Performance Max gets two extreme treatments. Google sells it as the future of paid media - a single, AI-driven campaign that handles everything. Critics dismiss it as a black box that cannibalises everything around it. Both takes are wrong, but in different ways.

We've now run PMax across enough accounts - DTC, ecommerce, B2B, lead gen, regulated sectors - to have a clear view of when it works, when it doesn't, and what the actual playbook looks like. This is that playbook.

Where PMax actually adds value

• Inventory you can't easily reach with Search. YouTube, Discover, Gmail, the Display Network, Maps. PMax is the easiest way to get incremental impressions on these surfaces with reasonable measurement.
• Shopping at scale. For ecommerce accounts with healthy feeds, PMax usually outperforms standard Shopping for non-brand queries. Asset groups can carry multiple product categories cleanly.
• Audience signal that's hard to use elsewhere. Audience signals - first-party data, custom segments, lookalikes - feed PMax in ways that take meaningful effort to replicate in Search.

Where it absolutely doesn't

• As a replacement for branded Search. If you let PMax run on brand traffic, it will eat it and report it as PMax success. Always. Without exception. This is rule one.
• As your only campaign. PMax is bad at exploration. Without a Search campaign feeding it real intent signals to learn from, it tends to optimise for the easiest conversions it can find - usually retargeting and brand-adjacent.
• For regulated lead generation without offline conversions. The "conversions" PMax learns on need to be qualified leads, not raw form fills. Without offline conversion uploads, PMax will cheerfully scale low-quality lead volume.

The four levers that actually matter

1. Brand exclusion

Add the account-level brand exclusion list to every PMax campaign on day one. Without this, PMax will harvest brand queries and credit itself for the conversions. The dashboard will look great. The incremental contribution will be close to zero.

If you've inherited an account where this hasn't been done, expect the reported PMax ROAS to drop by 40-70% once brand is excluded. That drop isn't PMax getting worse - it's PMax finally being measured against what it's actually contributing.

2. Asset group structure

The biggest mistake we see is one PMax campaign with one asset group covering all products. PMax can't optimise creative or audience signals across wildly different products in one bucket.

The structure that works:

• One PMax campaign per major business unit or product category. If you sell shoes and accessories, that's two campaigns minimum.
• Within each campaign, asset groups by customer type or product subtype. Don't go nuts - usually 2-4 asset groups per campaign is the sweet spot.
• Audience signals fed in deliberately. Customer match, website visitors, custom segments by intent. Without signals, PMax is essentially guessing for the first 4-6 weeks.

3. Feed quality (for ecommerce)

For Shopping-led PMax, the Merchant Center feed is more important than anything happening in Google Ads. Title structure, attribute completeness, custom labels for margin tiers, the GTIN, the brand attribute, the Google product category - all of it influences which queries PMax bids on and how aggressively.

A two-week feed-health audit before any PMax restructuring usually returns more value than the restructuring itself.

4. The campaigns running alongside it

PMax learns best when it runs alongside well-structured Search campaigns covering brand and high-intent generic. Search gives PMax real signals to learn from; PMax extends reach to the surfaces Search can't easily cover. Together they work. Apart they don't.

The structure we typically end up with:

1. Brand Search campaign - tight, low CPA, isolated from PMax via exclusion list.
2. Direct generic Search - high-intent queries by product category, Target ROAS bidding.
3. Discovery generic Search - exploration queries on Maximise Conversions.
4. PMax for non-brand reach - by product category, with audience signals.
5. Standard Shopping (where the feed and margin justifies it) at higher priority than PMax.

What "success" actually looks like

Once a PMax campaign is set up properly, the realistic expectation is steady contribution to non-brand revenue or qualified leads, with reach into surfaces (YouTube, Discover, Display) that would otherwise need separate work.

It's not "fire and forget". It's not "AI replaces account management". The successful PMax accounts we run still have weekly check-ins on asset group performance, fortnightly audience signal updates, monthly creative refreshes, and constant attention to whether PMax is competing with Search rather than complementing it.

What to do this week

• Check whether your PMax campaigns have the brand exclusion list applied. If not, add it.
• Look at the placement report. If PMax is spending heavily on YouTube placements that look like obvious junk, add them as exclusions.
• If PMax is your only campaign type, build out a basic Search structure alongside. Brand, generic, and an exploration campaign at minimum.
• Audit your Merchant Center feed quality if you're running Shopping. Title structure and custom labels are usually where the leverage is.

If you want a clear read on whether your PMax setup is actually contributing - or quietly cannibalising the rest of the account - that's a typical strand of the strategic review. We'll look at brand exclusion, asset group structure, feed health, and the campaigns running alongside.