Notes
Performance Max won't tell you if it's prospecting or harvesting
Run Performance Max as a prospecting campaign and you expect it to bring you new customers. Often it spends most of the budget somewhere else: on people already searching your brand, already on your remarketing lists, already in market and close to buying. That is not wasted money. Brand and existing-customer demand is some of the cheapest, highest-converting traffic you can buy. The catch is that PMax does not show you how much of its result came from finding new customers and how much came from harvesting demand you already had, so it is easy to read one as the other.
What Performance Max is actually optimizing
PMax is not a YouTube campaign, or a Shopping campaign, or a Display campaign. It is one campaign that buys across all of them at once: Search, Shopping, YouTube, Display, Discover, Gmail, and Maps. You hand it a goal, usually a conversion-value target with a ROAS attached, along with your creative and assets, and it spends wherever it can find the cheapest path to that goal across every inventory Google owns.
Smart Bidding takes that instruction literally, and the cheapest path to a conversion is a conversion that was already close to happening.
The cheapest path is the demand you already have
Someone typing your brand name. A past customer on your remarketing list. A person deep in market who recognizes your offer on sight. PMax finds that demand first and books it as its own result, because it is the easiest demand to win.
None of that is bad to buy. A brand searcher is not a guaranteed sale: a competitor may be bidding on your name, your organic listing may sit below the fold, and a paid ad puts that person on the page where they are most likely to convert. Capturing existing demand is often your best-converting spend. The problem is only what you conclude from it.
The wall when you scale
The trap shows up when you scale. A campaign that is mostly harvesting looks ready for more budget, so you give it more and wait for new customers to follow. They don’t. There is only so much existing demand to collect, and once you have the people already searching your brand and already on your remarketing lists, the extra budget has nothing left to convert. That is the wall people hit scaling PMax, and it comes from reading harvesting as prospecting. To harvest more later you have to prospect more first: bring relevant traffic in through Search, and build demand higher up the funnel with brand-oriented ads, so there is a bigger pool to collect from down the line. Scale the harvest without refilling the pool and the campaign stalls while the ROAS still reads fine, which is why the number never warns you.
What Google leaves switched on
Google pushes PMax hard and will not tell you to constrain it. Out of the box, URL expansion, content and placement settings, and broad inventory all widen where it spends. Turn URL expansion and content settings off and cap your placements, so the campaign works on the inventory you chose rather than wherever it finds the cheapest conversion.
The fix
Make PMax prove it can prospect. Add your brand terms as an account-level brand exclusion so it can no longer count brand searches as its own wins, and run brand defense as its own campaign where you set the budget and read the result directly. That single change turns a flattering blended number into an honest one, and everything else builds on it.
Then keep three things running:
- A prospecting base in Search and Shopping. PMax harvests the demand those campaigns and your brand-building create, so the base is what refills the pool. Starve it and PMax has nothing left to scale into.
- A benchmark against that base. Pit PMax against Search and Shopping and watch for cannibalization. If it is re-booking conversions Search would have won on its own, its standalone ROAS is fiction. Keep PMax only where it wins net of what it takes from everything else.
- Offline conversion and revenue data fed back in. Send Google what actually closed and what it was worth, so it optimizes toward new, profitable customers instead of raw conversion count.
Set up this way, the question stops being “what is my PMax ROAS” and becomes “what did PMax add on top of everything I already run.” That second number is the only one that tells you whether you are growing or paying to collect demand you already had.