[JDC05] Methods for the analysis of customer satisfaction and loyalty: the experience of ElectrictiÃ© de France

Conférences Internationales sans actes : 3rd world Conf. on Computational Statistics & Data Analysis, Limassol, Chypre, Limassol, Chypre,

Auteurs: E. Jakobowicz , C. Derquenne , V. Casacci

Mots clés: Satisfaction, loyalty, statistical methods

Résumé: Customer satisfaction and retention are key issues for organizations in todayÂ’s competitive market place. Furthermore the French electricity industry is entering a new transition period with the opening of the market. In response, ElectricitÃ© de France has set up a process of surveys to quantify customer satisfaction with an aim of retention and development of consumer loyalty. For five years, EDF is using structural relations models defined by Â“causalÂ” relations between nonobservable variables, known as latent variables, and observable variables, called manifest. The first correspond to the customer satisfaction facets, the latter to satisfaction surveys answers. The study of the links between variables is done using different conceptual marketing models coming from the theory of customer decision-making process. The choice of the model is of first importance because all the remaining studies depend on it. These models are usually estimated using Structural Equation Modelling (LISREL) or Partial Least Squares Path Modelling (PLS). An alternative approach to PLS has been developed within EDF, called Regression on First Principal Components (RFPC), which reduces the bias induced by the inner model. To go still further in research for nonfixed models, we set up Â“free modelsÂ” allowing to build blocks of manifest variables and statistically significant relations without any prior information. In addition, we developed an approach called Partial Maximum Likelihood (PML) which generalizes PLS to different types of variables or variables with different scales. A lot of work is still outstanding, like the analysis of models with longitudinal data, the treatment of filtered questions (generally the complaints in our models) or the possible use of other methods (like Bayesian networks). A presentation of the context will be given together with a short presentation of the different methods. Real-world applications will follow showing EDFs experience in these fields. We will conclude with some future prospects.

Equipe: msdma

BibTeX

@conference {
	JDC05,
	title	=	"{Methods for the analysis of customer satisfaction and loyalty: the experience of ElectrictiÃ© de France}",
	author	=	" E. Jakobowicz and C. Derquenne and V. Casacci ",
	year	=	2005,
}