Goal and objectives of the dissertationThe economic context of the last thirty years in Europe has changed the consumer into a real expert on consumption, so much that now his knowledge of markets, goods and services makes him suspicious of what firms offer. To adapt to this new consumer, firms strive to establish, maintain and enhance customer relationships. From this relationship context a word emerges: 'complicity' (meaning, 'close relationships'). This 'complicity' is rooted in commercial narratives used in customer destined contacts with the aim of making the latter feel an 'accomplice'.The purpose of this dissertation is to introduce the concept of 'complicity' relationships between customers, as a means of compensating for service failures.The aim is to answer the question: to what extent can relationships of complicity take the form of coalition between customers, and have a positive impact on consumers and companies? The objective is to model complicity in terms of antecedents, dimensions and consequences.MethodologyThis dissertation mobilizes both qualitative and quantitative methods. Though there is a vast theoretical background to the field of complicity relationships, to date there is a lack of conceptual clarity and definition. Therefore, two exploratory studies are carried out. The first one consists of twenty semi-structured personal interviews with 'experts'. The second one is an ethnographical study, based on participant and non-participant observations. Then, to validate the hypotheses and model, quantitative methods are utilized. A theoretical model of complicity is created, in terms of antecedents, complicity and consequences (stopgaps). A service failure or insufficiency triggers a need for social links. Combined with the cognitive proximity of consumers they bring about the relationship of complicity. The latter entails palliative actions both practical and emotional. These service recovery actions have a positive impact on the overall perceived satisfaction. The following step consists of inferring eighteen hypotheses. Eighteen hypotheses are inferred. Three existing scales of measure are utilized: service failures, emotions, satisfaction. Four scales of measure are created: the need for social links, cognitive proximity and functional palliative complicity, and complicity itself. Then a research protocol is elaborated and validated. To collect data, a questionnaire is created, consisting of fortyeight questions classified into four sections. Churchill's Paradigm, and Gerbin and Anderson's updated paradigm are followed. The pre-test aims at testing the psychometric properties of measurement tools. An exploratory factor analysis is conducted by means of principal component analysis. Validity is tested with SPSS through usual indices: Communalities, KMO test, the percentage of variance, Cronbach's Alpha. The methodology of the main research hinges upon the validation of the external model of measure. Dimentionality is tested, through three sets of indices: absolute indices (RMSEA, GFI, AGFI), incremental indices (NFI, CFI) and parsimony indices (X^sup 2^/df). Then confirmatory factor analysis is conducted; reliability and validity are tested with Cronbach's Alpha and Dillon-Goldstein Rho, less sensitive to the number of items. Convergent validity is tested by means of the T test associated with each factor contribution, the factor weight and the convergence Rho. Discriminant validity is tested through the method of split variance.Subsequently, the structural model is tested through the three sets of indices above mentioned, with LISREL. Three steps were followed: specification, identification and model estimation. Finally we confirmed ten hypotheses, by means of the Student's T and standardized coefficients. This methodology is applied in several grounds of research, encompassing a large number of service companies mainly experiential, tourist and international ones: a ski resort, hotels, restaurants, a sporting centre, a national railway company. …