Collective Intellect is developing new and innovative techniques to use the blogosphere as a predictive tool for reaction-based events. We’ve had predictive success in single state primaries and caucuses, as well as Super Bowl advertising; but the massive amounts of conversation in different topic areas surrounding Super Tuesday seemed to have too many variables which flooded and diluted the data in our current methodology. Above all else, these predictions are serving as an ongoing experiment to perfect CI’s ability to predict reaction-based events using social media.

A review of the data revealed that more of the projections would have been correct if we’d only used state-based blogs; unfortunately the data sample of 40-50 state-based blogs was too small on which to depend completely. Examples from past projections suggest that averaging state blogs with an overall national reading would balance any skewed data from the smaller state samples.

The overwhelming number of national “events” was responsible for skewing the data it was supposed to help balance. Below is an example of the two data sets for Missouri:

Ideally, we would find and use every state-based political blog, and forget about the national ones, but that is currently too time-consuming. We will continue tracking and forecasting the next few primaries, and publishing any relevant data in advance– sticking our neck out there, as Forrester’s Peter Kim put it– to ensure that this method will continue to work for individual races, and we will expand our current methods to account for the bigger events in the future.

Social media has proven incredibly valuable “to keep the marketer in tune with consumer moods in real-time,” but social media predictability is proving not only possible, but an invaluable tool for marketing, politics, advertising, and the corporate world. We used a similar methodology for Super Bowl advertising that proved to be an accurate gauge for advertisers who want to make sure their campaigns get optimal opportunity and value.

The share-of-voice graph above represents the top ten companies’ pre-game buzz and compares it to the top ten buzz 12 hours after the game. These companies’ positioning changed little if any after the spots ran, rendering our prediction – that the top four pre-game buzz leaders would also be the post-game leaders – off by only one company.

These results alone and what Collective Intellect has seen in past projects only reinforce our confidence in using the blogosphere to predict results. With a little more experience and tweaking, a real-time reading of consumer (and constituent) moods can be used to make accurate and consistent predictions for when those moods turn into results.

– Mark Lucier and Kevin Yordy

mark@collectiveintellect.com

kevin@collectiveintellect.com

Tags: Analyzing Consumer Generated Media, Brand Management, Mass Media/Culture //