[sfifs]

A reasonably dense geo, timestamp and IP address event stream over time (ie. Multiple apps sending over time reasonably frequently) should have sufficient information to associate ids. However most ad tech companies including I believe Google and FB are very leery of doing this "Harry" type of stunt and arrange data pipelines in a way that this cannot be done at any aggregation less than a few hundred IDs and export anything beyond segment data to prevent re-identification. I'm very surprised these people are touting it

[mastazi]

> However most ad tech companies including I believe Google and FB are very leery of doing this "Harry" type of stunt and arrange data pipelines in a way that this cannot be done at any aggregation less than a few hundred IDs

I would be interested to see any sources about this (I’m not questioning what is said in the comment, just genuinely curious about this topic)