Modeling product co-consideration relations: A comparative study of two network models

Customers often compare and evaluate alternative products before making purchase decisions. Understanding customer preference is an important step for choice modeling in engineering design. This study presents a network approach to model co-consideration relations between products in supporting engineering design decisions. The network approach of co-consideration represents each product as a node, and a link between two nodes implies the two products are co-considered by customers. We compare two network-based modeling techniques - the multiple regression quadratic assignment procedure (MRQAP) and the exponential random graph model (ERGM). Using vehicle purchase data in the 2013 China market, we evaluate the goodness-of-fit of the two techniques at both network level and link level. The analysis indicates that the ERGM outperforms the MRQAP model. Specifically, the ERGM is able to characterize the interdependence of product co-considerations through various network configurations and therefore has a better fit of the data. The insights of co-consideration models help to understand market segmentation and product competitions as well as other types of product associations.