Performance Max, Honestly: When It Works, When It Wastes Money
A field guide to Google's Performance Max campaigns — the data it needs, the exclusions nobody sets, and how to tell whether it is finding new demand or reselling your own brand traffic.
Google Ads · · 9 min read
Performance Max is the campaign type advertisers argue about most, and almost all of the argument comes from comparing accounts that are not comparable. A retailer with 4,000 SKUs, a clean product feed and 600 conversions a month is running a fundamentally different machine from a two-location dental practice with eleven form fills a week. Both are called Performance Max. Only one of them has enough signal to work.
What Performance Max actually is
Performance Max takes one budget and one goal and spends across Search, Shopping, YouTube, Display, Discover, Gmail and Maps. You do not choose placements or keywords. You supply asset groups — headlines, descriptions, images, video, a product feed if you have one — plus audience signals, and the system decides where each impression goes. In exchange for giving up control you get inventory that is difficult to reach any other way, particularly Shopping and YouTube in the same auction logic.
That trade only pays when the system has enough conversions to learn from. Our working threshold across managed accounts is roughly 30 conversions in a rolling 30 days before Performance Max behaves predictably, and closer to 50 before we trust it with a majority of budget.
The prerequisites we insist on before launch
- Conversion tracking that fires once per real business outcome — not on every page view, thank-you page reload or phone-number click.
- Values attached to conversions. Even estimated values (a booked survey is worth 4x a brochure download) transform how the bidding behaves.
- Enhanced conversions or offline conversion imports where the real sale happens in a CRM days later.
- A brand exclusion list, so Performance Max is not billing you for people already typing your name.
- At least one genuine video asset. If you supply none, Google auto-generates one, and the auto-generated versions consistently underperform anything filmed on a phone with intent.
The brand-traffic problem
The single most common reason a Performance Max campaign looks brilliant in month one is that it is quietly harvesting brand searches that would have converted anyway. Return on ad spend of 9x on a new campaign in a business with no new demand should be treated as a red flag, not a result. Add your brand terms to the account-level brand exclusion list, then compare incremental conversions against the baseline of the four weeks before launch.
In one home-improvement account we took over, 61% of Performance Max conversions were brand searches. After excluding brand, reported ROAS fell from 7.4x to 3.2x — and total company revenue did not move at all. The 3.2x was the truth, and it was still a good number. It just was not the number the previous report was celebrating.
Structuring asset groups so reporting means something
Because you cannot see placements, your asset group structure is your only reporting dimension. Split asset groups the way your margins split — by product category, by service line, or by customer segment — never by 'variation A / variation B'. If your gross margin on gutters is 22% and on full roof replacement is 41%, those belong in separate asset groups with separate target return on ad spend values, because a blended target quietly funds the wrong one.
Reading the reports you do get
- Search-terms insights: thin, but enough to catch obvious irrelevance and inform negative keyword lists at the account level.
- Asset performance ratings: treat 'Low' as a prompt to replace, not to panic — ratings are relative within the asset group.
- The listing groups report: on feed-based accounts, this shows which products absorb spend. Chronic offenders are usually low-margin items with high click volume.
- Segment by device and time. Performance Max leans heavily mobile, and if your booking flow is painful on mobile you will read that as a campaign problem rather than a landing page problem.
When we do not use it
We keep clients on standard Search campaigns instead when monthly conversions sit below 20, when the offer is highly consultative and lead quality varies enormously by keyword, or when the client cannot supply imagery and video of any usable standard. In those accounts, exact and phrase-match Search with tight negatives and manual review of search terms simply produces better qualified pipeline for the same money.
A realistic ramp plan
- Weeks 1–2: launch at roughly 20% of account budget with brand excluded and a target cost per acquisition set 20% above your allowable figure. Change nothing.
- Weeks 3–4: review search-term themes, add account negatives, replace the two weakest assets in each group.
- Weeks 5–8: if cost per qualified lead is within range, increase budget in 20% steps no more often than every four days.
- Ongoing: refresh creative monthly, review listing groups fortnightly, and re-verify that conversions still map to real revenue in the CRM.
Performance Max is neither a miracle nor a scam. It is a leveraged bet on the quality of your conversion data. Improve the data and it improves; feed it noise and it will spend your budget confidently in the wrong direction.
Want this applied to your account?
Book a free 30-minute call and we will review your current setup against the points above.
Book a strategy call