That's what I want to know. I would assume the article is talking about monitoring which IP's request which domains, but the overwhelming majority of requests that a registry sees have to be from downstream resolvers, right? In which case they don't know who's using the site, they only know that "somebody who uses this resolver" is using the site.
In most cases there are probably multiple levels of resolvers, so it might even be as coarse as "somebody on the west coast of the US is using this site," which isn't terribly useful.
By what mechanism is this possible?