Unfolding large-scale marketing data

Marketers use multidimensional unfolding to understand the relationship between customer preferences and product positioning through a joint display of customers and brands on a map. In today's information age, unfolding marketing data is challenging, as marketing data can be large in size and high-dimensional in nature. Moreover, the unfolding model is always subject to the curse of the degeneracy problem. We propose a new approach to unfold customer-by-brand transaction data and customer-by-customer network data in a reduced space. The proposed approach can recover the true configuration with reasonable accuracy, is scalable in terms of the number of estimated parameters, and can produce non-degenerate solution. We compare its performance with existing approaches by simulation experiments and real data analyses with interesting results.