I think there are a couple of quite important things there.
With regard to the data collected from Dr. Kogan's survey, that would seem to be primarily psychological profiling of people who completed the survey, but then crucially linked to their Facebook profile. I think that's important, because if you then take that data and add it to other datasets they may have acquired, and the email lists that the campaign group itself had acquired, you can then not only merge all those datasets together, but you can also tie it back to the Facebook user profile, which helps you to retarget the individual back through Facebook, where Facebook is effectively acting like a giant telephone exchange, and you're programming the data in, not generically to categories of people, but to named individuals.
Mike Schroepfer was saying that the Kogan datasets wouldn't have included email addresses. That could be right. They could have just found those emails in other ways. But from the evidence that we had yesterday from Chris Vickery, I think it's clear that you can have many different datasets that you process together through your targeting tool and then use that for the basis [Technical difficulty—Editor] Facebook.