Performance Max: when it works and when it quietly wastes money

Performance Max produces some of the strongest results I've seen in e-commerce accounts, and some of the most flattering-but-meaningless reporting I've seen anywhere.

Both things are true. The difference is setup.

What PMax actually does

It runs across Search, Shopping, Display, YouTube, Discover, Gmail and Maps from a single campaign, using automated bidding and asset combinations chosen by Google.

You give it assets, a feed, a budget and a goal. It decides nearly everything else.

The credit-stealing problem

This is the issue that matters most, and it's the one people miss.

PMax will bid on your brand terms unless you stop it. Someone searches your business name — a person already coming to you — PMax serves an ad, they click, and PMax books a conversion.

Your PMax ROAS looks superb. Your total revenue hasn't moved. You've paid for traffic you were getting free, and the reporting rewards it.

Fix: apply brand exclusions at campaign level. Then compare PMax performance before and after. The drop tells you how much of your "performance" was branded traffic all along.

Every account I've audited that ran PMax without brand exclusions was overstating it. Every one.

When PMax genuinely works

• E-commerce with a healthy product feed. Feed quality is the single largest input. Good titles, accurate attributes, real images.
• Accounts with sufficient conversion volume. It needs data. Thin accounts don't give it enough to learn.
• Broad, mainstream products. Where Google's audience signals have something to work with.
• When you've already saturated Search and want incremental reach.

When it wastes money

• Low conversion volume. Below roughly 30 conversions a month, it's guessing.
• Lead generation with poor lead quality controls. It optimises for form fills, and it will find you people who fill in forms. Whether they're worth anything is your problem, not the algorithm's.
• Businesses with wildly different margins across products. One PMax campaign with one target can't respect that.
• When you can't exclude brand. See above.

Setup that limits the damage

Exclude brand terms. Non-negotiable.

Use asset groups deliberately. Structure them by margin tier or product category, not one group with everything.

Feed the conversion signal you actually care about. For lead gen, that means passing back qualified leads, not raw form fills. This is more work and it's the difference between PMax working and PMax generating noise.

Add negative keyword lists at account level where your account supports it.

Set a genuine target, not "maximise conversions" with no constraint.

Reading the reporting honestly

PMax reporting is deliberately limited. You get less channel-level visibility than you'd want.

The one comparison that cuts through: total account revenue and total account spend, before and after. If PMax "added" R200,000 in attributed revenue and total revenue rose R40,000, you know what happened.

Always evaluate at account level, not campaign level. That principle applies well beyond PMax, but PMax is where it matters most.

The honest summary

PMax is a good tool with a reporting model that flatters it. Use it, exclude brand, judge it on incremental revenue rather than attributed ROAS, and it earns its place.

Skip that discipline and you'll spend a year very pleased with a campaign that's quietly buying customers you already had.