Whatever data is available needs to be interpreted cautiously. We could become very enthusiastic about strategies that apparently have strong data support that actually show very little impact.
The graphs provided by Duncan show that something was phenomenally successful in driving traffic approximately 40 days ago. What the graphs do not show (because you need to extract other data to do so) is that the additional 2000 visits produced 18 new user registrations over that three day period. Of those 18 people, ONE has been "seen" on the site after 25 October.
Initial visits (eyeballs) are great. Registrations (clicks) are better. Regular users are what builds the site. Then we can move on to consider activity levels (voting, editing, asking, answering).
It is too easy to get excited about essentially meaningless statistics just because they can be gathered.