O efeito de recomendações e argumentos de prova social na intenção de compra e experiência do cliente no comércio eletrônico

• Ano de defesa: 2016
• Autor(a) principal: Fabio Roberto Ferreira Borges
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Prova Social; Comércio Eletrônico; Sistemas de Recomendação; Gestão da Experiência do Cliente; Boca a boca Eletrônico; Comércio eletrônico; Administração de empresas; Marketing na internet
• Link de acesso: http://hdl.handle.net/1843/BUBD-AAQH4D

E-commerce allows consumers to find different market offerings with little effort, in such a way that the management of customer experience has become a new competitive frontier. Social influence in this environment has taken an even more important role in the purchasing process, by means of customer reviews, approval ratings and other elements. E-commerce companies make use of Recommender Systems (RSs), which often use behavior patterns of community users to extract useful information and translate them into recommendations. Such recommendations implicitly employ the social influence underlying the process by how the recommendations are produced by calculations, and can go along with arguments which mention the social proof (social-proof argument) or mention an external data source to validate the recommendation (evidence of external validation). In this work, we carried out three experiments in order to understand the effects of using behavioral similarity in recommendations (levels: high and low) and social proof arguments (the conditions: absent and present) (manipulated factors), on the recommendation appraisal, purchase intent and customer experience (dependent variables). The interaction between the manipulating factors was also tested, in order to verify whether social proof moderates the effects of behavioral similarity. The three studies adopted 2 x 2 factorial models, testing the interaction of manipulating factors in samples of 128, 120, and 136 experimental subjects. To calculate the behavioral similarity this work proposed a recommendation system for collaborative filtering. Using MANOVA and ANOVA, also MANCOVA and ANCOVA, the results showed that the behavioral similarity in recommendations provides better evaluation of recommendation, higher purchase intent, and better perception of customer experience when social proof is present. We supported the hypotheses of interaction of the manipulated factors for the three DVs on the two first studies: Study 1: Recommendation Appraisal (F(3,124) = 9.68, p < .05, and ² = 0.53), small effect size (d = .224); Purchase Intention (F(3,124) = 10.93, p < .05, and ² = .48), small effect size (d = .224); Customer Experience: (F(3,124) = 12.99, p < .05, and ² = .102), medium effect size (d = .337); and Study 2: Recommendation Appraisal (F(3,116) = 10.44, p < .05, and ² = .072), medium effect size (d = .278); Purchase Intention (F(3,116) = 8.86, p < .05, and ² = .049), small effect size (d = .226); Customer Experience (F(3,116) = 4.89, p < .05, and ² = 0.46), small effect size (d = .220). No interaction of the manipulated factors was found for the three DVs when social proof was used with external validation on Study 3: Recommendation Appraisal (F(3,131) = 1.61, p > 0.05 e ² = 0.012); Purchase Intention (F(3,131) = 1.61, p > 0,05 e ² = 0.012); Customer Experience (F(3,131) = 0.40, p > 0.05 e ² = 0.003). Thus, this work adds to the understanding of the effects of recommendations on consumer behavior, based on behavioral similarity and use of social-proof arguments.