Performance Max launched in late 2021 as the campaign type that spends your budget across Search, Shopping, YouTube, Display, Gmail, Discover and Maps while reporting almost nothing about where the money went. I have run it since the forced Smart Shopping migration, and for three years the deal was simple: hand over the budget, accept one blended number, ask no questions.
Between January and April 2025, that deal changed. Google shipped the transparency features advertisers had been requesting since day one, and a year of running accounts with them has changed how I structure every e-commerce account I manage. Not because PMax got better. Because I can finally see what it was doing all along.
The 2025 transparency drops, in order
January 23, 2025: Google announced campaign-level negative keywords rolling out to all PMax advertisers. Account-level negatives existed before, but campaign-level control is what lets you keep brand queries out of one campaign while still allowing them in another.
March 2025: Google raised the PMax negative keyword cap from 100 to 10,000 per campaign, matching Search campaigns, as reported by Search Engine Land. The original 100-keyword cap was the tell that January’s release was a concession rather than a conviction. The 10,000 cap made it an actual tool.
April 30, 2025: the big one. Google announced channel-level reporting in open beta, with clicks, conversions and spend broken out for each surface PMax buys, plus a full search terms report described as similar to what Standard Search and Shopping campaigns get, per Search Engine Journal.
Of the three, the search terms report changed my week-to-week work the most. Channel reporting mostly confirms suspicions you already had about where the budget leans. The search terms report hands you decisions.
Worth saying plainly: none of this changed how PMax spends. It changed what you can prove about how PMax spends. Those are different things, and the second one is what audits, client conversations, and structural decisions actually run on. A black box you can read is still a black box, but at least now the receipts exist.
What the search terms report showed me in week one
The first thing I did when search terms visibility landed in one of my e-commerce accounts was sort by conversions and look for the client’s brand name. It sat right at the top, exactly where the skeptics always said it would be.
PMax had been taking credit for people who searched for the store by name, people a branded Search campaign would have converted at a fraction of the cost. I added the brand as a campaign-level negative. Reported ROAS on the PMax campaign dropped. Total account revenue barely moved. That gap is the point: the distance between those two numbers is roughly how much the blended reporting had been flattering itself.
I will not attach a percentage because it varies enormously with brand strength, but I have repeated this exercise across several accounts and the direction is always the same. If you have not negated your brand in PMax, your PMax ROAS is partly fiction.
Google’s number versus the independent data
Google’s headline claim, published in February 2023, is that advertisers using Performance Max achieve on average over 18% more conversions at a similar cost per action. Treat that the way you treat any platform grading its own homework: as the upper bound of a self-selected sample.
The best independent dataset I know is Optmyzr’s study of 9,199 accounts and 24,702 campaigns, published in October 2024. Two findings from it drive my structure decisions.
In Optmyzr’s data, accounts reaching at least 60 conversions in 30 days performed significantly better with Performance Max than accounts below that threshold, and Standard Shopping posted impressive ROAS wins when the two campaign types ran side by side.
Both findings match what I see managing $80K-plus a month. PMax is a signal-hungry system; starve it of conversions and it guesses with your money. And Standard Shopping keeps winning on ROAS wherever someone actually does the query sculpting work.
My working rules, one year in
- Under roughly 60 conversions a month, use Standard Shopping. Below the signal threshold you are paying tuition for an algorithm that never graduates.
- Brand negatives on day one, at campaign level. This is now trivial to do and there is no longer an excuse.
- Start feed-only where creative is weak. A mediocre auto-generated video does not deserve budget just because the campaign type can make one.
- Run both campaign types. Standard Shopping for the query tiers you want to price yourself, PMax for the long tail and its cross-channel surfaces.
- Work the search terms report weekly. Junk queries become negatives. The cap is 10,000 now; use it.
- Check channel reporting monthly. If a surprising share of conversions comes from surfaces you cannot independently evaluate, that is a research task, not a victory lap.
- Re-test the split every quarter. Google keeps tuning the auction relationship between PMax and Standard Shopping, and a structure that won in Q1 is not guaranteed to win in Q3. The reports now exist to check; checking is the job.
Grade the homework
PMax in mid-2026 is a legitimate tool that finally submits to auditing. It is still not a strategy. The advertisers who got burned by it were rarely burned by the algorithm itself; they were burned by not being able to see what it did with their money.
Google took more than three years to show its homework. It finally has. Now grade it, in your own account, with your own search terms report open.