Determining the nature of the behavior that can be predicted using various variables, the accuracy of such predictions, and the timescale in which that behavior can be predicted are all areas of interest to researchers, academics, and practitioners. For businesses, accurately forecasting sales is of immense value. This thesis builds on established literature and aims to predict the sales of the sports brand Nike. In this attempt, the thesis examines the forecasting capabilities of search engine volume index and customer satisfaction index. Specifically, the thesis aims to answer the following research question: “To what extent can search volume index, derived from Google Trends, and customer satisfaction index, derived from the American Customer Satisfaction Index, predict the quarterly revenue of the sports brand, Nike?” In order to answer this question, two statistical models and one Machine Learning model were trained and tested. The modeling techniques were chosen based on established literature, the intuitiveness of the techniques, and the findings in the exploratory data analysis. The data and models were visualized, analyzed, and processed using R-Studio and Microsoft Excel. In the results, this thesis found that two out of the three models performed satisfactory results. These were the hybrid models, SARIMAX and NNAR, which both contain features from more than one forecasting technique. Moreover, these results are in line with prior studies approaching prediction problems in the fashion industry. Also, the models demonstrated the usefulness of the Google Trends index and customer satisfaction index in predicting sales. However, the customer satisfaction index has some considerable limitations in this study, and its contribution must be seen more skeptically. At last, this thesis found that the optimal time lag was one month for the search term index and one quarter for the customer satisfaction index. In a possible explanation, this thesis points to the purchasing process model, where customer satisfaction affects the purchase prior to the information-seeking. However, it acknowledges that further research is needed and points to several possibilities. The contribution of this thesis is two-fold. The first is methodological, where some of the approaches and techniques in predictive problems have been further examined. The second contribution is an applied one, where the proposal of this thesis can benefit companies such as Nike and the academic field of predictive analytics.
|Educations||MSc in Business Administration and E-business, (Graduate Programme) Final Thesis|
|Number of pages||70|
|Supervisors||Niels Buus Lassen|