• by Laurie Sullivan  @lauriesullivan, April 18, 2023

Artificial intelligence (AI) has helped Google developers build what the company calls a “privacy-preserved” ad-targeting model that shows the promise of becoming just as successful as third-party cookie targeting.  

Developers of Google Ads and Display & Video 360 have been experimenting with interest-based audience (IBA) solutions. This also includes affinity, in-market, custom audiences and demographic segments on Google’s display network, using a combination of privacy-preserving signals.

Consumer privacy has been a focus for Google, as the industry approaches Chrome’s deprecation of third-party cookies.

Tests are underway to prove performance. In Q1 2023, Google’s ads platforms conducted an experiment to understand how interest-based audience (IBA) solutions perform when they rely on a combination of signals that preserve consumer privacy.

“These signals included contextual information, the Topics API from the Privacy Sandbox and first-party identifiers such as Publisher Provided IDs,” says Dan Taylor, vice president of Global Ads. “Our research did not compare the performance of third-party cookies to the Topics API alone but rather a broader suite of signals available in a privacy-first world.”

Two experiments took place using Chrome traffic. One relied on third-party cookies. In the other, Google replaced third-party cookies with a combination of what it calls “privacy-preserving signals for IBA solutions.” However, it continued to rely on third-party cookies for other use cases such as measurement and remarketing.  

The tests compared targeting audiences with Google Ads display campaigns using IBA, with and without AI. In the early stage of testing, there was a slight decrease in effectiveness, but the degree by which it decreased was less than expected. Google views this as positive progress.

For example, in terms of the percentage change of reported metrics using AI in Google Ads, based on the spend by advertisers, the tests show a reduction of 1.654% when compared with using third-party cookies. Without AI, tests show a reduction on average of 7.248%

When comparing conversions per dollar, with AI-powered optimization the tests show a reduction of 0.454%. Without AI, tests showed a reduction of 2.446%.

When comparing the conversion rate, with AI-powered optimization the tests show a reduction of 1.899%. Without AI, the tests show a reduction of 1.132%.

Click-through rates showed a reduction of 8.043% with AI, and 8.256% without.

Similar “success” -- a relative term in this case -- was seen in tests for Display & Video 360 display campaigns using IBA. Success because this new model preserves consumer privacy.

The percentage change of reported metric, spend by advertisers, came in at 0.24% less with AI-powered optimization and 0.08% without using Display & Video 360 display campaigns using IBA.

When comparing conversions per dollar, with AI-powered optimization the tests show a reduction of 1.91%. Without AI, the tests show a reduction of 1.63%.

When comparing the conversion rate, with AI-powered optimization the tests show a reduction of 6.07%. Without AI, tests show a reduction of 3.24%. Click-through rates showed a reduction of 1.62% with AI, and 0.365% without.