Segmentação de clientes com base em dados contextuais e atitudinais
Resumo
Resumo: A segmentação de clientes é um conceito que busca identificar grupos homogêneos. Originalmente construído a partir de dados demográficos e geográficos, foi ampliado com o uso de dados comportamentais e com o surgimento do e-commerce, possibilitando a análise de segmentos com maior volume de dados e atributos, desde que válidos, úteis e interpretáveis. Assim, o objetivo desta tese é compreender os critérios e atributos de segmentação de clientes que representem aspectos contextuais e atitudinais, com uma abordagem post-hoc. Para isso, foi realizado um levantamento de artigos publicados, o qual identificou duas abordagens (a priori e post-hoc) e 27 métodos de segmentação mais pesquisados. Em seguida, realizou-se uma pesquisa bibliográfica, que resultou na listagem de 229 atributos, e, uma análise textual, baseada na metodologia de Flick (2009), que resultou na descrição de 27 atributos, sendo 19 contextuais e 8 atitudinais. A descrição desses atributos foi feita com base no significado de cada um, visando identificar fatores contextuais e atitudinais. Para validar o uso desses atributos, aplicou-se a abordagem post-hoc, com análise de cluster, usando o algoritmo K-means, em um conjunto de dados artificial de 50 mil instâncias. Com isso, foram formados de três clusters homogêneos, representando 53% (cluster 0), 32% (cluster 1) e 15% (cluster 2), onde foi possível observar sete atributos diferenciadores entres os clusters, possibilitando um relacionamento mais preciso com os clientes, com base nos dados contextuais e atitudinais descritos. Conclui-se que, no contexto e no uso da segmentação, a abordagem post-hoc é particularmente indicada para a análise de grandes conjuntos de dados e identificação de grupos com padrões não óbvios. Além disso, entre os métodos de segmentação mais frequentes, o algoritmo K-means prevalece (28%). Também os atributos descritos e analisados pelo experimento revelam padrões relacionados aos aspectos contextuais e atitudinais dos clientes, diferenciando-se das abordagens convencionais baseadas em atributos demográficos e geográficos prédeterminados. Essa diferenciação destaca a utilização de atributos contextuais e atitudinais para a segmentação, com foco na compreensão do comportamento e necessidade dos clientes. Abstract: Customer segmentation is a concept that seeks to identify homogeneous groups. Originally built on demographic and geographic data, it has been expanded with the use of behavioral data and the emergence of e-commerce, making it possible to analyze segments with a greater volume of data and attributes, if they are valid, useful and interpretable. Therefore, the aim of this thesis is to understand the criteria and attributes for customer segmentation that represent contextual and attitudinal aspects, using a post-hoc approach. To this end, a survey of published articles was conducted, which identified two approaches (a priori and post-hoc) and 27 of the most researched segmentation methods. This was followed by a literature search, which resulted in a list of 229 attributes, and a textual analysis, based on Flick's methodology (2009), which resulted in the description of 27 attributes, 19 of which were contextual and 8 attitudinal. These attributes were described based on the meaning of each one, with the aim of identifying contextual and attitudinal factors. To validate the use of these attributes, a post-hoc approach was applied with cluster analysis using the K-means algorithm on an artificial data set of 50,000 instances. As a result, three homogeneous clusters were formed, representing 53% (cluster 0), 32% (cluster 1) and 15% (cluster 2), where it was possible to observe seven differentiating attributes between the clusters, enabling a more precise relationship with customers, based on the contextual and attitudinal data described. The conclusion is that, in the context and use of segmentation, the post-hoc approach is particularly suitable for analyzing large data sets and identifying groups with non-obvious patterns. Furthermore, among the most frequent segmentation methods, the K-means algorithm prevails (28%).The attributes described and analyzed by the experiment also reveal patterns related to the contextual and attitudinal aspects of customers, differing from conventional approaches based on predetermined demographic and geographical attributes. This differentiation highlights the use of contextual and attitudinal attributes for segmentation, with a focus on understanding customer behavior and needs.
Collections
- Teses [29]