Verbal interaction has been the subject of a growing interest among language professionals in Europe since the CEFR was published in 2001; in linguistics, verbal interaction has long been studied. In the Bakhtinian approach, it is even considered "the fundamental reality of language". All types of interaction share the fact that they are dynamically co-constructed by participants. How then can we assess or certify interactional competence on an individual basis when dynamic instability prevails? What criteria can be defined in order to deconstruct interactional competence into specific operational criteria, if interaction is intrinsically multidimensional? These are the questions that we address in this paper. To do so, this paper presents the insights gained as a result of the co-operation between two certification systems: CertAcles (Spain) and CLES (France), both belonging to NULTE ("Network of University Language Testers in Europe"). These certification systems have agreed to collaborate extensively, sharing their constructs and assessment routines. As a result, CertAcles is shifting towards more contextualized tasks and CLES is considering adopting descriptive assessment scales for interaction (C1 level). We hope to demonstrate that the materialization of scientific collaboration of this kind can help improve individual systems.