A lot has been made over the last year and a half about Google’s shift to remove third party cookies. In January 2020, Google announced the depreciation of these cookies from Chrome browsers by 2022. A move seen as one to begin prioritising user privacy over advertising dollars.
The announcement was met with initial panic that the ‘cookie apocalypse’ would bring an end to digital targeting capabilities as we knew it. The more pragmatic saw it as a change that was desperately needed as a ‘cookie’ was an archaic piece of code that was never designed to do all the things it became required to do.
Then there are those who saw this, less as an end of an era of audience targeting, and more that Google would evolve cookies into something more advanced. How could Google essentially kill off a massive revenue-generating part of their advertising empire? (This would obviously not be unprecedented, simply check out Killed by Google to see some examples).
While there was some radio silence from Google throughout 2020 around potential solutions to replace the cookie, a lot of work was seemingly done behind the scenes. Google began testing a potential solution, referred to as ‘Turtledove’, which looked to replicate some of the features that cookies provided, while ensuring privacy of user data is maintained.
An ambitious goal to be sure. Turtledove set out to establish a new system that would connect information about the user to the browser, rather than being held by the advertiser. Advertisers could then look to serve ads based on user interest, but would not be able to combine these interests with any other information about a user, such as content they are looking at or who they are.
With bird names apparently a major focus, Google worked closely with a number of their partners to develop this solution further. Criteo’s ‘Sparrow’, Magnite’s ‘Parrot’, NextRoll’s ‘Tern’, and Google Ads’ ‘Dovekey’ were all connected to Turtledove in an attempt to make it the best viable solution. This has worked to potentially facilitate Google’s cookieless solution, FLEDGE and FLoC.
FLEDGE, while a devolutionary term compared to the origins of initial products within Turtledove, looks to advance the early tests and provide audience-based targeting opportunities in line with what is available through cookies. More specifically, FLEDGE is designed to provide remarketing opportunities for advertisers.
FLoC, or Federated Learning of Cohorts, appears to be Google’s solution to audience targeting without cookies. The vision for this solution is that user information is anonymously collated, based on digital consumption, to build out an understanding of interests and behaviours. Based on these signals, people are put into one particular ‘cohort’ with all those people who share similar interests and behaviours. A user can only be in one cohort at a time, and these groups would be refreshed weekly.
If you are thinking to yourself right now “that doesn’t seem overly different to what we are already doing”, or “how is that more private than what we do now?” then you probably won’t be surprised to hear that European auditors agreed with you and pushed back on this being a viable privacy-first solution.
These auditors did not accept the early version of FLoC, which Google presented to them in early 2021. The decision was that FLoC does not do enough to prevent companies from ‘fingerprinting’ users, essentially allowing people to extrapolate information down to an individual user.
Google went back to the drawing board after this ruling, announcing they would be delaying the depreciation of cookies until 2023. Since then, Google has revealed a detailed timeline for these changes and when each step in the process will begin. The timeline was a welcome guide, as it was something that was conspicuously absent when they first announced the depreciation at the start of 2020.
Google has constantly maintained that their ‘cohort’ construction is the future of audience targeting. However, in August there was a potentially huge development around the FLoC offering, as one of Google’s tech lead managers, Josh Karlin, stated that it “might make sense to stick to topics instead of cohorts”. This statement implies that there may be some within Google that are not aligned with the FLoC offering, or that it might simply be getting too hard to develop a solution that provides targetable audience data in a privacy-first world.
Contextual targeting has been around for a long time now. The irony is that the usage of contextual targeting within digital was diminished with the rise of audience targeting, and now with audience targeting being diminished we could revert back to contextual targeting.
Without audience targeting, the big underlying revelation here is that it could create a more even playing field for other DSPs looking to challenge for Google’s crown. Google has managed to develop a stranglehold on performance media over the years by offering rich audience targeting at scale. If they lose this, they may see increased competition from a number of other DSPs that can offer similar, or better, contextual based solutions.
Oracle’s Grapeshot solution, for example, has been the leader in contextual targeting for a number of years now. While Google’s contextual solution focuses on the existence or nonexistence of specific keywords, Grapeshot explores for relevant keywords but also looks at the semantic terms around them to understand the context with which the word is used. In functionality, Google would block an article because it mentions the word ‘gun’, while Grapeshot may not block it as it has assessed that the article is actually about DIY using a ‘glue gun’, for example.
There is still a lot of time for Google to develop the right solution for them, with 2022 now outlined as the year for testing, before things begin rolling out in late 2022/early 2023. The decision to deprecate cookies came as a bit of a surprise to many, but seemingly everyone agrees it is the right move to ensure user privacy. As you can see, there have already been a lot of twists and turns that have gotten us to this point, and there are bound to be a few more before the rollout in 2023.