Everyone says shift budget to PMax. In lead-gen, I do the opposite.
Open almost any agency playbook and you will find the same line: PMax has the lowest CPA in the account, so move more budget there. It sounds obvious, the numbers seem to back it, and in a lead-gen account I think it’s mostly wrong.
Here is what that low CPA usually is. When I set up PMax in a lead-gen account, I run it as a layering tool, with all the primary Search keywords loaded as exact-match negatives so it can’t quietly eat my Search traffic. That means a big slice of what PMax converts on is warm and branded, people who were already coming, retargeting that was always going to close. So of course the cost-per-lead looks beautiful. It isn’t a smarter algorithm making a better decision. It’s cheaper traffic that the channel happened to sit on top of.
If you read that low CPA as “this campaign is winning, feed it,” you reward PMax for borrowing credit it didn’t earn. And it gets worse, because once PMax depends on a converting term, you can’t add that term as a negative in PMax and move it to the relevant Search campaign without starving the campaign you just decided to scale. You end up obscuring your own Search performance and feeding a black box at the same time.
PMax in lead-gen is a discovery tool, not a budget destination.
So when PMax converts on a search term that should have been an exact-match keyword in a Search campaign, I don’t celebrate the cheap lead. I harvest it. The cycle is dull and it works: pull PMax’s converting search terms, cross-check each one against the live Search keyword lists, and for anything not already covered, add it as exact-match in the right Search ad group, then add it as an exact-negative back in PMax. The term moves to the channel where I can actually see it, bid on it, and write to it. PMax goes back to finding the next one.
Only after that harvest do I let myself ask the budget question. Is PMax still driving genuinely incremental leads, or was it living off terms I’ve now pulled into Search? If it’s still incremental, a modest bump to feed the discovery is fair, and I’ll frame it exactly that way. If it isn’t, the budget stays flat. “Scale the best-performing campaign” is not a reason on its own. It’s the line that ignores the whole architecture.
There’s a second job PMax quietly does well here, and it’s the one people mistake for waste. In most lead-gen accounts I tightly exclude the fence-sitters from Search: the people comparing rental instead of buying, the ones circling a subscription or a marketplace listing, the concurrent-curious. On Search that intent is a money leak, so it goes on shared negative lists. But those lists are coupled to the active Search campaigns only, never to PMax. PMax, with its mixed channels and lower CPC, converts a real share of those mixed-intent users at an acceptable cost. So when I open a PMax campaign and see a wave of comparison or rental queries, that isn’t the campaign misfiring. That’s the campaign doing the exact job I built it for.
One caveat, because the principle is specific to lead-gen. In e-commerce none of this applies. If the feed is solid, PMax-shopping replacing or augmenting standard Shopping is completely fine, and there the “more budget” instinct is often right. The layering logic only holds where leads, not a product feed, are the thing you’re buying.
So no, I’m not moving budget to PMax because the dashboard says it’s cheapest. I’m reading where that cheapness comes from first. After twenty years in this, the campaign with the prettiest CPA is usually the one asking the fewest hard questions.