Many marketing subjects call for treating different "level" variables within a single model (and therefore to resort to, as we shall demonstrate, "multilevel" models) 1 such as, for example:• international marketing problems involving consumer segmentation (Ter Hofstede, Wedel and Steenkamp, 2002;N'Gobo and Guinguant, 2006) or, for example, demonstrating the existence of moderating effects within the relationship between the perceived quality of websites and its main drivers on the individual level (Steenkamp and Geyskens, 2006) requires the integration of national characteristics (country level) in addition to individual characteristics (consumer level); • measurement of the ROI from relationship marketing programs requires assigning variables for each of the levels involved: customer, salespersonThe authors would like to thank the anonymous reviewers of this article as well as the editor of Recherche et Applications en Marketing for their comments and suggestions. The authors are, however, solely responsible for any remaining errors. Jean-Claude Ray would also like to thank Section 37 of the CNRS for allowing him to benefit from a delegation for a period of two years to conduct a research project involving the implementation of multilevel models. The authors can be reached at the following e-mail address: daniel.ray@grenoble-em.com 1. While the term multilevel seems to be the norm, these models are also referred to as mixed linear models. These are often hierarchical linear models (HLM).
ABSTRACTRecent marketing literature relies more and more on multilevel models. Yet, this method, which applies to data structured in levels (e.g., consumers, nested within stores, themselves nested within retailers), is not straightforward to understand and to implement. Through a marketing example, the present text introduces the reader to the objectives, interest and limits of multilevel modeling.