Buyers Question Inventory Quality, Transparency on Google’s AI-powered Performance Max

Ad buyers are finding that large chunks of impressions spent on Google’s artificial intelligence-fueled product, Performance Max (Pmax), are ending up in less desirable places, like on open web inventory and branded search terms.

One ad buyer, who requested anonymity to protect industry relations, found that around 80% of impressions on a Pmax campaign ended up on open web inventory. The rest landed on Google’s owned and operated properties like Search and YouTube. Five buyers told Adweek that Google-owned properties generally perform better, and they would rather not run ads against display inventory.

“Knowing Google and Google products, really, it’s not surprising at all because they have so many instances of pushing ad dollars where brands don’t intend it to be, without transparency,” the buyer said.

In the two years since Google rolled out Pmax, it has become one of the company’s buzziest buying tools. With the goal of optimizing toward performance goals, such as online sales and lead generation, the tool’s algorithms place ads across a wide swath of Google inventory, including Search, YouTube, display, Gmail and Maps. Ad buyers continue to find the tool, which takes some control from buyers around placement, lacking in transparency, a critique that’s particularly pertinent following a recent Adalytics report that found ads were unknowingly targeted to children.

While Pmax gives buyers insight into which websites their ads run against on the open web, it doesn’t tell advertisers on which YouTube channels their ads air, making it hard for advertisers wanting to audit their buys to make sure they’re not showing ads to children. YouTube has refuted the Adalytics’ report’s findings.

Inventory quality on Pmax

For a lead gen campaign at digital agency Markacy, 20% of impressions were served on open web inventory, while approximately 5%-10% of impressions were served on non-Google properties across multiple ecommerce campaigns, said vp of media Chris Rigas.

While Rigas said the agency prefers to avoid open web inventory, more concerning is how Pmax can optimize toward branded search keywords, which are search terms of the brand’s own name. While these are effective at converting customers, they capture people who would have clicked anyway. Incremental performance improves when the agency asks Google to remove these branded search terms, said Rigas.

A Google spokesperson said that buyers looking for incrementality can optimize for finding new customers, rather than existing ones, via a customer optimization goal feature.

Apps are not a prime conversion driver, so why in the world would Google optimize for that, if not to buff up CPMs?

Anonymous ad buyer

Three buyers told Adweek they saw between 33% and 90% of impressions in their Placement Reports, which Google rolled out in January 2022, were delivered to open web inventory, which buyers generally consider low quality.

However, Placement Reports are not reflective of an entire Pmax campaign since they don’t include search impressions. When buyers attempted to include Search impressions, the proportion of open web inventory was sometimes lower, though Google does not provide a full breakdown of all impressions by inventory type.

“Apps are not a prime conversion driver, so why in the world would Google optimize for that, if not to buff up CPMs and ensure delivery?” said another buyer, who declined to be named because they didn’t receive authorization from their agency. The buyer’s agency noticed Pmax inventory weighted more toward open web than it would prefer.

Other buyers told Adweek that overall Pmax performance is good enough and display inventory is sufficiently low such that they aren’t concerned.

“As long as the performance is decent, [the proportion of open web inventory] is kind of a meh issue,” said Adam Lovallo, vp at performance agency BMG360. “All display sucks, but it generally is only a small percent of spend, so we don’t mind. That said, we’d rather not buy display.”

A Google spokesperson said that wherever Pmax ads are delivered are the best places to meet advertisers’ goals.

“Performance Max optimizes for advertiser value based on their goals, not a specific percentage of inventory per channel,” a Google spokesperson told Adweek.

How to know where Pmax ads ran

The transparency of Pmax, while improving, is another frustration with the tool, buyers told Adweek.

Rigas and the first anonymous ad buyer understood Pmax inventory breakdowns by triangulating several different Google reports or by using third-party APIs, as all Pmax placement reporting for a campaign does not exist in one centralized place.

Advertisers can learn where the Pmax algorithm placed their campaigns via two reporting tools. According to the Google spokesperson, Placement Reports show the impressions that ran on open web inventory and Google’s owned & operated properties YouTube, Maps and Gmail, but critically not Google Search.

Advertisers can also learn about their Pmax search impressions via Search Term Insights, a tool released more recently that can also be used for non-Pmax Google campaigns.

Three buyers told Adweek that they did not realize that Pmax Placement Reports did not include Search inventory.

“Google’s explanation and the way they surface the data is not very clear,” Rigas said, noting that clearer Placement Reports could help buyers with attribution, brand impact and creative optimization. “So Google being so opaque with that data is not ideal.”

Google also does not break down Pmax inventory by spend or performance, only by impressions, Lovallo said.

“The reporting for normal campaign types is way more granular,” Lovallo said. “That’s kind of by design. They sort of built Pmax to be a black box.”