Automotive Customer Satisfaction Data Analysis Using Logistic Regression

It is standard practice in the automotive industry to use the Customer Satisfaction (CS) metric, defined as the percentage of “high satisfaction” ratings, i.e. the percentage of customers who rate a vehicle feature either 9 or 10 on a 10 point scale. Based on the observation that this is equivalent to a transformation from discrete to binary, this paper introduces logistic regression as a natural choice for statistical analysis of CS data. The methodology proposed in this paper uses penalised maximum likelihood for model fitting and the Akaike Information Criterion (AIC) for model selection. AIC is also used for optimal selection of the shrinkage parameter. The paper also shows how this methodology can be used to identify factors associated with low customer satisfaction.