These kinds of articles remind me of the Black-Scholes financial model of LTCM fame, in that the very understanding of and subsequent ubiquity of resulting analysis invalidates the model itself. Multi-agent social systems are inherently recursive in this way. I suppose the data collected is relevant up until a particular threshold level of social media marketing experts saturate the strategies developed as a consequence of these findings. A more interesting analysis would include the life cycle validity of the data collected. Just some food for thought.