Two-Stage Clustering Approaches for Customer Profiling: A Practical Framework

Organizations often have difficulties to extract knowledge from data and selecting appropriate Machine Learning algorithms in order to develop robust User Profiles and segments. Moreover, marketing departments often lack a fundamental understanding on data-driven segmentation methodologies. This paper aims to develop a practical framework of Unsupervised Machine Learning (UML) algorithms for User Profiling with respect to important data properties. We conduct a systematic literature review on the most prominent UML algorithms and their requirements regarding the data properties. The proposed framework provides two-stage clustering approaches for categorical, numerical, and mixed types of data with respect to the data size and data dimensionality. In the first stage, a hierarchical or model-based clustering algorithm is applied to determine the number of clusters. In the second stage, a non-hierarchical algorithm is applied for cluster refinement. The two-stage clustering approach alleviates the drawbacks of solely using a hierarchical or non-hierarchical clustering procedure. The framework can support researchers and practitioners to determine which UML algorithms are appropriate for developing robust User Profiles and data-driven segments. The framework contributes to literature regarding approaches and methodologies for UML and data-driven segmentation in a marketing context. Future research can test the proposed framework on varying types of data, data sizes, and data dimensionality by conducting a case study.