Predicting Smartphone Upgrades through Deep Learning Classification Models

Informações

Código: MKT258
Divisão: MKT - Marketing
Tema de Interesse: Tema 05 - Teoria, Epistemologia e Métodos de Pesquisa em Marketing

Autores

Vinicius Andrade Brei (Prog de Pós-Grad em Admin/Esc de Admin – PPGA/EA/UFRGS - Universidade Federal do Rio Grande do Sul) brei@ufrgs.br
Leonardo Nicolao (Prog de Pós-Grad em Admin/Esc de Admin – PPGA/EA/UFRGS - Universidade Federal do Rio Grande do Sul) leonardo.nicolao@ufrgs.br
Maria Alice Pasdiora (Prog de Pós-Grad em Admin/Esc de Admin – PPGA/EA/UFRGS - Universidade Federal do Rio Grande do Sul) mariaalice.pasdiora@gmail.com
Rodolfo Coral Azambuja (Prog de Pós-Grad em Admin/Esc de Admin – PPGA/EA/UFRGS - Universidade Federal do Rio Grande do Sul) rodolfoazambuja@gmail.com

Resumo

This manuscript explains which variables are more relevant to predict which consumers will replace their products (smartphones) in the future. Data was collected through a longitudinal consumer panel, which measured upgrade decisions and independent variables classified into eight main groups: ownership, enjoyment, desire, perceptions about the smartphone, context, individual traits, and demographics. We tested two types of non-linear, state-of-the-art machine-learning models to explain upgrade behavior: Extreme Gradient Boosting (decision-tree) and two integrative Deep Learning models. Results provide a comprehensive, yet parsimonious model showing ownership, enjoyment, and context variables as the most relevant to determine which consumers are more prone to replace smartphones. Our findings enhance previous understanding of upgrade decision theory by taking a holistic approach and bridging different theoretical accounts. Further, results contribute to marketing theory and practice, shedding new light on the understanding of consumer decision making when upgrading products. Managers may apply findings to identify which variables are more relevant to influence consumer decision process regarding new products. Therefore, they can adapt their marketing strategy. Consumers, in turn, may be aware of how their behavior can be shaped by marketing actions and, then, react accordingly, making better decisions.

Abrir PDF