Google’s Privacy Sandbox Origin Trials Are Officially in the Experimental Stage

This week, Google’s new vision for privacy and the economics of the Internet reached a milestone toward becoming a reality. Two months into the origin trials for Privacy Sandbox, 50% of the beta population, or developers who agreed to be early testers of Chrome’s new privacy software, are now eligible to test products like Topics, Fledge and the Attribution Reporting API. Since Google began the origin trials in early April, it has been increasing eligible testers, with the goal of reaching the 50% threshold. At that point the tech giant will split testers into two groups: an experiment group and a control group, product lead for chrome privacy sandbox Vinay Goel told Adweek. But the path hasn’t been seamless, and will continue to be rocky, sources tell Adweek. Last week, a bug in the Topics API made the product temporarily unavailable for testing. And as more developers root around in the sandbox, and more information about Google’s products leak into the market, several industry sources have raised concerns that Sandbox tools may have unintended negative consequences for publishers. That’s when they have the opportunity at all to sign for the trials, which many publishers lack the time or resources to do, said one publishing ad-tech source who spoke to Adweek. “It’s a highly technical problem with a bunch of highly technical solutions, and publishers don’t have a lot of free engineering hours,” the source said. Meta Looks to Core Apple in Comments Filed With the NTIA Critiques of Privacy Sandbox, and the original trial process, are of increased importance as the clock is ticking on the 2023 deadline for cookie deprecation. With Chrome’s large market share and the cookie framework underpinning digital advertising for decades, Google’s process to reimagine targeting through the lens of privacy will reverberate across the industry. De-bugging Topics In a blog post last week detailing progress with the origin trials, Google revealed it had discovered a crashing bug within the Topics API—its contextual targeting solution—and would temporarily disable the API in origin trials. Another ad-tech source, who also requested anonymity to protect industry relationships, said the company gave stakeholders little information about the cause. Goel told Adweek that the root of the issue was from Chrome’s back end. Google assigns websites that are sensitive and not suitable for advertising a “null” Topic, which did not compute with the system designed to receive this signal, Goel said. The issue was resolved within 24 hours, he added. While this particular bug was not discovered by developers testing Privacy Sandbox in origin trials, Google wants as many testers as possible to detect other possible glitches. “There is no guarantee that the code is bug free,” Goel said. “It really depends if developers are able to identify the bug and whether they report it to us.” Naming rights Even if Topics did work as intended, it may not be useful, our publishing source told us. Topics assigns only one topic per domain name. For example, The Wall Street Journal might be tagged as “business.” But that’s too narrow, the source said. “A lot of these proposals are specifically targeted at large userbase multi-domain entities. If you’re an ad network and you run on multiple domains, these proposals have a lot more utility for you,” the source said. “If you’re not, the efficacy is very small.” Goel disagreed, saying that every domain receives several topics, each with a different weight of relevancy. A single general interest publisher domain could correlate with several different topics, each with different weights—such as travel or shopping. Another bone of contention is that Chrome gets to assign the topics to each publisher, a power that publishers would rather have themselves, the second ad tech source said. Chrome addressed this concern in a document about origin trials feedback that it submitted as part of its obligation to the UK’s Competition and Markets Authority. “Chrome remains open to input on the taxonomy,” the document states. Data leakage for small publishers? Another concern is that Topics systemically undermines smaller publishers, according to Don Marti, vp of ecosystem innovation at publishing network Café Media. For example, Topics might tag a viewer of a shoe specialty website as interested in shoes. When that user goes to a big, general interest publisher like YouTube, YouTube gets a valuable piece of information—this is someone interested in shoes—but because YouTube is a very broad site, it has no niche topic to offer the bidstream, benefitting from small publishers’ data without giving any of their own. “It enables tagging [of small publishers’] audiences to be monetized elsewhere,” Marti said. Not all in the publishing space think this is a major problem. The issue of data leakage is not new, said the publishing ad-tech expert, and while Topics may not eliminate the problem, it may be an improvement. Goel said that in the previous example, YouTube would only know a user was interested in shoes if the user who visited the shoe website also visited YouTube, allowing less data leakage than the current system. He added that Topics benefits small publishers. Without Topics, a specialty shoe website might only receive contextual advertisements around shoes. But with Topics, the website might know that their user who is interested in shoes is also interested in tennis—Topics works by choosing three topics for each user to share with advertisers and publishers every three weeks.