Google explores first-party replacements for third-party cookies

The company believes an interest-based cohort model could be nearly as effective as cookie-based ads.

Media buyers and planners, as well as their clients, have been trying to think of how they will be able to effectively and accurately target and measure ads once tech giants like Google and Apple succeed in their goals of eliminating third-party cookies from the web.

Today, Google provided an update of tools as part of its efforts to ensure advertisers have the ability to do their work effectively online, while still respecting user privacy.

Google has described “Federated Learning of Cohorts” as an “effective privacy-focused replacement” for third party cookies. The simple version is that, through Chrome, users will be assigned a “cohort ID” based on their web activity, which, in order to preserve anonymity, can only be used by advertisers if a minimum number of other users also have the same ID. From here, companies observe the behaviour of a cohort of similar people, instead of observing the browsing behaviour of individuals, describing the API as hiding individuals “in the crowd,” uses on-device processing to keep a person’s web history private within their browser.

Google claims advertisers can expect at least 95% of the conversions per dollar spent with FLoC compared to cookie-based advertising. Compared to random groupings, these cohorts achieve 350% improvements over ad recall and 70% improvements in precision in the company’s initial tests.

In an example, the company also suggests this kind of targeting could be more effective than that based on web history – two websites could easily attract users that are interested in cars and bikes, but targeting done based on visiting either site will result in a mix of interests, while FLoC-based targeting could ensure ads can reach only bike enthusiasts.

The company plans to make cohorts available in March for brands, marketers and advertisers for further testing, with cohort testing with advertisers in Google Ads expected in Q2.

“A key part of this process is engaging with the community to experiment with these new mechanisms, test whether they work well in various situations, and develop supporting implementations,” Chetna Bindra, group product manager, user trust and privacy at Google, tells MiC. Last week, the company launched new experiments around how real-time bidding may work in a cookieless world, inviting ad exchanges, DSPs and advertisers to participate.

FLoC is a part of Google’s Privacy Sandbox, which has proposals for all the ways marketers may be able to develop and deploy their respective audiences without utilizing third-party cookies, which will be eliminated from the Chrome browser next year.

As part of the Sandbox, Google has also published a new proposal called FLEDGE (First Locally-Executed Decision over Groups Experiment), specifically focused on remarketing. An expansion of last year’s “TURTLEDOVE” proposal, FLEDGE takes into account industry feedback, including the idea of a “trusted server” that’s specifically designed to house information about a campaign’s bids and budgets. Though it is still implementing, some changes are needed to provide features for web-based advertising, Google plans to make FLEDGE available for testing through origin trials later this year with the chance for ad tech companies to try using the API under a “bring your own server” model.

Chrome has proposed several technologies within the Sandbox that would enable marketers and partners to measure their campaign performances without third-party cookies, Bindra notes. An example would be event-level reporting that allows bidding models to recognize patterns in the data, and aggregate-level reporting which supplies accurate measurement over groups of users.