Word Travels Fast: Using Text Analytics to Drive Insight from Unstructured Power Sports Data

This paper is a client-based text analytics project performed on public data from the Internet. The client, Polaris Inc., is a leading manufacturer of outdoor power sports vehicles and equipment. The client was interested in mining text data from industry forums and Twitter to capture helpful insights that wouldn’t otherwise be seen in internal feedback data streams. It is thought that external insights from public data could someday be combined with internal metrics to provide a holistic view of quality throughout the manufacturing value chain and enable the business to better align its efforts to the voice of the customer. The client was also interested in the results of a competitive benchmarking activity conducted between itself and two other major brands in the market space: Can-Am and Arctic Cat. Text data was obtained through forum scraping and tweet harvesting in the open source statistical programming language R. The data was subsequently cleaned, prepared and analyzed for trends through topic modeling, clustering and sentiment analysis. A competitive benchmarking activity was conducted through a comparison of major signals. Several ideas were discovered and discussed, such as: the quality of specific vehicle components, vehicle operation guidelines, and the emotional perception of brand ownership. The project’s conclusion consisted of several suggestions for follow-up studies as well as directly addressing business questions on the client’s behalf.