Blog · Google Ads
Google Ads

Performance Max, without handing Google a blank cheque

Prem Jasuja By Prem Jasuja, Founder · 8 min read · August 2026

Performance Max is the campaign type Google pushes hardest, and for good reason: it will spend every dollar you give it. Left on defaults, a PMax campaign is a black box that chases the cheapest clicks across Search, Display, YouTube, Gmail, Maps and Discover, then reports a healthy-looking cost per conversion that quietly includes traffic you would have won for free. The campaign is not broken. It is doing exactly what you told it to do, which is nothing in particular.

Used well, Performance Max is genuinely powerful. It finds pockets of demand a manual account never would, and it does it at scale. The difference between a PMax campaign that prints revenue and one that burns budget is not luck or budget size. It is guardrails. Below is the setup we use to keep PMax honest, and the checks that tell you within two weeks whether it is working.

Performance Max does not know what a good customer looks like. Your job is to teach it.

First, understand what PMax actually optimises for

Performance Max bids toward whatever conversion action you feed it, using Google's automated bidding across all inventory at once. That is the whole model. If you feed it a weak signal, page views, form-opens, newsletter sign-ups, it will find thousands of cheap versions of that weak signal and call it success. This is the same trap that drains manual accounts, and we wrote about it in why your ad report is lying to you. With PMax the stakes are higher, because you have handed the algorithm the keys to six placements instead of one.

So the single most important decision happens before you build a single asset: what counts as a conversion, and what is it worth. Everything else is downstream of that.

The five guardrails that make PMax profitable

1

Brand exclusions, or you pay to win your own name

By default, Performance Max will happily serve on searches for your own brand, then take credit for conversions you would have gotten organically or through a cheap brand Search campaign. It inflates your reported ROAS and hides how the campaign performs on real prospecting. This is the number one reason PMax looks amazing in month one and mediocre by month three.

How to fix it: request access to brand exclusions from your Google rep, or add brand terms to an account-level negative list, and run a separate exact-match brand Search campaign so you can see the true cost of new demand. Keep brand and non-brand spend in different buckets. Then judge PMax on prospecting alone.
2

Conversion values, not just conversion counts

If every lead is worth the same in Google's eyes, the algorithm will chase the easiest, cheapest lead every time. A demo request from an enterprise buyer and a tyre-kicker downloading a free guide should not carry the same weight. When you attach values, PMax starts optimising toward profit instead of volume.

How to fix it: use value-based bidding. Assign different values to different actions (a booked call is worth more than an email capture), and where you can, import real revenue back from your CRM so the model learns which clicks actually turn into money. Target ROAS beats target CPA for any business where deal sizes vary.
3

Tight asset groups built around themes, not everything in one bucket

An asset group is PMax's version of an ad group. Dumping every product, headline and image into a single asset group forces Google to average across audiences that have nothing in common. You lose the ability to see what works and the model loses the ability to match the right message to the right person.

How to fix it: build one asset group per distinct offer, audience or theme. Give each a full set of strong headlines, descriptions, images and video (if you skip video, Google auto-generates a weak one for you). Use search themes to point each group at the intent you actually want.
4

Audience signals that guide, not guess

Audience signals are hints, not targeting. PMax uses them as a starting point and then expands. Give it a vague signal and it wanders. Give it your best-customer data and it has a real head start on finding lookalikes who convert.

How to fix it: feed it your customer match lists (actual buyers, not all sign-ups), your high-intent website visitors, and tightly defined in-market and custom segments. The better your first-party data, the faster PMax finds people who look like your best accounts. Refresh these lists monthly.
5

A product or lead feed you actually maintain

For ecommerce, the Merchant Center feed is the engine of Performance Max, and a stale feed with missing attributes, bad titles or wrong prices quietly caps your performance. For lead-gen, the equivalent is a clean set of assets and a landing experience that matches each theme. Google can only work with what you give it.

How to fix it: keep titles keyword-rich and accurate, fill every attribute Google offers, fix disapprovals the day they appear, and make sure the campaign points to pages built to convert, not your homepage. A great feed into a weak page still loses.

How to launch PMax without lighting money on fire

New campaigns need a learning period, and PMax is hungrier than most. Starve it and it never exits learning; over-feed it before tracking is solid and it learns the wrong lesson. Here is the sequence that keeps the first month sane.

The launch checklist

  • Confirm conversion tracking fires on a real business event, and reconcile it against your CRM before you spend a rupee.
  • Turn on brand exclusions and split brand into its own campaign.
  • Start on a target CPA or ROAS you can actually afford, then loosen it once data comes in, do not start loose.
  • Build at least two or three themed asset groups, each with full assets including video.
  • Budget enough to clear roughly 30 conversions in the first month, or the algorithm never learns.
  • Leave it alone for two weeks. Editing daily resets learning and wastes the spend you already paid for.

Reading the results honestly

Google's PMax reporting is deliberately thin, but you can see more than the dashboard suggests. Run the campaign-level and asset-group insights, check the search categories your traffic is coming from, and pull the placement report through a script or your rep if you suspect junk Display and app inventory. If most of your conversions trace back to brand or to a handful of low-intent placements, the campaign is not prospecting, it is harvesting demand you already had.

The deeper problem is that Performance Max can mask the same leaks that drain any account, cheap clicks, soft conversions, unmaintained feeds, just behind a prettier interface. If you want the full list of where paid budget disappears, we broke it down in where 40% of your ad budget quietly disappears. PMax makes those leaks easier to hide, which is exactly why the guardrails matter more here, not less.

Run Performance Max the lazy way and it is a spend machine. Run it with brand protection, real conversion values, themed assets, sharp audience signals and a maintained feed, and it becomes one of the most efficient ways to buy new customers at scale. The tooling is the same. The result is night and day.

Is PMax working, or just spending?

Book a free audit and we will pull back the black box, show you where your Performance Max budget is really going, and tell you the two changes that will move it most.

Book your free ad audit

Keep reading