The increasing complexity and volume of organizational data have led to the emergence of the Data Mesh paradigm, a data architecture with a federated governance aimed at addressing the limitations of traditional monolithic data systems that has overlapping principles with the microservices architectural style. Although related work exists, the majority of architectural approaches regarding Data Mesh are conceptual, technology-centric or vendor specific. This paper introduces a Data Mesh Reference Architecture (RA) using the ArchiMate enterprise architecture modeling language, designed to assist organizations in implementing (or migrating towards) data mesh solutions. The RA comprises three main components: domain architecture, self-serve data platform architecture, and federated governance, which reflect the main Data Mesh principles. Through a systematic literature review, four data mesh archetypes (Pure, Semi-Pure, Hybrid, and Distributed) were identified, along with challenges, limitations, and motivational factors for adoption. A questionnaire-based validation among experts confirmed the RA’s utility, quality, and variability. However, practical validation was not conducted within this study. The study contributes to both literature and practice by offering a structured approach and a set of reference models for designing data mesh architectures. Future research can contribute to practical validation, assessment of RA-driven design efficiency, and extending the RA with domain-driven solution architectures.