Performance Max in 2026: how to actually control it.

Performance Max is now the default campaign type for most Google Ads accounts. Google positions it as the future of the platform, and to a large extent it is. It is also the campaign type advertisers understand least, because Google deliberately hides most of what it is doing. The reporting is thin, the placements are opaque, and the official guidance is more or less to trust the machine.

We do not trust the machine on faith. PMax is a powerful tool that, left ungoverned, optimizes toward whatever is easiest rather than whatever is most profitable. The most common failure pattern is simple: PMax quietly absorbs your branded search and your existing retargeting audiences, reports a beautiful ROAS, and claims credit for revenue that would have arrived anyway.

You have more control than Google implies

The framing that PMax is uncontrollable is convenient for Google and wrong for advertisers. The levers exist. They are just not surfaced in the place you would expect.

• →Account-level and campaign-level brand exclusions, so PMax stops cannibalising branded demand
• →Account-level negative keyword lists applied to PMax (now supported), to cut obviously irrelevant traffic
• →Asset group structure built around margin and theme, not one undifferentiated mega-group
• →Search themes and audience signals used as direction, not as hard targeting
• →Feed-only versus feed-plus-assets decisions made deliberately per product set
• →Geographic and language settings audited, not inherited from a template

Separate brand from Performance Max

The single highest-leverage thing you can do with PMax is stop it from eating your brand. Branded searches convert at very high rates because the intent is already there. When PMax is allowed to serve against them, it reports inflated efficiency and the algorithm leans further into the easy win. The result is a campaign that looks like your best performer while doing very little incremental work.

Make it report what it is doing

PMax reporting is deliberately shallow, but you can force more visibility. Search term insights, channel and placement breakdowns, and asset group level performance are all retrievable with a bit of work. The discipline is to look at them weekly and act on them, rather than accepting the headline ROAS number the campaign reports to you.

Asset group architecture

Treating a PMax campaign as one giant asset group is the structural equivalent of pointing it at your whole catalogue and hoping. We build asset groups around margin tiers and product themes so that budget can be steered toward the products that actually deserve it. High-margin hero products do not belong in the same asset group as clearance stock, because they do not deserve the same bidding behavior.

What good Performance Max governance looks like

• →Brand excluded and verified before launch
• →Asset groups segmented by margin and theme
• →Search term and placement reports reviewed weekly
• →Negative lists maintained, not set once and forgotten
• →ROAS read against incremental contribution, not platform-reported credit

Done this way, PMax becomes what it should be: a powerful demand-capture engine you direct, rather than a black box you hope for. When we scaled Awesome Books across millions of titles and multiple regions, disciplined PMax governance was a large part of how we grew sales while holding strict ROAS and CPA targets. The campaign type did not change. The way it was controlled did.