the platform: server-side conversion tracking, offline conversion imports from the CRM once deals close, and a Looker Studio view that compares PMax periods against a holdout. If you can’t run a clean holdout, at least compare year-over-year at the channel level before and after launch.

When it genuinely works

Performance Max earns its place when you have real conversion volume, accurate values flowing back in, and enough creative to fill several distinct asset groups. Under those conditions it is very good at finding demand you weren’t targeting manually.

Below roughly thirty conversions a month, you’re asking a machine-learning system to learn from noise. Spend that budget on search and fix your tracking first.