Google’s been busy in the Privacy Sandbox.
Consider: Google’s Chrome browser has a new bird-themed proposal called FLEDGE that builds on TURTLEDOVE. The company is also making its proposal for interest-based cohorts – or FLoCs [aka, federated learning of cohorts] – available for public developer testing starting in March. And finally, on Monday Google shared the latest results of its own experiments with the FLoC API.
Google says its simulations demonstrate that with cohorts marketers can expect to see around 95% of the conversions per dollar spent when compared with cookie-based advertising.
FLoCs are created by using on-device machine learning to group users into cohorts based on their browsing behavior so that individuals are, at least theoretically, indistinguishable from the other people in their cohort.
“Doing something like that was counterintuitive just nine or 10 months ago,” said Chetna Bindra, Google’s senior product manager for user trust, privacy and transparency. “But the performance we’re seeing with FLoC shows that it’s nearly as effective as third-party cookies.”
For the birds?
There’s been some skepticism in the advertising community, however, about the viability of FLoCs as a replacement for third-party cookies.
In October, Google released early results from its FLoC tests indicating that interest-based cohorts generate big improvement in recall and precision over random clustering.
But, as James Rosewell, CEO and founder of 51Degrees and a vocal member of the W3C’s Improving Web Advertising Business Group, pointed out in an AdExchanger column shortly thereafter: “Who spends money randomly?! It appears that Google, which is known for its algorithmic prowess, somehow does not think marketers and publishers will see through this.”
According to Bindra, the goal of Google’s previous test was primarily to analyze different types of clustering algorithms. More recent testing has focused on seeing how interest-based cohorts stack up against cookies.
The results reinforce Google’s conviction that FLoC and other proposals within the Privacy Sandbox “represent the future of how our ads and measurement products will work on the web,” Bindra said.
Google declined to share, though, whether it plans to deploy any of these same or equivalent cookie-replacement technologies across its own services, such as YouTube or search.
Getting ready to fly
But putting aside Google’s O&O, the first widescale test of how the FLoC API will function on the open web is getting ready to kick off through a new proposal called FLEDGE from Chrome that was recently added to the Privacy Sandbox.
FLEDGE – like fledging, get it? – outlines an early prototype for ad serving based on Chrome’s original TURTLEDOVE framework and encompasses a bunch of different components from other sandbox proposals, including Criteo’s SPARROW, Dovekey from the Google Ads team, Magnite’s PARRROT and Nextroll’s TERN.
[Quick primer: TURTLEDOVE suggests moving the ad auction into the browser and serving ads based on FLoCs rather than cookies. In response, SPARROW called for placing control over bidding, rendering and reporting with an independent, trusted third-party server. The Google Ads team responded with Dovekey, which introduces the gatekeeper concept from SPARROW into TURTLEDOVE. PARRROT proposes refining TURTLEDOVE further by allowing publishers to retain control of the auction. And TERN calls for reducing the number of background network calls a browser will need to make to a DSP.]
“There have been marked signs of progress here, and FLEDGE is trying to take the best of all of these in order to think through which will be the critical features for web advertising while also ensuring that privacy protections are in place,” Bindra said.
Chrome plans to make FLoC-based cohorts available in March through origin trials, which allow developers to safely experiment with web features, and to begin testing them with advertisers in Google Ads in Q2.
Fully fledged FLEDGE experimentation could start as soon as Q2.
But if you thought you were done with bird proposals for the moment, well, you’re not. Chrome also recently published an anti-fingerprinting proposal called Gnatcatcher. (Gnatcatcher, by the way, refers to a family of around 15 species of small insect-eating New World birds … the more you know.)
Chrome is still refining its Gnatcatcher proposal, which allows groups of users to send their traffic through a privatizing server with the intent of hiding a device’s IP address so that it can’t be used for targeting but is still available for what Bindra called “legitimate purposes,” including spam and fraud prevention.
Testing, testing
The plan from here on out, she said, is to continue testing, experimenting, talking with publishers and advertisers and gathering feedback from discussions at the W3C.
As that happens, Google is well aware that the clock is ticking down towards 2022.
“We’re seeing a lot of engagement and we are making a lot of progress,” Bindra said. “We know this is an aggressive timeline, but we will only do this once we have privacy preserving mechanisms in place.”
What’s unclear, though, is whether Google will look to achieve a minimum level of stakeholder consensus within the working and community groups of the W3C before actively deploying any of the sandbox APIs in Chrome.
Also unclear is the potential effect that the investigation launched by the UK’s Competition and Markets Authority (CMA) in early January into Google’s forthcoming third-party cookie phaseout could have on progess and eventual deployment of sandbox proposals. The investigation was triggered after the CMA received several complaints about how Google’s Privacy Sandbox proposals could impact competition.
Although Bindra said she couldn’t comment on a particular investigation, she did say that “we welcome the CMA’s involvement.”
“This is a huge evolution and we’re looking forward to working together with the industry and the CMA to continue to develop new proposals,” she said. “We understand that this is a very big shift for the ad-supported web.”
