The exact targeting options that will exist after cookie deprecation are unknown, because Google’s Privacy Sandbox proposals are all on a delayed timetable. The main Privacy Sandbox targeting-related proposals are:
- Federated Learning of Cohorts (FLoC). Machine learning within the browser analyzes the web history of that browser and matches it up with thousands of other browsers with a similar web history. Each cohort, or FLoC, is assigned an identifier that can be used to target ads, and the assignment is updated regularly.
- First Locally-Executed Decisions over Groups (FLEDGe). This is the Privacy Sandbox proposal to replicate retargeting.
The Privacy Sandbox proposals are meant to cover impressions to users that aren’t authenticated (i.e., logged in) or enriched with data in some other fashion (e.g., as part of a publisher-created audience segment based on behavior over time on a single domain). We expect these proposals to eventually constitute a “bottom tier” of targeting options, where advertisers can continue to reach users at a one-to-one or otherwise microtargeted level in some logged-in environments and can use probabilistic methods of identifying users in some other environments as well.
Google’s delay wasn’t exactly unprecedented. Apple had originally announced its AppTrackingTransparency (ATT) framework would go into effect in September 2020, and that change was put off until April 26, 2021. Under ATT, users must opt in to allow advertisers to track them across devices, on a per-app basis. Previously, users had been able to opt out of sharing their Identifier for Advertisers (IDFA). App developers and other advertisers expected opt-ins to be low.
Mobile measurement partner AppsFlyer reported that during the last week of June, 40% of US users who were shown an ATT prompt opted in to tracking, translating to an average of 36% per app. Not all users are shown the prompt—some aren’t, and are simply set not to be tracked, like those who had already turned on Limit Ad Tracking (LAT).
AppsFlyer also reported finding that IDFA availability could be affected by the way app developers set up their SDKs to load. If developers make sure to call their attribution SDK a few seconds after displaying the prompt, they can make sure to capture all possible IDFAs. AppsFlyer reported that at the end of June, apps that made sure to trigger its SDK after the prompt had a 25% IDFA availability rate in the US, compared to a rate of 14% among apps where the SDK wasn’t timed this way.