I'm assuming you’ve scrapped most of the user profile of those 2M account. If so, it could be interesting to see a few other things with just that information, without having to cross the files with anything else. Things such as distributions by:

-Number of posts: for example ->  0,1,2,3,4,5,5-10, more. This is rather more specific for the immense amount of newbie accounts, and would allow us to see when they stop (after how many posts).

- Last Active :grouped by some sort of criteria  such as -> last 30 days, 30-90, 90-180, 180-365, > 365.  That would give us an idea, for each rank, how many are active. The “active” measure with this definition is not of a Boolean nature, but rather more like a “state”.

Of course, that is your prerogative if you see it fit and think it can throw-in some more information.

Thanks for the info so far. I’ve already referenced it twice today on separate posts to strengthen my point of view on two different matters.