Google Performance Max asset groups: how to structure for 2026

In PMax, asset groups = sub-campaigns. Wrong structure = wasted budget. Here is the structure that works.

Rule 1: One asset group per product category

Do not dump all products in one group. Group by category (e.g., "Running shoes", "Hiking boots", "Casual sneakers") so PMaxlearns category-specific signals.

Rule 2: Audience signals are suggestions, not rules

Feed PMaxyour best-customer lookalikes, past purchasers, and competitor intent. PMaxtreats these as starting hints, not targeting.

Rule 3: 5+ videos per asset group

YouTube + Display need video. No video = worse performance. If you lack video, use the free video builder in Google Ads.

Rule 4: Minimum 20 headlines + 5 long headlines

More asset variations = more combinations = better auction coverage.

Rule 5: Segregate brand vs non-brand

Exclude brand keywords from PMax. Run brand in separate Search campaigns. Otherwise PMaxtakes credit for cheap brand conversions.

Frequently asked questions

Is this approach right for early-stage companies?

Most frameworks in this space assume a certain level of operational maturity, dedicated team members, established measurement infrastructure, some history of experimentation to build on. Pre-seed and seed-stage companies often lack these prerequisites and need a lighter-weight adaptation. For brands doing under $3M in annual revenue, focus on three or four of the principles that matter most for your specific business model rather than trying to implement the full framework at once. Rigor matters more than coverage at this stage.

How does this work for B2B versus B2C businesses?

The underlying principles around google pmaxasset groups apply across both contexts, but execution differs meaningfully. B2B google adstypically has longer sales cycles, multiple stakeholders per deal, and consideration periods measured in months rather than minutes. Measurement frameworks need longer windows. Attributionbecomes more complex. The same core strategic logic applies, but the tactical implementation looks different. We've worked extensively in both contexts and can flex the approach accordingly.

What changes when we integrate this with existing systems?

Every implementation requires integration work, systems don't exist in isolation. Analytics platforms, CRM, email systems, ad accounts, BI tooling all need to talk to each other for this to work at scale. Plan for 2-4 weeks of integration work at the start of any implementation. Shortcutting this phase creates data quality issues that compound and undermine the entire program over 6-12 months. We've seen teams skip integration work to move faster, only to spend 6 months later reconciling measurement discrepancies that could have been prevented upfront.

When should we reconsider the approach?

Every 6 months, run a structured review against the principles outlined here. Ask whether the market has shifted meaningfully, whether your business model has evolved, whether competitive dynamics have changed. Frameworks should evolve with context. A rigid commitment to any specific approach, including ours, eventually becomes the problem rather than the solution. The teams that outperform long-term are the ones that update their operating model based on evidence, not the ones that defend past decisions.