4 ms·

Google does and the cohorts are opaque and data generated based on history. They're not actually using this specific approach but roughly what they're doing is

Google does and the cohorts are opaque and data generated based on history. They're not actually using this specific approach but roughly what they're doing is like taking your browser history, representing it as some vector of numbers and then clustering nearby points together and giving each cluster a cohort id. Cohort ids are also updated periodically (roughly weekly). The clusters are most likely to be surface level interests. The id itself doesn't give you that though and you'd want many users with same cohort ids, correlate there behaviors, to try to guess what underlying browse trend it means which is a lot better than the current state of things of effectively having a user id with 3rd party cookie.

matthewdgreen 5y ago
It seems to me that decoding FLoC IDs will be a big business, given their value. So people will do their best to reverse the system and keep an up-to-date table for at least some of the more valuable/interesting cohort IDs, if not all of them. In the worst case you can do this just by paying volunteers with known interests and recording their FLoC IDs. Presumably that will be done enough that at least some of the data is available for sale. My further understanding is that FLoC IDs will be available to any website. So even if you’re too small to run a global tracking network, you can grab the FLoC ID from any visitor and, assuming that decoding services exist, turn that into useful data instantly.