Abstract : In recent years, the Web has evolved into an exchange platform. Customer Relationship Management (CRM) must follow this evolution and connect CRM tools to social networks in order to place companies in the center of all the exchanges. We propose, in this article, a community detection approach that identi fies clusters of customers of a company using their explicit and implicit behaviour. Our contribution is the definition of a composite pro le that integrates various informations gathered from di erent applications, such as the information system of the company, the existing CRM, or Twitter. We de ne a similarity measure, between a user and a tag, that takes into account the rating and consultation of resources, as well as actions on social networks and user contacts. We validate this approach against a test database and we discuss results and future works.