What Performance Max Is (and Why It's Misunderstood)

Performance Max (PMax) is Google's AI-first campaign type that serves ads across all Google inventory — Search, Shopping, Display, YouTube, Gmail, and Maps — from a single campaign. It uses Google's automation to optimise bids, placements, and creative combinations toward your conversion goal.

Launched in late 2021 and made the mandatory replacement for Smart Shopping in 2022, PMax has a polarising reputation. Advocates point to genuine ROAS improvements; critics point to the black-box nature and difficulty attributing what's actually driving results. Both are correct. The answer is knowing how to set it up and interpret it.

The core tension: PMax gives Google's algorithm maximum freedom to find conversions anywhere in its network. This is powerful when your conversion data is rich and your creative is strong — and frustrating when either is thin. The setup decisions you make at the start determine which experience you get.

Asset Groups: The Architecture That Determines Performance

Asset groups are the fundamental building block of PMax. Each asset group contains your creative assets (headlines, descriptions, images, videos, sitelinks) and the audience signals that tell Google who to target. Google then mixes and matches these assets across placements.

Recommended asset group structure for e-commerce:

  • By product category: Separate asset groups for each major product category (e.g., "Men's Footwear", "Women's Apparel"). This lets you use category-specific headlines, images, and URLs — critical for relevance.
  • By audience temperature: Some advertisers separate brand/remarketing audiences from prospecting — though Google now handles this more automatically with PMax's customer value goals.
  • By margin tier: If some products have significantly higher margins, separate asset groups allow different target ROAS settings per group.

Minimum assets per asset group (aim for more): 15 headlines, 5 descriptions, 3 landscape images, 3 square images, 1 portrait image, 1 video (use auto-generated video if none available, but create a proper one — auto-generated video consistently underperforms).

Audience Signals: Guide the Algorithm Early

PMax audience signals tell Google's algorithm where to start looking. They're not hard targeting constraints — Google will expand beyond them — but they're critical for helping the algorithm find converting audiences faster, particularly in new campaigns.

Effective audience signals to add:

  • Customer match list (existing customers from CRM email export) — the highest-signal audience Google can use
  • Website visitors (all visitors, purchasers, cart abandoners) — historical intent signals
  • Competitor keyword themes — signals the algorithm toward users researching your category
  • In-market audiences relevant to your product category
Don't leave audience signals empty: A PMax campaign with no audience signals starts completely cold — Google learns from scratch using your conversions alone. This can mean 2–4 weeks of suboptimal spend while the algorithm calibrates.

Search Themes and Brand Exclusions

Google added Search Themes to PMax in 2023 — a critical feature that lets you specify keywords you want the campaign to appear for in search results. Add 10–15 relevant search themes per asset group to ensure your search coverage includes your priority queries.

Brand exclusion is essential: By default, PMax competes with your brand search campaigns and can take credit for brand conversions that would have happened anyway. Apply a brand negative keyword list to your PMax campaign to prevent this. In Google Ads: Campaigns → PMax Campaign → Settings → Brand exclusions.

Without brand exclusions, PMax ROAS figures are often inflated by brand conversions — making it appear more efficient than it actually is for incremental customer acquisition.

The Reporting Problem and How to Work Around It

PMax's biggest legitimate criticism is its limited reporting. You can't see which placements, keywords, or audience segments are driving conversions — Google surfaces only high-level asset performance data.

Reporting workarounds:

  • Search terms insights: PMax does surface aggregated search category insights (not individual terms). Review these weekly in Insights → Search Terms to understand what queries are triggering your ads.
  • Asset performance ratings: Assets are rated Best/Good/Low. Systematically replace Low-rated assets — this is the primary optimisation lever within the campaign.
  • Auction insights: Shows which competitors your ads appear alongside — useful for competitive intelligence.
  • Segment by network: In the campaign view, segment by "Network" to see the split between Search, Shopping, Display, and YouTube. A healthy PMax typically drives 50–70% of conversions from Shopping and Search.

The incrementality question: The hardest PMax problem is determining how much of its attributed revenue is incremental vs. would have happened anyway through other campaigns. Run PMax holdout tests (remove it from 10–20% of your audience for 4 weeks) to measure true incrementality before committing major budget.