Apply
Performance Max

Performance Max for eCommerce: how to set it up and actually make it work

Nik Armenis · October 2026 · 11 min read · ← Back to Resources

Performance Max is Google's most powerful campaign type for eCommerce — and the most misunderstood. Brands either avoid it because they've heard bad things, or they set it up carelessly and get bad results that confirm the bad things they heard.

Done properly, PMax is the highest-leverage channel in most eCommerce Google Ads accounts. This guide covers everything you need to know to set it up and make it work.

What PMax actually is (and isn't)

Performance Max is a goal-based campaign type that accesses all of Google's inventory — Search, Shopping, Display, YouTube, Gmail, and Discover — from a single campaign. Google's algorithm decides where to show your ads and how to allocate budget across inventory types, based on your conversion data and the signals you provide.

It's not a magic button. It's an algorithm with inputs and outputs. The quality of those inputs determines the quality of those outputs.

Asset groups: the structural unit of PMax

Asset groups are the building blocks of a PMax campaign. Each asset group contains the creative inputs (headlines, descriptions, images, videos, logos) and the products (from your feed) that Google uses to generate ads.

How many asset groups do you need?

More than one. One catch-all asset group with your full product catalogue means the algorithm is matching generic creative to every product category, which produces mediocre results across the board.

The right approach is to segment by product category (clothing, accessories, footwear) or by margin tier (high-margin vs. lower-margin). Each asset group gets creative specific to the products it covers, and you can set different ROAS targets for asset groups with different margin profiles.

What goes into each asset group

  • Headlines (15, at least 8 unique): Mix product-specific, benefit-led, and brand headlines. Avoid repetition.
  • Descriptions (5, at least 3 unique): Expand on headlines with more context. Include offers, USPs, delivery.
  • Images: At minimum — landscape (1.91:1), square (1:1), and portrait (4:5). Product images on white background for Shopping; lifestyle and brand images for Display.
  • Logos: Square and landscape versions of your logo on a white or transparent background.
  • Videos: Optional but recommended. 10-30 seconds for YouTube. If you don't provide video, Google auto-generates one from your images — usually poorly.
  • Final URL: The most relevant category page or collection, not your homepage.
  • Product feed link: Connect your Merchant Center feed and filter to the relevant products for each asset group.
  • Audience signals: See below.

Audience signals: the most important input you probably skipped

Audience signals tell the algorithm where to start looking for customers. Without them, it has to learn from scratch using your conversion data, which takes longer and costs more.

The strongest signals are first-party data: customer purchase lists, email subscribers, recent site visitors from high-intent pages (product pages, checkout). These represent people who have already demonstrated interest or intent, and the algorithm uses them to find similar audiences across Google's network.

Weaker signals — Google's interest categories, in-market segments — are better than nothing but won't give you the same advantage. Build your own lists first.

Brand exclusions: apply them immediately

By default, PMax will bid on your branded search queries. This inflates your reported ROAS by capturing easy conversions from people who were going to buy anyway, and diverts budget from new customer acquisition.

In your PMax campaign settings, find "Brand exclusions" and add your brand name and any common variations. This forces PMax to focus on non-brand inventory where it's actually generating incremental value.

Setting the right ROAS target

Start with a target at or slightly below your current average performance, not an aspirational goal. The algorithm needs to learn from real conversion data, and an overly aggressive target restricts volume and slows learning.

Once performance stabilises — typically 4-8 weeks — you can begin tightening the target incrementally (5-10% at a time, with at least 2-week observation periods between changes).

Key principle: PMax ROAS targets should be grounded in your contribution margin, not benchmarks or aspirations. If your margin is 45%, you need a ROAS above 2.2x just to break even on ad spend. Set your target with a meaningful buffer above that.

Diagnosing PMax underperformance

When PMax underperforms, the cause is almost always one of five things:

  • Bad conversion data: If what PMax is optimising toward doesn't accurately reflect actual purchases, nothing else will fix it
  • Brand cannibalisation: High ROAS but not growing? Check if PMax is eating brand traffic
  • Weak signals: No first-party audience data means the algorithm is flying blind longer than it needs to
  • Generic creative: If your asset groups don't have category-specific headlines and images, the Creative Quality score will be low and performance will suffer
  • ROAS target too aggressive: Too-high targets restrict volume, reduce learning data, and lead to erratic performance

Work through this list before assuming "PMax doesn't work for us." The issue is usually fixable.

Want your PMax setup reviewed?

We audit PMax campaigns against all five common failure modes — and fix what's actually limiting performance.

Apply for an Audit