Performance Max Isn't Autopilot: How E-commerce Brands Should Actually Run PMax

Performance Max is the campaign type Google most wants you to run and the one founders understand least. It promises to find buyers across Search, Shopping, YouTube, Display and Gmail with one campaign and a target. For a busy e-commerce team, "give us a goal and we'll handle the rest" sounds like a gift.

It isn't a gift. It's a trade. You hand over control of placement, targeting and reporting, and in return Google spends your money efficiently against its definition of success — which is rarely the same as your contribution margin.

Run well, PMax is one of the most powerful tools in an e-commerce account. Run on autopilot, it's the fastest way to spend a lot of money on traffic you were going to get anyway.

The black-box trap

The most common PMax failure has nothing to do with bids. It's this: PMax eats your brand traffic and reports it as new performance.

Left alone, PMax will happily serve to people searching your brand name, win those near-certain conversions, and post a glorious ROAS. You celebrate, you add budget, and you slowly realise total revenue isn't moving. You're paying premium prices to "acquire" customers who already typed your name into Google.

So the first rule of PMax is the same as the first rule of any campaign: separate brand from non-brand. Use brand exclusions (now available natively) or account-level negatives so PMax has to earn incremental sales instead of harvesting existing demand.

Feed it good signals or it guesses

PMax doesn't have keywords. It has signals. The quality of those signals decides whether it finds your next customer or wastes spend exploring. Three things matter most:

• Your product feed. In an e-commerce PMax campaign, the feed is your most important asset — more than the headlines, more than the images. Titles, product types, GTINs and attributes are how PMax understands what to sell and to whom. A neglected feed is a blindfolded campaign.

• Conversion values. Send margin, not revenue, and PMax will optimise toward profit. Send revenue, and it will scale your cheapest, lowest-margin products because they convert easily.

• Audience signals. These are suggestions, not targets — but good first-party lists (purchasers, high-value customers, cart abandoners) give the algorithm a warm start instead of a cold one.

Structure for control

"One PMax campaign for everything" is the default, and it's almost always wrong for a brand with a real catalogue. You lose the ability to set different targets for different parts of the business.

Instead, structure asset groups and campaigns around margin and priority:

1. Group high-margin or hero products separately so you can push them with a more aggressive target.

2. Keep low-margin or clearance lines in their own group with a tighter ROAS goal, so they don't drain budget from the products that actually fund growth.

3. Use listing groups to control which products are eligible, rather than dumping the entire catalogue into one pool.

This is the difference between telling Google "make sales" and telling it "make these sales profitably."

Measure incrementality, not just ROAS

PMax's reported ROAS is the number it most wants you to look at and the one you should trust least, because it includes the brand and remarketing traffic that would have converted anyway. The question that matters isn't "what ROAS did PMax report?" It's "what happened to total revenue and new-customer acquisition when PMax ran?"

Run experiments. Use Google's built-in lift tests, or simpler before-and-after holdouts at the account level. Watch new-customer rate, not just blended ROAS. If PMax scales and total new customers don't move, it's harvesting, not hunting.

The honest summary

Performance Max rewards the brands that treat it like every other part of a serious account: clean inputs, deliberate structure, ruthless measurement. It punishes the ones who believe the autopilot pitch.

You don't beat PMax by avoiding it. You beat it by refusing to hand over the controls.