The PMax problem.
Performance Max is the fastest way to waste $50,000 on Google Ads. Google will tell you the opposite. Both can be true.
PMax — Google's black-box campaign type that runs across Search, Display, YouTube, Discover, Gmail, and Maps from a single asset group — is now the default recommendation for almost every account. Rep calls, automated emails, the interface itself — everything pushes you toward PMax.
And it does work. Sometimes. For certain accounts. When set up by someone who knows what they're doing. Which is the problem: the pitch is "just hand Google the creative and let the algorithm figure it out," but the reality is that the algorithm figures out how to spend your money quickly, not how to spend it wisely.
What PMax actually does
Under the hood, PMax is an arbitrage engine. You hand it creative assets, a target (usually conversions), and a budget. It then goes looking for the cheapest impressions across Google's entire inventory that it can plausibly attribute a conversion to.
That last part is where it gets ugly. "Plausibly attribute" is doing a lot of heavy lifting.
Because PMax decides for itself what "conversion" means inside the campaign — based on your account's historical data — it will spend heavily on branded search, remarketing lists, and already-captured demand. Then it'll report a 6:1 ROAS back to you. And the ROAS is real, in the sense that the conversions happened. But most of them would have happened anyway.
PMax is very good at taking credit for conversions your other efforts already earned.
The autopilot trap
Here's the pattern I see in most audits. A founder or marketing lead launches PMax. It spends $10k in a month. It reports $60k in attributed revenue. The ROAS looks great. The team cheers. They double the budget.
The next month, the campaign spends $20k. Attributed revenue goes to $90k. But total company revenue didn't move proportionally. Organic traffic conversions are down. Direct traffic conversions are down. Branded search — which was "free" — is now being bought.
That's the autopilot trap. PMax didn't create new demand. It cannibalized existing demand, bought your own brand traffic, and slapped its name on every conversion that walked in the door.
The founder has no way to see this from inside the PMax reporting interface, because PMax specifically hides the breakdown by channel and placement.
Five guardrails to keep a human in the loop
None of this means PMax is useless. It means PMax needs to be steered. Here are the five things every account I run PMax on has in place.
1. Tight asset group boundaries
One PMax campaign per product line or buyer persona. Not one global PMax campaign trying to do everything. Separating the asset groups forces the algorithm to learn within a smaller signal space, which is more controllable.
2. Aggressive negative keywords
Yes, PMax supports negative keywords now. Most agencies still don't use them. Add your brand as a negative. Add competitor brand names as negatives. Add your existing customers' company names as negatives. Force PMax to go find new demand.
3. Watch branded search cannibalization
Set up a script or dashboard that tracks your branded search CTR from PMax separately from your non-brand search campaigns. If PMax is eating into your branded traffic, you're paying for impressions that would have converted anyway.
4. Custom labels for real attribution
Tag every conversion action with a custom label so you can measure PMax performance against your actual business goals — not Google's definition of conversion. New customer acquisition is not the same as repeat purchase is not the same as a form fill from an existing lead.
5. Account-level alerts on CPA drift
Set alerts for week-over-week CPA increase. PMax will tell you "it's learning" for months. An alert tells you when the learning stops learning and starts losing money.
When PMax actually works
PMax works well when: you have strong conversion signal from your pixel already, your asset library is deep (15+ images, 5+ videos, multiple ad copy variants), you're running it alongside proper Search campaigns that are pre-capturing branded demand, and you have someone actually watching the account weekly.
PMax does not work when: you set it up and forget about it. Which, again, is how most accounts run it.
The accountability test
Here's the test every PMax campaign should pass: if you turned it off for a month, would your overall revenue drop by the amount PMax claims to be generating?
Most teams never test this. The ones that do are usually shocked at the answer. Try it. Pause PMax for two weeks. Watch what happens to your total company revenue, not your attributed campaign revenue. That's the real number.
If your PMax campaigns feel like they're "working" but you can't explain exactly why, you probably need a second set of eyes on the account. Book a free Google Ads audit and I'll walk through your account with you in 30 minutes. No pitch, no sales. Just the real numbers.
