BeschreibungFor marketers it is important to know what their customers think about their products and services and whether or not many other (potential) customers share or value these opinions. One readily available source of customer opinions about a product is online customer review data. We analyze a data set from Amazon that contains reviews on various brands of tablet computers. We use the Latent Dirichlet Allocation (LDA) to identify topics within reviews and investigate how the various topics affect the number of helpful votes of reviews for tablet computer brands. This way, we like to shed light into two types of research questions: First, which particular review contents/topics affect the number of helpful votes of a review and if so, is the effect positive or negative? Second, which type of count model best models this relationship? The modeling is carried out as follows: The identified topics serve as predictors for the various types of count models, such as Poisson, Negative Binomial models, and others. From our LDA, we identify seven different content categories like, e.g., usage behavior or brand comparison. We base our decision on the optimal count model on model criteria. Some review contents (e.g., usage behavior like reading) have a positive impact on helpfulness for one brand but no effect for another brand. Based on these results, we give some examples on how marketers can benefit from our findings.
|Zeitraum||20 Juni 2019 → 22 Juni 2019|
|Ereignistitel||INFORMS Marketing Science Conference|