Most advertisers running Performance Max in 2026 are working from incomplete information. Not because PMax is broken — it isn’t. But because the numbers Google shows you and the numbers that actually matter are two different things.

PMax now accounts for 19.9% of total Google Ads spend, up from 16.7% in 2024. That’s a 19% increase in a single year, according to Fluency’s 2026 benchmark report. In 2025, PMax delivered 51.1 billion impressions for Fluency clients, compared to 7.8 billion from Google Search. 71% of advertisers now use it, up from 60% in 2024.

The experiment is over. PMax isn’t a test campaign type anymore. The question is what you’re actually getting for that budget.

19.9%of total Google Ads spend goes to PMax in 2026
71%of advertisers now use Performance Max (up from 60%)
51.1Bimpressions delivered by PMax in 2025 vs 7.8B from Search

What the ROI Data Looks Like Outside of Google’s Dashboard

Cassandra analyzed 253 Marketing Mix Models across 59 advertisers, covering $383 million in total media spend. The goal: measure actual incremental ROI — what your spend caused, isolated from attribution overlap, seasonality, and cross-channel halo effects.

Campaign TypeMedian Incremental ROI95% CI
Search Non-Brand5.21xNarrowest range
Performance Max4.64x3.11x – 6.41x
Search Brand4.14x

Those numbers are consistently lower than platform-reported ROAS — by a factor of two to five times — because they measure incremental impact rather than last-click attribution. Platform ROAS tells you what Google attributed to a campaign. Incremental ROI tells you what would not have happened without it. Those are different questions with very different answers.

The honest read: PMax is consistent. The 95% confidence interval runs from 3.11x to 6.41x. You’re unlikely to hit the extremes. For advertisers who want predictable performance across Google’s full inventory without managing separate Search, Shopping, Display, and YouTube campaigns, that consistency has genuine value.

What PMax doesn’t offer is transparency into where that budget went. In Cassandra’s analysis covering $66 million in PMax spend, most advertisers could not break down what portion went to Search, Shopping, Display, or YouTube. You’re buying a bundle and trusting Google’s algorithm to distribute it.


The Transparency Problem Is Getting Fixed, Slowly

PMax launched with a structural criticism baked in from day one: you hand over creative assets, audience signals, and budget, and Google decides everything else. That was a reasonable trade when PMax was generating strong platform-reported ROAS. It became harder to defend when advertisers started asking how much of that ROAS was just capturing branded traffic they would have gotten anyway.

Google has been addressing this incrementally:

None of this changes how PMax works at its core. But it gives you enough data to manage the campaign rather than just monitor it. That’s a meaningful difference.

The cannibalization problem: A PMax campaign running without brand exclusions will frequently serve ads to users already searching your business name — inflating ROAS considerably. The fix is straightforward: apply brand exclusions. Many accounts still don’t have them in place.


The Cassandra benchmarks are clear on one thing: Search Non-Brand outperforms PMax on incremental ROI at the median. That’s not a knock on PMax (4.64x is a solid return), but advertisers who treat PMax as a replacement for well-structured Search campaigns are likely leaving returns on the table.

The more useful framing is that PMax and Search serve different functions:

Optmyzr’s research found that PMax campaigns allocating 50% or more of account budget achieve 625% ROAS at maturity versus significantly lower performance for lower-allocation campaigns. The implication: PMax needs volume. If you’re running it on a tight budget below 30 conversions per month, standard Shopping or Search campaigns typically outperform it because they don’t require the learning-phase data PMax depends on.


What to Actually Do With This

PMax works. It delivers consistent returns, it scales across Google’s full inventory, and the transparency improvements in 2026 make it genuinely more manageable than it was a year ago. But consistent doesn’t mean optimal. The benchmarks exist now — use them to set expectations that don’t depend on Google’s attribution logic.

Sources

Fluency 2026 Google Ads Benchmark Report → Optmyzr Performance Max Benchmark Report →