FloC, FLEDGE and Topic API Failure
The certain death of third-party cookies, marketers seek AdTech that aligns with increasing polices while actually providing good results for their businesses. The kind of regulation is GDPR, which is a factor of EU privacy and human rights laws. The general consensus is that these developments are vital to defend internet users, but the changes create a massive challenge for marketers still reliant on third-party data in advertising campaigns.
The majority of the top web browsers such as Safari, Edge and Firefox took action to block third-party cookies, but Google Chrome postponed this to mid 2023. With huge browser market share of around 67%, Google’s actions have a substantial role in the direction of the sector. Moving just as fast as Google are Oracle and IBM, with AI technology focused on targeting and advertising to users based on their shared behaviors.
Google’s primate attempt at a solution was Federated Learning of Cohorts (FloC), which aimed to maintain user privacy by masking individuals in a crowd of users. This new method of tracking end user behavior was meant to ensure advertising relevance for consumers while hiding individual user information from advertisers. The solution earned rapid backlash, as concerns that their behavioral cohorts contain enough entropy to pinpoint an individual. At the same time, behavioral cohorts open the door to discrimination and the degradation of civil liberties, but the results during the test were questionable.
The mediocre results pushed Google to abandon FloC and attempt to replace cookie tracking with a new “solution” called Topics API and FLEDGE. Another attempt at replacing cookies, FLEDGE makes ad auction decisions in the browser as opposed to at the ad server level. In theory this will lead to less user data being harvested for the purpose of building user profiles. While this move makes sense given the vulnerabilities of FloC, there are effective and secure cohort solutions that are not behavioral, but instead built around engagement with keywords.
Cohorts in Advertising 101
Cohort-based advertising is implemented by targeting anonymized segments of internet users united by a common criterion. An example of such criteria would be “All customers that purchased an iPhone in 2021” or “Every customer acquired through last month’s PPC campaign.” Cohorts deliver insights that can improve short-term marketing efforts, as behaviors change on a daily basis, and help group users into baskets of shared characteristics.
A common misconception around cohorts is the false comparison to general demographics such as age, gender, and income. In actuality, cohorts separate groups of people within their demographic, and instead sort by shared life events and time. This is where personalization comes in, enabling advertisers to create tailored campaigns based on these events and specific moments in time.
This functionality has led many marketers to explore cohorts as a way to group consumers based on their behavior, which is easily classified by events and moments in time. This changes the nature of audience targeting, moving away from the need of uncovering needles in a haystack. Using cohorts, as data is collected and more campaigns are deployed, the accuracy increases in theory.
We are still in a very experimental phase when it comes to advertising cohorts. Marketers are beginning to recognize where to utilize this approach and where it falls short. The early focus on behavioral cohorts seemed to have some promise, but now the cracks are becoming apparent. This doesn’t mean that cohorts should be written off, as there are ways to re-conceptualize cohorts and deliver real results in alignment with GDPR and CCPA.
Shortcomings of Behavioral Cohorts
While advertising cohorts are still relatively new, the shortcomings of behavioral cohorts are becoming abundantly clear. Experienced marketers and advertisers are witnessing the downsides of utilizing this type of approach, and how it is leading to real ethical concerns.
Generalizations of Broad Audience
Behavioral cohorts are a form of aggregated data, failing to provide the granularity that 1:1 targeting can. This fact leads to concerns around generalizations, as the data collected can only provide a broad-brush stroke when it comes to pinpointing performance results related to specific audiences. Broad pieces of information aren’t as helpful when examining effectiveness and developing marketing strategies.
Assembling groups based on behaviors can lead to the generalization of an audience and reduce the potency of an advertising campaign. As data is collected and campaign optimization is carried out with new cohort audiences factored in, it is possible to trend in the wrong direction based on assumptions being made.
Using certain types of behavioral cohorts opens the door to the possibility of browser fingerprinting. This practice involves the collection of many small pieces of information that might seem discrete individually, but when combined create a unique identifier for that user. As a user’s browser begins to differentiate itself from others through their actions, it becomes easier to fingerprint.
Fingerprinting is one of the main push backs that caused Google to abandon FloC, despite their original plans. In theory, as a user is grouped into a cohort with thousands of others, the cohort ID doesn’t allow for a single user to be distinguished from the pack. However, if trackers start with a cohort ID, there are only a few thousand other browsers for it to move through to recognize a unique identifier. This privacy issue is notoriously difficult to defend against, with major browsers spending years working to reduce the potential threat. One of the ways to reduce the risk of fingerprinting is to remove sources of entropy, which is what a behavioral cohort is.