Recommender systems (RS) play a central role in everyday activities by generating personalized recommendations that support users’ choices. In this regard, evaluating whether these recommendations are fair is crucial. However, this is challenging because fairness criteria vary across stakeholders and existing approaches span many, often incompatible conceptions and operationalization of metrics as well as trade-offs. To address this problem, we propose a stakeholder-centered auditing framework that elicits and formalizes stakeholder perspectives on fairness in RS. We focus on the music domain, where artists and other item providers, listeners, and music streaming services (MSS) have potentially conflicting goals. Furthermore, we follow a three-stage mixed-methods approach: (1) data collection through literature review and semi-structured interviews; (2) framework development through the analysis and synthesis of collected data; and (3) evaluating the framework through scenario-based evaluations with stakeholders. Through this approach, we expect the outcome of a structured auditing approach that improves fairness assessments by redefining what is considered to be appropriate measures to capture fairness and its trade-offs.