ENBIS-12 in Ljubljana

9 – 13 September 2012 Abstract submission: 15 January – 10 May 2012

Ask the Right Questions, or Apply Involved Statistics? Thoughts on the Analysis of Customer Satisfaction Data

11 September 2012, 11:50 – 12:10


Submitted by
Thomas Hochkirchen
Dr. Thomas Hochkirchen (Ford Motor Company), Dr. Martin Blankenagel (divis GmbH)
For years, there has been much talk about customer satisfaction – why companies should care, how we should measure and how we should analyse our customers’ satisfaction levels. Statisticians have published many papers on appropriate modelling methods to identify the “real” drivers of satisfaction.

Our discussion’s focus is on the experience we made in the automotive industry, which has become incredibly competitive and thus spends much effort on understanding the voice of our customers. However, automobiles do not only carry passengers but as well lots of emotions, making separation of emotional and technical aspects of a product a challenging task. Statistically, this becomes visible in a high degree of multicollinearity between potential input variables, which makes “naïve regression modelling” a dangerous game to play.

While we’ll provide a choice of tools suggested to overcome the technical difficulty of highly correlated inputs, we do as well want to trigger a discussion if we typically start from the right starting point: If we ask the right questions (in customer, not in engineering speak!), the multicollinearity issue will disappear – and with it, the need for unnecessarily complicated analysis methods. This in turn enables easier communication of findings and therefore increased acceptance of statistics as a tool with real business benefit.

Return to programme