How to assess the quality of evidence produced by the evaluation of advisory services dedicated to occupational safety and health (OSH)? Besides the “level of evidence” (i.e. statistical robustness), another very important dimension of the quality of the evidence is its “relevance” (i.e. its ability to deal with the issue that is considered a priority, to take-into-account the context of the intervention and the structural sources of inequalities, to have access to reliable data to design and implement the evaluation). The goal of the research was to characterize what different types of stakeholders (farmers, farm workers, advisors, managers...) consider to be a relevant evaluation of OSH advice in agriculture. The methodology combined three sources of data in France: (i) semi-structured interviews (n=33) with different types of actors, (ii) administrative and statistical data, (iii) scientific literature, regulations, official reports. The results show five contrasted perceptions of what would be a relevant evaluation system for farm OSH advisory services. They also reveal antagonistic positions about the relative importance of some topics (e.g. pesticides exposures) and the importance to be given to different types of beneficiaries (e.g. non-permanent workers). They confirm that the trade-offs between level of evidence and relevance are a sensitive issue for the design and the implementation of the evaluation of OSH advisory services.