Analysis of customer satisfaction survey data

Author(s):  
P. Rotella ◽  
S. Chulani
Author(s):  
Lin He ◽  
Christopher Hoyle ◽  
Wei Chen

Choice modeling is critical for assessing customer preferences as a function of product design attributes and customer profile information. Previous works have focused upon the use of survey data in which respondents are presented with a set of simulated product options from which they make a choice. However, such data does not represent real purchase behavior and these surveys require significant time and additional cost to administer. For these reasons, an approach to estimate a choice model using widely available customer satisfaction survey data for actual purchases is developed. Through a close examination of customer satisfaction survey data, several key characteristics are identified, including the lack of defined choice sets and missing choice attributes, the use of subjective measures such as ratings by customers to describe product attributes, multiple collinearity among many of the product attributes, and potentially insufficient attribute variation in the product designs evaluated by the respondents in the survey. A mixed logit based choice modeling procedure is developed in this paper to incorporate the use of both survey ratings as subjective measures and engineering attributes as quantitative measures in the model utility function. In order to accurately reflect choice behavior in actual market conditions, heterogeneity in customer preference is explicitly considered in the demand model. A case study using the Vehicle Quality Survey data acquired from J.D. Power and Associates demonstrates many of the key features of the proposed approach. The estimation results show the mixed logit model to be successful in modeling customer choices at the individual level, demonstrating the potential of being integrated with engineering models for engineering design.


Sign in / Sign up

Export Citation Format

Share Document