A key challenge in systematically collecting data on intermediary agri-food value chain actors is that value chains take the form of a network, with actors linked by a series of transactions. Moreover, we have limited ex ante knowledge about the structure or scale of these networks, which complicates the construction of valid sampling frames and limits traditional random sampling approaches to collect data. To address these challenges, we adapt the respondent-driven sampling approach to collect data on intermediary agri-food value chain actors within their transaction-linked network and implement this approach in the arabica coffee and soybean value chains in Uganda and the rice and potato value chains in Bangladesh. We observe meaningful heterogeneity in the structure and scale of agri-food value chains across commodities and countries. Focusing on traders, we show that the respondent-driven sampling approach generates a larger sample of traders who differ in observable characteristics (i.e., value added, enterprise scale, and financial access) compared to a sub-sample of traders generated in a way that mimics traditional random sampling approaches used to study traders. We conclude by discussing how this respondent-driven sampling approach, applied within transaction-linked networks, can provide a useful data collection method for studying intermediary agri-food value chain actors.