It’s common knowledge that Google Chrome will deprecate third-party cookies in the near future — likely in 2022. Possible alternatives to preserve addressability are being floated, among them initiatives springing from Chrome’s Privacy Sandbox. But are these alternatives, and the many others being proposed by adtech and data vendors, compelling to marketers?
To make sense of Federated Learning of Cohorts, or FLoC (and Fledge and Turtledove), we started with what the developers on GitHub’s Web Incubator Community Group are saying. A FLoC cohort is a large number of people grouped by the browser based on browsing history. “The browser uses machine learning algorithms to develop a cohort based on the sites that an individual visits. The algorithms might be based on the URLs of the visited sites, on the content of those pages, or other factors. The central idea is that these input features to the algorithm, including the web history, are kept local on the browser and are not uploaded elsewhere — the browser only exposes the generated cohort. The browser ensures that cohorts are well distributed, so that each represents thousands of people.”
Birds of a feather FLoC together
What is generated, in other words, is an addressable cohort of anonymous users which is not uploaded (or sold or traded). That’s the idea, anyway. As the GitHubbers point out, this approach is not risk-free: it does, after all, involve tracking and data collection — quite possibly including sensitive data — and there is the potential of tying cohort information to personally identifying information. But it certainly seems to be an improvement on directly tracking an individual’s behavior, which is what dropping third-party cookies seeks to do.
We had to speak with some experts to get a handle on all this.
Michael Schoen is SVP and GM Marketing Solutions at Neustar, the identity resolution provider. Neustar last year launched its own alternative approach to identifiers — not a solution, but a collection of strategies it calls Fabrick. “FLoC is going to continue to develop into a couple of related proposals. Already there’s been an evolution that is referred to as Fledge, which is very similar.” (Fledge, in turn, evolved from Turtledove.) “Slight technical differences, but to the lay person they’re going to look pretty much the same. They’re a family of solutions that leverage techniques that come from differential privacy to allow useful insights to come from data that only gets exposed in an aggregate way.”
Essentially, these are techniques for creating lookalike audiences, said Schoen. “The early results that Google has shared are promising. You need to consider those results with a couple of teaspoons of salt, because often they’re done in comparison with completely untargeted advertising — and no-one does completely untargeted advertising online. The reality is, marketers don’t care what individual users are doing, they care that their marketing dollards are used effectively.”
Schoen seemed cautiously optimistic: “What Google has shared as an update is reflective of them making some progress towards their aspiration. They want to ensure that the use cases for the third-party cookie that marketers care about have some form of adequate replacement before the third-party cookie goes away.” Neustar will incorporate whatever emerges from FLoC into its Fabrick solution.
Neustar has been playing in the Privacy Sandbox too, offering something it calls PeLICAn (Private Learning and Inference for Causal Attribution). “The PeLICAn prospoal was similar to FLoC.” It’s a way of providing non-click-based attribution by leveraging the propensity data available in the FLoC cohorts. “We’re optimistic it will end up being part of the Privacy Sandbox solution set,” said Schoen.
What contextual advertising can’t do
What remained initially unclear, however, was the advantage of FLoC over some precise and refined form of contextual advertising, where reasonable assumptions can be made, say, about the properties of the cohort following football news or searching for camping equipment. “There are clearly some products that have a natural affinity with content,” Schoen agreed, “although I continue to believe we’ll see a resurgence of contextual targeting because of these changes. But there are some products that don’t have a natural affinity to specific pieces of content; a lot of CPG goods do not necessarily have an obvious correlation to consumers of sports or news content, etc. There may end up being a correlation, but the correlation is not obvious to begin with.”
Secondly, said Schoen, there’s a need for lookalikes simply to extent reach. “There literally may not be enough ad placements alongside, say, American football content, so to identify those consumers when they’re consuming other entertainment content can still be useful.”
The precise use case for third-party cookies has been to enable targeting of consumers of a specific type of content in a different context. “With FLoC you’re going to see something similar, you just do it in a way that is more opaque.”
Advertisers may be confused and concerned
A more skeptical view came from Patrick O’Leary, founder and CEO at boostr, an enterprise software company that makes solutions for the sell-side of the ad industry, including pure digital, out-of-home, broadcast and audio media. Unusually, it combines a CRM and advertising order management in a single platform.
“It has a lot of implications,” he said, “like can you cross-device target somebody? The modern consumer has their phone and their desktop browser, and may be watching TV at the same time, and people have been trying to solve that consumer journey as they switch across devices. That’s an important question about FLoC.”
Despite the many types of cookie, O’Leary nevertheless regards the cookie as a single standard that can be applied across formats, devices and hardware. “Now we’re unbundling. Does this just fit Chrome? What’s going to happen on my Apple or Android device, on my Roku or set-top box? Verizon Media has announced they they have an identity solution, the DSPs have identity solutions. If I’m an advertiser, I’m going to be confused and concerned that none of this is going to be interoperable.”
The feedback O’Leary is hearing from publishers and advertisers is that they’re trying to understand this new identity environment and how it can be leveraged. In addition to the operational challenges, there is real concern about revenues and margins. “Will the money pull out of the market while everybody re-tools? There are a lot of unanswered questions on change management from Google.”
O’Leary, like Schoen, expects a resurgence of contextual advertising, but has a more positive outlook on it. “I’m a mountain biker. When I go to the mountain bike website, I appreciate the ads because they’re products and companies I care about, in an environment where it makes sense. I don’t want to be on the mountain bike website and see an ad for a diaper for my kid who grew out of diapers three years ago.”
Contextual advertising, said O’Leary, might actually perform as well as, or better than, “all that stuff” advertisers have been buying from “middlemen in the adtech world.”
This story first appeared on MarTech Today.