Writing

How to structure a Performance Max campaign so the asset groups actually pull their weight

By Christopher Corben · Surry Hills, Sydney NSW

Performance Max hands a lot of control over to Google, and the instinct for most advertisers is to pour everything into one campaign, drop in a pile of headlines and images, and hope the machine sorts it out. It usually does not, or at least not well. The lever you still hold is structure: how you split asset groups, what audience signals you attach to each, and how clean your product feed is going in. Get those three right and Performance Max behaves; get them wrong and the campaign quietly spends its budget on whichever pocket of demand is cheapest rather than whichever is most valuable.

Start with asset groups, because they are the closest thing Performance Max gives you to ad groups. The mistake is treating one asset group as one campaign. A better way is to think of each asset group as a theme with its own creative story, and to split along lines where the messaging genuinely changes. An e-commerce store might run separate asset groups by product category, because the imagery, the headlines, and the buyer differ between, say, running shoes and recovery gear. A services business might split by the distinct problems it solves. The test for whether a split is worth making is whether the creative inside each group would actually be different. If you would write the same headlines for two groups, they belong together. Splitting for the sake of it only thins out the data each group has to learn from.

Audience signals are the next lever, and they are widely misread. A signal is not a targeting setting, it is a hint about where to begin looking. Google will explore beyond it, but a precise starting point shortens the learning period and steers the early spend toward people who resemble real buyers. Build signals from your own data first: customer lists, site visitors who reached a cart or an enquiry form, past converters. Custom segments built from competitor and high-intent search terms are a useful second layer. What you want to avoid is a single broad signal stretched across every asset group, which tells the system almost nothing. Match the signal to the theme, so the recovery-gear group is seeded with recovery-gear intent, not a generic store-wide audience.

For anyone selling products, the feed does more work than the creative. Performance Max leans heavily on the merchant feed, and weak titles, missing attributes, or sloppy product types cap how well the campaign can ever perform. Spend time on descriptive titles that include the terms people search, fill out attributes like brand, colour, and size, and use product types and custom labels so you can segment the feed into listing groups that mirror your asset-group themes. That alignment, where the feed segmentation and the asset groups tell the same story, is what stops a single bestseller from soaking up the whole budget while the rest of the catalogue sits idle.

None of this is set-and-forget. Use the asset-group reporting and the placement and search-category insights to see where spend is actually going, and pull the things that do not belong, such as branded traffic you would have won anyway or low-value placements, into their own treatment. The point of a deliberate structure is not to micromanage the algorithm. It is to give it clean themes, honest signals, and a tidy feed, so that the automation has something good to optimise toward rather than guessing in the dark.