Abstract:

 

Computer-aided design of new products is made possible by the confluence of a number of factors in advanced analytics and computational speed. Recent interest in AI and its sub-branch, Machine Learning, has drawn attention to this area. Although the interest in this area and the goals have remained relatively constant over several decades, we are much closer to achieving these goals today than ever before. In a recent technical report, we discussed how Linear Programming can now be used in combination with Graph Theory and TURF to create optimal product bundles from unimaginably large numbers of possibilities. Another example of interest, in this report, is unfolding which is a computationally intensive analytic technique based on a great idea proposed by Clyde Coombs in the 1950s. Unfolding of liking or purchase interest involves the determination of ideal and product coordinates in a low-dimensional drivers of liking space. This method is a key to understanding what drives consumer liking and hardly anyone would disagree with the process model that it specifies or the objective it attempts to accomplish.

 

After about a half-century of frustrating failure to achieve nondegenerate solutions to the unfolding problem (degenerate solutions are ones with little or no interpretability) the issue was finally resolved. One of the successful methods for unfolding, developed in 2001, is Landscape Segmentation Analysis® (LSA) which is based on a probabilistic similarity model involving latent product and individual ideal point coordinates. Other methods have also been proposed since then. Once an LSA map is generated, it can be used as a computer-aided design tool to predict the success of new products based on available descriptive sensory data, or other descriptive data, without the expense of conducting a consumer test. Since the method also provides individual ideal point locations, LSA facilitates segment identification and can be synergistically linked to newer methods of analysis, such as Machine Learning, to characterize segments. This facility for synergistic analytics makes LSA an attractive candidate to link to newer and older methods such as Conjoint Analysis, MaxDiff, and Decision Trees or Random Forests. In this technical report, we will explore the capability of using LSA as a computeraided design tool and show how it can be used to predict the performance of new prototypes or marketing concepts and to associate them with emerging segments.

​

​

​