Fostering plant biodiversity using companion plants (CP) is a promising alternative for the sustainable management of crop health in horticultural systems. To evaluate the contribution to pest and disease control, CPs are tested in field experiments carried in conditions close to those used on commercial farms. Such experiments provide scientific knowledge on the achievement of the targeted function, but also raise many operational issues. We therefore consider that experimenters who design and test systems with CPs support the production of experience-based knowledge. Our objective is to highlight and characterize this knowledge. We interviewed experiment leaders, covering 21 locations in metropolitan France and the West Indies. We defined a "modality of use of companion plants" (MU) as the combination of the CP, its technical management, including spatial and temporal arrangement in the field, and the main expected functional processes assigned to CPs. 46 MUs were investigated. MU scale was relevant for experimenters to report on interactions between practices and CPs, and to properly assess CP success or failure. We showed that experimenters reflect on both anticipated barriers and on barriers they actually faced when implementing CPs in the field. The diverse obstacles encountered were mainly related to CP growth, labour, crop management, or sanitary control difficulties. Experimenters also identified improvements. This knowledge from practical experiments completes scientific knowledge on ecological processes and should be useful to stakeholders intending to adopt CPs. Building on such actionable knowledge would be valuable for the sustainable management of crop health in horticulture.