There’s a massive difference between Farmville et al and the OFA app. Facebook had limits in place that constrained the ratio between the number of friend profiles that were allowed to be accessed and the number of people that had explicitly authorized the app. They were a black box on this issue, but for example I had apps that were stopped from accessing deep profile information of friends (likes, occupation, etc.) when I reached 10:1. OFA’s ratio was in the 200:1 neighborhood.

It was somewhat possible to overcome this by spreading the collection out of a number of months or years, which is what I believe the Kogan app did. But even in that case, with the data collection spread out over a long period of time, they had nowhere near the data that the Obama campaign was allowed access to. Finally, the CA data was years old, while the Obama data was allowed to stay fresh right up to election day because they had no API limits.

So these apps with much larger install bases than Obama could have ever dreamed of had access to less data than OFA did because their ratio was not allowed to be as ridiculously asymmetric as Obama’s was. They still had access to large amounts of data, but only because of their massive authorized install bases (which in all of the cases you mention were far larger than the OFA install base). But none had the entire US social graph - with the exception of OFA.