Three Essays on Big Data Analytics in Marketing

This thesis investigates how the immense amount of real-time and retrospective data can contribute to marketing theories and practices in branding, advertising, and communications. In the first essay, brand management in social media is studied due to its vast potential as well as the companies’ interests in utilizing it for branding purposes. More specifically, the research examines the sentiments toward a brand, via brand authenticity, to identify both the reasons for positive or negative sentiments on social media and its polarity. Practically speaking, while firms need insights about users’ sentiment towards their brand, knowing just that it is a simple positive or negative sentiment does not provide them with enough information. They might ask themselves why users like us or what the reason is behind the negative sentiment. Using three qualitative studies, along with the latent semantic analysis (LSA) and the support vector machine (SVM), the findings illustrate the effectiveness of the proposed procedure of brand authenticity sentiment analysis to predict both the brand authenticity dimensions and their level of sentiment. The second essay is concerned with the immense amount of locational and destination-based data for advertising purposes. That comes from the fact that more and more apps have access to users’ locations and destinations. As the context, the sharing economy is selected to test if a ridesharing app, i.e. Uber, can alter one’s destination by providing relevant destination-based ads. Examining construal levels (i.e. spatial distance and cultural distance) show the effects on the relationships between attitudes towards destination-based advertising and redemption of marketing incentives as well as app reuse intention. Two experimental mock apps that mimic a well-known ridesharing app in North America are used. The findings provide implications for the construal level theory, the theory of planned behavior, and how practitioners can use it to alter planned behavior, i.e. planned destinations. The third essay is also in the context of the sharing economy. Building on the speech act theory, the sale description impact on consumers’ social cognitions of service providers is investigated to find how they could be used to generate new content. Findings suggest the role of linguistic concreteness, service provider type, and sentiment analysis on the perceived level of warmth and competence of the service providers. Using two text mining methods (frequency analysis and modified LDA), the differences between the four kinds of property descriptions and their hosts are explained. Then, the findings are employed to better train the Natural language generation (NLG) algorithm. This shows how Long-Short-Term Memory (LSTM), as well as cleaned input, could help service providers generate more engaging content to communicate their service. The research findings on each of these streams could contribute to several theories and facilitate further inquiries into big data analytics in marketing. This research also provides marketing practitioners with reliable and valid theories, models and decision support systems to gain insights and propose appropriate strategies to strengthen their firm