Predicting Net Promoter Scores with Text Mining

Attended the UC data mining symposium a few weeks ago, will be posting some things of interest that came from that.  A rep from SAS described work they had done with text mining.   Text is interesting because it is called 'unstructured' data, though text clearly has structure, though of a different type.  So it requires some specialized methods to tease out the structure.

That's not a new thing, we used methods called 'content analysis' to do that back in the 80s. In this very instructive example, SAS uses their text mining capabilities to predict promoter scores from from social data. A very useful thing.  Well done piece.

From Customer Risk to Corporate Strategy ( Full article requires registration)  Using Text Analysis and Predictive Modeling to Improve Promoter Scores

Two assets significantly influence success or failure of a company. Those are customers and their continued patronage, and employees and their knowledge (as well as their work productivity). This paper describes how to evaluate the likelihood of continued customer patronage versus the risk of losing it. It also acknowledges corresponding loyalty, measured in the form of promoter scores. The paper proposes a strategic analytic roadmap for how SAS enables you to use promoter score survey results to identify customer migration value. The strategic value your organization gains from such analyses enables you not only to understand why customers are or are not likely to promote your product or service – but also their likelihood to do so in the future.... "