An Integrated Latent Variable Choice Modeling Approach for Enhancing Product Demand Modeling

Abstract

In today's highly competitive economy it is increasingly important to consider customer desires in engineering design on a systems level, that is, there is a need for integrating business decision-making and engineering decision-making. Building upon the earlier work on using the discrete choice analysis approach to demand modeling, in this work, the discrete choice analysis method is enhanced by introducing latent variables to include the customer's attitude and perception in a demand model. The latent variable approach better captures psychological factors that affect the purchase behavior of customers and facilitates the understanding of the relationship between customers' desires and product features. The approach is expected to enhance the predictive accuracy of demand models and help verify a designer's intent by assessing the contributions of various product attributes to the customer's perceived product performance. In this work, the existing binary latent variable discrete choice model is extended for multinomial choice and the mathematical formulation is developed to integrate a latent variable model and a multinomial logit choice model. The demand modeling of passenger vehicles with emphasis on studying the impact of engine design attributes is used to demonstrate the potential of the proposed approaches. It is important to keep in mind that the example's emphasis is on demonstrating the approaches rather than the results per se.

Navigating redesign and market desirability implications when considering increased product variety

“…Arguably, this challenge has been best met by research in DCA. Wassenaar first integrated DCA into decision-based design (Wassenaar and Chen 2003;Wassenaar et al 2004), andMichalek, Feinberg, andPaplambros (2005) and Michalek, Ceryan, and Paplambros (2006) used DCA to design both a single product and a product line. Nested logit models were used by Kumar, Chen, and Simpson (2009) to represent heterogeneity in the development of product families, and the advantages and challenges of using a Hierarchical Bayes (HB) multinomial logit (MNL) model (Train 2003) were explored by Sullivan, Ferguson, and Donndelinger (2011).…”

Creating targeted initial populations for genetic product searches in heterogeneous markets

“…These works saw the first application of the logit model (Wassenaar and Chen 2003;Wassenaar et al 2004), experimental methods for profiling the market and mapping to the technical space (Hoyle et al 2008;), exploration of model assumptions and their implication on results (Shiau et al 2007;Donndelinger, Robinson, and Wissmann 2008), integration with existing design-decision tools (Michalek, Feinberg, and (Williams, Kannan, and Azarm 2011). However, these works focused on the design of a single product.…”

Enhanced Targeted Initial Populations for Multiobjective Product Line Optimization

“…Choice-based conjoint studies and discrete choice analysis [8][9][10] ushered the next evolution with the added level of realism associated with selecting from a set of alternatives. These works saw the first application of the logit model 11,12 , experimental methods for profiling the market and mapping to the technical space 13,14 , exploration of model assumptions and their implication on results 15,16 , integration with existing design-decision tools 17,18 , exploration of market heterogeneity 19 , application toward mass customization 20 , and the effect of retail channels 21 . However, these works focused on single product design.…”