Decoding consumer preferences in wine

: predictive analytics and machine learning in analyzing Portuguese wine consumer ratings

Abstract

The increasing significance of electronic markets and platforms has revolutionized how consumers interact with and purchase products such as wine. This change provides more information availability to distinguish it from others. Consequently, it is easier to convey a distinct message for characterful products. Nevertheless, it also results in drawbacks like information overload and a loss of the ability to differentiate. This thesis uses advanced analytics to decode the most important factors for consumer preferences in the Portuguese wine market. At the same time, it addresses the challenges and opportunities presented by the information-rich environment of electronic marketplaces. Specifically, I conducted a study to identify, using predictive analytics tools, the essential qualitative product features of Portuguese wine that matter for consumer satisfaction. To do this, robust predictive models were built using individual consumer review ratings and the descriptive characteristics of Portuguese wines from the electronic marketplace and platform Vivino. Ultimately, relative feature importance was determined using Random Forest, AdaBoost, Gradient Boosting, and XGBoost. Additionally, the study incorporates user comments via topic modeling into the predictive models. As a result of this analysis, the study revealed consumer participation, user engagement, sensory perception, generalization, and price description as driving factors for consumer satisfaction. In summary, all models demonstrated similar outcomes, recommending a focus on extrinsic rather than intrinsic product attributes to differentiate from other product groups. These findings can be used further for strategic market decisions and research.

Date of Award	23 Jan 2024
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Nicolò Bertani (Supervisor)