‘We Cannot Diagnose the Patient's Illness…But Experience Tells us What Treatment Works’

Abstract

Having spent a lot of money collecting data to better understand the satisfaction of their customers, many clients want to know, with some certainty, what will happen if...? The traditional statistical techniques used to answer this question frequently struggle to cope with the complexity of real survey data, and in particular the interrelationships that exist between the various measures which make up 'satisfaction'. In providing a solution, some analysts venture where others fear to tread, and many clients are obliged to follow whether they know the risks or not. The authors have developed a practical and intuitive solution to building 'what if' scenarios using an empirical approach, which overcomes many of the technical problems associated with analysing complex customer satisfaction data. Results are transparent to the client and can be explained without compromising the truth. As a consequence, the authors feel that clients will make better investment decisions. This paper describes the analytical approach to creating 'what if' scenarios and provides case studies, using real datasets, which had previously proved trouble some. Implications for the ubiquitous key driver analysis are also discussed.

An expert system for perfume selection using artificial neural network

“…The ANNs approach has been applied more recently to consumer satisfaction and loyalty analyses (e.g. Audrain, 2002;Hackl & Westlund, 2000;Willson & Wragg, 2001). Gronholdt and Martensen (2005) applied ANNs in customer satisfaction analysis to identify existing patterns in the data, and synergies between the drivers of satisfaction.…”

Analysis of a customer satisfaction survey using Rough Sets theory

“…In general, we are unable to derive such conclusions from the collected survey data because the scale of questions in questionnaires is usually measured on an ordinal or nominal scale. Several statistical and intelligence methods have been applied to customer satisfaction analysis like conjoint analysis (Danaher, 1997), partial least squares (Ryan et al , 1999), targeted bootstrapping (Willson and Wragg, 2001) and artificial neural network approach (Hackl and Weslund, 2000). In recent years, we have seen increased attention being given to satisfaction analysis in the intelligence methods literature, where a simple description on how to model the customer retention problem using the Rough Sets (RS) model was given (Kowalczyk, 1997).…”