Google announced this year that it would eliminate automatic third-party cookies on its Chrome browser. The company joins Apple and Mozilla, which earlier made users opt in to the technology on their browsers. While Google’s move may be positive for users who want privacy, it’s bad for companies that want to target ads to specific audiences.
Boston University’s Tesary Lin and Chicago Booth’s Sanjog Misra evaluated the alternatives for advertisers. Building an analytical framework and conducting an empirical experiment, they find that advertisers have few good options for constructing accurate user profiles.
“The system was already broken and imperfect,” Misra says. “Any time there is fragmentation, anything you want to measure is going to be off to some degree. Now the rules have changed, and it will exacerbate that fragmentation to a much greater degree.”
Cookies are tags from websites that live on a person’s computer or internet-capable device. When you search for shoes online, the search page and the sites you visit leave cookies that can be collected to tell a company such as Zappos that you’d be a good target for an ad.
Because people use more than one device, a third-party company can collect cookie data to see a person’s browsing history. These companies use a data-linking strategy to cross-reference cookies to see where names, email addresses, and other identifying features overlap and then stitch together a more complete user profile. Advertisers can track if an ad is effective, even if that ad is viewed on a smartphone by someone who bought the shoes on a laptop a day later.
But the picture that emerges is never perfect. People may use different email addresses on a work computer and on a tablet at home, keeping those fragments from being connected.
With Google and Apple controlling about 85 percent of the browser activity in the United States, according to web-traffic analyst Statscounter, the new restrictions make third-party cookies essentially obsolete and increase the fragmentation, leaving companies less certain about how to target ads effectively. Even companies that use data linking will have to rely more on partial links, increasing bias in their assessment of an ad’s effectiveness, the researchers write.
Could companies run trials to estimate the effectiveness of their ads on consumers, and make adjustments on the basis of a series of assumptions? The researchers conducted an experiment using data from an online seller of durable goods that included about 390,000 observations of display ads, search ads, and social media ads. They find that this approach doesn’t work because the assumptions are untenable. Assuming that all users have three devices, for example, will seriously skew the results if some users have only one device and others have seven. At best, results are approximations for large populations.
The researchers propose a third option, stratified aggregation, which would pull together fragments of data into higher-level demographic groups. This method might tell a company how well an ad works for men in Mumbai, but it wouldn’t drill down to the individual level to allow for more granular targeting.
Companies with paywalls or those that require log-ins will have an advantage, the researchers write, because they’ll be able to track users’ activity directly. Others will face significantly more fragmentation bias that will require new data-aggregation methods.
“Without these solutions, the post-cookie world can render data analytics difficult or even impossible for most firms, while a few established firms who can obtain complete user data using a login wall can have a stronger incumbent advantage,” the researchers write.
More from Chicago Booth Review
We want to demonstrate our commitment to your privacy. Please review Chicago Booth's privacy notice, which provides information explaining how and why we collect particular information when you visit our website.