For decades, advertisers relied on third-party cookies to follow and understand how consumers behave online. These tracking data packets provided them with valuable information that helped them figure out how to target and display their ads. However, as users became more aware of the implications of ad surveillance, more and more tech companies started incorporating smart cookie-blocking technology to give them peace of mind.
As a result, many cookie laws and numerous ad-blocking apps started to emerge. This new norm has pretty much left the cookie tracking system to die. And, as the ad surveillance era seemingly comes to its end, Google has come up with some interesting ideas for data sources in the future of marketing. That's where the Federated Learning of Cohorts (FLoC) enters the picture.
The Privacy Sandbox is Google’s new proposal for how to conduct behavioral targeting without third-party cookies doing all the leg work. The FLoC might be the most dangerous element within this initiative. It’s meant to be the new way for browsers to do user profiling. It will observe and label what millions of people do online.
At Trustpage, we believe businesses should start paying more attention to their security policies as more data collection techniques like FLoC develop, and we believe that consumers have a right to online privacy. As the ad surveillance era faces its imminent death, we must demand more transparency. That's why we — and many other tech companies — are taking a stand against Google's FLoC in order to protect our users and stay true to our mission.
Now, before you jump to any conclusions, we want to start by making one thing very clear: Google's great at what they do, and we rely on many of their products and services to keep Trustpage going. But there are some things they've come up with that we fundamentally disagree with, FLoC being one of them. This doesn't make them evil, or a bad company, it just makes for an opportunity—we think—to improve.
What Is FLoC?
Before we define the future of ad tracking according to Google, let's get some context.
Google introduced the Privacy Sandbox initiative in 2019. This project presents an outline of what online privacy should look like in the near future. It pretends to replace cookies by instating protocols to fill the gap they'll leave behind — and by that, we mean to satisfy the needs of marketers and advertisers. It puts revenue ahead of user privacy.
Since Google’s announcement, the ideas behind its Privacy Sandbox, which includes FLoC, have been discussed in the Web Advertising Business Group of the World Wide Web Consortium (W3C). The issue here is that the Web Advertising Business Group is mainly composed of tech vendors, so there's an obvious conflict of interest.
Google and these other ad-tech representatives have come up with a myriad of proposals in which they design the functions FLoC will serve in the targeted advertising world once cookies are gone for good.
The Federated Learning of Cohorts (FLoc) is an API that will integrate with Chrome to replace third-party cookie tracking in helping marketers perform interest-based tracking. The project was announced in January 2020. It claims to offer several privacy advantages to internet users while still providing them with a tailored advertising experience.
How Does FLoC Work?
A browser with this controversial API enabled would gather data on people's browsing preferences and habits. This information would later be used to classify users into different groups. This way, people who have somewhat similar interests and browsing habits would belong to the same "cohort." Each member in a group would have their own cohort ID, which would later be used by different sites to better target their ads.
Every cohort would represent an audience made up of thousands of users. Its members would all have things in common, helping the FLoC SimHash algorithm — a technology that estimates similarities between sets — classify them more accurately.
However, as if this whole cohort analysis approach wasn't dystopic enough, FLoC would also have a pretty comprehensive summary of who users are, what they like, what they recently searched for, and more.
Initially, Google experimented with 8-bit identifiers, which compare data of up to eight bits and offer a low output when a match is made. This method yielded a limited possible number of cohorts, as they were limited to 256. Yet, new documentation hints the number of cohorts will be much higher. After all, the more cohorts there are, the more specific the data on their users can become.
There's no conclusive information about how small each cohort might be or whether some groups will be blended together if they don't have enough members.
The SimHash algorithm is meant to recalculate cohorts within the FLoC every week using the most recent information from the previous weekly cycle. This method would make cohorts only useful for short-term identification. However, the accumulated data would result in a more robust indicator of user habits over the long haul.