Desveaud, Kathleen (2021) Proposition of a comprehensive review methodology to build integrative and representative meta-models in Marketing research‎: An application to Customer Brand Loyalty antecedents. École Doctorale Sciences de Gestion TSM (Toulouse).

Text Restricted Registered users only Download (13MB)

Abstract

Le résumé en français n'a pas été communiqué par l'auteur.

In the past few years, we saw a dramatic increase in marketing publications, with the development of various theoretical frameworks and concepts to explain a whole bunch of phenomena. A perfect example is the long-standing topic of brand loyalty for which we can find a plurality of underlying explanations and theories. Facing such diversity and richness of findings, it is beneficial to aggregate them into a coherent framework to provide a common and intermediary knowledge base, between past and future research. Literature reviews are helpful in doing so, but this thesis reveals that despite a variety of useful methodological approaches applied in marketing, we actually lack a protocol specifically designed to study the extensiveness of findings on a given topic in a consistent and representative manner. A great number of excellent literature reviews are available to enhance our knowledge on a given topic, but to date, none has embraced the goal of aggregating the whole body of knowledge through the proposition of a representative meta-model of past literature. What we referred as meta-models in this thesis are integrative insights representing the key concepts and their associated relationships in a structural model format. Given the specificities of marketing research, they are very efficient in reporting on existing knowledge for many marketing phenomena like customer brand loyalty. We propose in this thesis to help improve marketing research practices by fostering the development of such meta-models. To this end, we produced four research papers, providing key methodological contributions. In the first paper we introduce an innovative method borrowed from the Information Systems discipline; the Grounded Theory Literature Review method. While Grounded Theory has been widely used in marketing research, the Grounded Theory Literature Review method, on the other hand, is a brand-new review method for marketing and of great value to identify, categorize and select the key concepts previously studied to explain a given phenomenon, in a rigorous and systematic fashion. We then test and illustrate this approach by applying it to brand loyalty antecedents in a second paper. In a third paper we introduce a complementary new method, this time borrowed from Quantitative Psychology; the OneStage MASEM. While the first method results in the development of a meta-model with its key concepts and categories, this second method results in the quantification of the meta-model. In a fourth paper, we then apply this quantitative method on brand loyalty to test its limits and refine our protocol. Consequently, we propose as a contribution a two-step methodological protocol, composed of two innovative methods to be applied successively, for any fragmented and rich topic. On top of this methodological contribution, the application of this protocol to brand loyalty also provided new theoretical and managerial contributions. We proposed an integrative meta-model of loyalty’s key antecedents gathered into four key categories namely (1) brand associations with the concepts of brand identification, self-congruence, and perceived CSR, (2) brand offer with the concepts of service quality and perceived value, (3) brand experience with the concepts of satisfaction and delight and finally (4) brand relationship with the concepts of trust and attachment. We show that loyalty is formed in a stepwise process in which brand associations and offer components, constitute fundamental variables that arrive upstream of the process with the higher total effects, while experience and relationship components, have a more direct effect on loyalty and arrive downstream of the process. A more surprising result is that contextual variables (e.g., product type, country, sample characteristic) have very little effect on the model as a whole, attesting for the almost absolute nature of the proposed meta-model.