Google Lifts the Lid (a Little) on AI Buying Tool Pmax, Giving Buyers More Ways to Test

Google has rolled out an A/B testing feature for its AI-powered ad-serving tool Performance Max. This lets advertisers test how certain changes impact campaign performance, such as changing creative assets or the landing page, giving buyers more visibility into performance, four ad buyers told ADWEEK. Google announced in late October that it plans to roll out "new experiments" in Pmax, letting advertisers add more assets (images, text, and video) to retailers in beta in November, as well as Final URL expansion (letting Google dynamically select the most relevant landing page). Some sources said they have had access to The ‘Performance Max Optimization Experiments’ feature for just over a month, gaining early access via a whitelist program, where Google grants select users priority access. “A/B testing helps advertisers in a big way. Advertisers have taken a set-it-and-forget approach with Pmax because of the lack of insights," said Andy Goodwin, head of paid search at Brainlabs. "This functionality, once it's fully available, allows them to be more rigorous and have a renewed focus on Pmax.” While marketers have been happy to use Google’s three-year-old Pmax, they have also criticized its black-box nature. Often set and then forgotten, media buyers have been unable to optimize campaigns continuously. “One of the reasons why a lot of clients were not wanting to adopt Performance Max [is because] it was very black box,” said Michelle Merklin, vp, paid search, Tinuiti. “There wasn't the ability to understand the impact of making changes within a Pmax campaign.” But split testing shows the impact of campaign changes in real-time. By contrast, before-and-after comparisons can be unreliable and don't account for external factors like competitor activity or shifts in business operations. Despite Google offering more controls, some ad buyers remain skeptical about the reliability of the results. Expanded testing elements Google’s latest update responds to a long-standing call from brands and advertisers for greater control, sources told ADWEEK. The tech giant first teased split testing for Pmax about 18 months ago, though the feature has faced repeated delays, according to Goodwin. Now, advertisers can set a base campaign on Pmax and run an experimental campaign simultaneously, testing elements like creative assets or bidding strategies, sources said. Once the test begins, Google divides traffic equally between the two campaigns, serving either the base or experimental ad to users across search, display, YouTube, and other Google platforms. After the test period, advertisers can see which campaign performed better. More nuanced insights At Tinuiti, nearly 10 clients have requested access to the split testing feature, with many looking to test campaigns that will drive new customer acquisition, according to Merklin. Other elements marketers can test include smart bidding strategies (adjusting bids based on goals like cost-per-acquisition or return on ad spend), final URL expansion, audience signals (inputting custom audiences, like first-party data, to target most relevant audience), and search themes (defining keywords or categories to align with specific search behaviors). Merklin said that this feature offers a clearer understanding of how changes impact performance while accounting for external variables like seasonality. The result is more reliable insights for the advertiser. Over reporting performance Not all buyers are fully on board with the feature, though. An anonymous executive from a marketing agency said that while Pmax A/B testing has become standard in optimization calls, agency adoption has been slow because buyers still want more control over test elements. Unlike traditional A/B tests, where two separate campaigns are run against each other, Pmax A/B tests are conducted within the Pmax account—which supports multiple Pmax campaigns across search or display. This lack of separation makes it difficult to isolate the impact of specific changes. “This is an issue with transparency,” the executive said. “We have seen Pmax cannibalizing search volume from search campaigns running concurrently. We have no insight into whether it is truly valuable or if it’s just an inefficient use of media dollars." Rain the Growth Agency, which previously tested split testing for Google Ads outside of Pmax, found that Google’s reports tend to overperform, especially in the B segment of the test. “Performance has always looked better in those scenarios,” said Stacia Fulginiti, director, paid search + YouTube. “So, we’re wary of fully leveraging these experiments.” A Google spokesperson responded "The advertiser can define a traffic split using the experimentation tool, which ensures that neither arm of the experiment is inherently favored."

Platforms

Google Lifts the Lid (a Little) on AI Buying Tool Pmax, Giving Buyers More Ways to Test

Brands have been clamoring for more transparent ways to test the powered ad-buying tool

Create Your Free Account

Registration Includes:

Trishla Ostwal

About

Subscriptions

Events

Publications