Performance Max is Google’s most automated campaign type: a single campaign, budget and goal that spans Search, Shopping, YouTube, Display, Gmail, Discover and Maps. For a D2C brand it can be the highest-leverage channel you run, or a quiet budget incinerator. The difference is almost entirely in the inputs you hand the machine.

Google markets Performance Max as “set a goal and let AI do the rest.” That is true the way “just add water” is true of bread. The system is genuinely powerful, but it optimises toward whatever signal you feed it. Feed it a weak signal and it will spend happily, winning branded searches you would have got for free, chasing discount hunters and accidental clicks, then report it all as success.

§01Your product feed is the whole game

For ecommerce, the product feed is the spine of Performance Max. Titles, descriptions, attributes, images and price competitiveness decide which queries you appear for and how Shopping renders you. A campaign sitting on a thin or messy feed cannot be rescued by bidding tweaks, the model simply has less to work with. Treat the feed as primary creative, not plumbing.

  • Front-load product titles with the words buyers actually type, category, brand, key attribute, not internal SKUs.
  • Fill every attribute Google offers: GTIN, brand, product type, colour, size, material. Each one is a targeting and reporting lever.
  • Use clean, well-lit images; they win the Shopping click before a word of copy is read.
  • Watch the price competitiveness and best-seller reports, they show where you are losing on margin versus demand.

“For ecommerce, the product feed is the spine of Performance Max.”

§02Give the algorithm a conversion worth chasing

Performance Max optimises toward your conversion action with ruthless literalism. If every purchase counts the same, it will chase volume, including cheap, low-margin orders. The fix is value-based bidding: pass real revenue (ideally profit-adjusted) back as the conversion value, then target a return on ad spend rather than a flat cost per acquisition.

Where you can, separate new-customer value from repeat purchases. A first order from someone who never buys again is worth less than one from a future loyalist, and Google’s new-customer-acquisition setting lets you tell it so.

§03Asset groups, signals and exclusions

Structure asset groups around themes or margins, not single products, group hero ranges, seasonal lines and clearance separately so you can read and steer them. Audience signals are hints, not hard targeting: seed them with your customer lists and high-intent in-market segments to speed up learning, then let the system expand.

  • Add brand exclusions so Performance Max stops claiming credit for searches of your own name.
  • Exclude irrelevant placements and, where allowed, low-value geographies.
  • Use account-level negative keywords to keep it off informational and competitor-brand terms.

§04Reading a campaign that hides its data

The longest-standing complaint about Performance Max is opacity: it blends channels and withholds the search-term and placement detail advertisers expect. You are not powerless, the search-terms insights, asset-group reporting and the insights tab reveal more than the headline view, and scripts can surface channel and placement data Google does not show by default.

The honest takeaway: Performance Max rewards operators who feed it clean data, a meaningful value signal and tight exclusions, and punishes those who treat it as a black box. Get the inputs right and it becomes the closest thing Google offers to a profitable autopilot.