Performance Max for lead gen: how to stop paying for junk leads.

Performance Max is the most powerful campaign type Google has ever shipped, and for lead gen accounts it's also the easiest way I know to burn a budget without noticing for weeks. The mechanism is simple to describe and hard to feel the consequences of in real time. PMax automates targeting and placement decisions across every inventory type Google owns at once: Search, Display, YouTube, Discover, Gmail, Maps. You hand it a budget and a conversion goal, and it decides where your ads show, who sees them, and how much to bid, with almost none of the visibility a standard Search campaign gives you.

That would be manageable if PMax optimised the way older campaign types did, gently, with room to course-correct. It doesn't. PMax optimises ruthlessly and literally toward whatever conversion signal you feed it. Give it a raw form fill as the goal and it will go find the cheapest form fills available anywhere in Google's inventory. Often that means a Discover placement served to someone scrolling their phone on a commute, or a YouTube pre-roll viewer who clicks out of habit, or a Gmail promo tab impression from someone who was never in-market at all. None of them look wrong in the interface. They show up as conversions, cost per lead drops, and the account looks like it's performing. It isn't. It's just gotten very good at making your cheapest possible action happen, and for most lead gen businesses the cheapest possible form fill and the most valuable one are not the same person.

PMax doesn't find good leads or bad leads. It finds whatever conversion event you told it to find, at the lowest possible cost, from wherever on the internet it can get it.

What makes this harder to catch than it should be is that the account dashboards were never designed to warn you. Cost per lead trending down and conversion volume trending up both read as wins in every reporting view Google gives you by default. Nobody gets an alert saying "your leads got worse while your numbers got better." You only find out three or four weeks later, when the sales team starts asking why the pipeline looks fuller than the calendar of actual booked calls, or when a client on a monthly retainer starts questioning why cost per lead is at an all-time low while revenue sits flat. By the time that conversation happens, the account has usually been running on the wrong signal for long enough that the machine learning models have locked in on exactly the wrong pattern, and unwinding it takes longer than setting it up correctly would have in the first place.

I've seen this exact pattern before in a completely different vertical: self-storage operators who feed Google raw "get a quote" clicks instead of actual move-ins, and watch the algorithm chase quote volume from people who were never going to rent a unit. The fix in both cases is the same principle. The algorithm isn't broken. It's doing precisely what it was told, and the fix is in what you tell it, not in fighting the automation.

Feed it a qualified conversion action, not a raw form fill

The single highest-leverage change you can make is upstream of the campaign entirely: what event PMax is told to optimise toward. A raw form-fill pixel firing the moment someone hits submit gives the algorithm no way to distinguish a genuine buyer from a student filling in fake details for a school project. Replace it with a qualified conversion action, one that only fires after a lead clears a basic quality bar: a CRM status change, a phone number that passes validation, a lead score threshold, a sales-accepted flag pushed back from your CRM via offline conversion import. If your leads aren't binary but come in tiers of value, use conversion value rules to weight them, so a lead from a high-intent segment is worth meaningfully more to the bidding algorithm than a low-intent one, and PMax optimises toward value rather than raw count. This is the same discipline that sits behind everything we build under server-side conversion tracking: the algorithm can only be as smart as the signal you're willing to give it, and a flat, unweighted form-fill event is the least informative signal you can hand a system this aggressive.

Use account-level negative keywords to close the obvious gaps

Since 2024, Google has allowed account-level negative keyword lists that apply across Performance Max and Search together. It's a blunter instrument than campaign-level negatives, and it won't give you the granularity a tightly managed Search campaign offers, but it's the main lever you have for keeping PMax out of territory you already know is wrong. If you run a B2B service and PMax keeps surfacing against "free," "jobs," "DIY," or your own brand terms in a way that cannibalises organic traffic, put those on the account-level list. Build this list from your search term insights, not from guesswork, and revisit it every time you spot a new pattern of waste. It's not a substitute for a qualified conversion signal, it's a second line of defence for the queries that shouldn't be entering the auction at all.

Treat audience signals as guidance, not hard targeting

Performance Max lets you feed it audience signals, custom segments, remarketing lists, customer match data, to give the algorithm a starting point. It's tempting to treat these the way you'd treat targeting in a manual campaign, as a fence the system stays inside. It isn't one. PMax uses audience signals as a hint about where to start looking, then expands well beyond them the moment it finds cheaper conversions elsewhere, because expansion is the entire premise of the campaign type. Build your signals as tightly and accurately as you can, upload real customer match lists, use your highest-intent remarketing audiences, but don't assume the presence of a signal means the campaign is restricted to it. It isn't, and the gap between what you configured and what the system actually did is exactly where budget disappears unnoticed.

Review placement and search term insights on a weekly cadence

PMax gives you less visibility than a standard Search campaign, but it doesn't give you zero. The placement report and the search term insights report are the two places where the black box cracks open enough to see what's actually happening, and they're worth a fixed weekly slot rather than an occasional glance when performance dips. Look for placements that are structurally wrong for your business: content mismatches, apps with no plausible connection to your audience, YouTube channels aimed at a different demographic entirely. Look for search terms that technically match your keywords but signal the wrong intent. Catching these early, before they compound into a spend pattern, is the difference between a PMax campaign that stays efficient and one that quietly drifts for a quarter before anyone checks the numbers. Automation earns trust through review, not through being left alone.

Know when to pull back to Search

Sometimes the honest answer is that a lead-gen account isn't ready for Performance Max yet. If you genuinely can't supply a qualified conversion signal, because the sales cycle is too long, the CRM integration isn't built, or you simply don't have enough historical data to define what a good lead looks like, PMax has nothing better to optimise toward than the raw form fill you're trying to avoid feeding it. In that situation a well-structured standard Search campaign, with tight keyword match types, proper negative lists, and manual or portfolio bidding you can actually inspect, will usually outperform PMax until the tracking catches up. There's no prize for running the newest campaign type on an account that isn't instrumented for it. Get the qualified signal built first, prove it out on Search where you can see every lever, then hand the algorithm something worth chasing.

Most of the PMax accounts I audit are burning money in exactly the ways described above, and it's rarely obvious from inside the interface because every number the platform shows you looks like progress. If you want a fast read on where your own account stands, run it through the PPC Waste Finder for a free estimate of what's leaking. And if the fix needs to go deeper than an audit, that's the conversation we have on the PPC service page.

Filed under: PPC · Google Ads · Performance Max · 2026