An honest assessment of Performance Max from an agency that runs it daily: where it genuinely earns its budget, where it wastes it, and the guardrails that decide which.
Performance Max is neither the scam its critics describe nor the autopilot Google sells. We run it across client accounts every day, and the honest summary is this: it is a powerful system that behaves exactly as well as the inputs and guardrails you give it, and the accounts where it disappoints are usually the accounts where it was switched on and walked away from.
Where it genuinely works
Ecommerce with a strong product feed is the clearest case. Given clean feed data, real conversion values and enough volume for the algorithm to learn, PMax reaches across Shopping, YouTube, Display and Search in ways manual campaigns cannot, and it finds converting inventory a human buyer would not have tested. It also works when your conversion signal genuinely reflects margin, because the machine optimises whatever number you feed it, for better or worse.
Where it wastes budget
The failure cases are just as consistent. Lead generation businesses feeding it a soft conversion, such as a form fill with no quality signal, teach it to find form fillers rather than customers. Accounts without brand exclusions let it quietly harvest branded demand and present the result as incrementality. Low budgets spread across its channel mix never give the learning system enough signal. And businesses whose value arrives offline, weeks later, are invisible to a system optimising on what it can see today unless offline conversions are imported properly.
The guardrails that decide the outcome
Four controls separate the good deployments from the expensive ones. Brand exclusions, so the campaign has to earn new demand rather than repackage existing demand. Conversion values that reflect margin or lead quality, imported from your CRM where the truth lives. Asset group structure that mirrors real product economics rather than the website menu. And search term visibility through the reporting Google now exposes, reviewed on a schedule, because waste in PMax hides in aggregate numbers. These are the same disciplines described in our Google Ads audit checklist, applied to a campaign type that makes them harder to skip.
The honest test
The only question that matters is incrementality: is PMax finding revenue you would not otherwise have had, or repackaging revenue you already owned? Branded uplift studies and holdout tests answer it; the campaign's own dashboard never will. We run those tests as part of performance and incrementality testing, and the results decide whether PMax deserves more budget, tighter guardrails or retirement from the account.
If PMax is a growing share of your spend and you could not say with confidence which of those three it deserves, ask us for a second opinion. It is one of the fastest questions in paid media to answer properly, and one of the most expensive to leave unanswered.
