The angiosperm Apiaceae tribe Scandiceae includes four major clades—subtribes Daucinae, Ferulinae, Torilidinae, and Scandicinae—that originated ca. 20 Mya. Although all four subtribes are highly supported in molecular analyses, and morphological data indicate a sister relationship between Daucinae and Torilidinae, their branching order has not been resolved using standard Sanger multilocus data. Therefore, in this study, we test the utility of genomic RAD seq data in resolving deep phylogenetic relationships (up to 20 Mya) in Apiaceae subfamily Apioideae, with special emphasis on tribe Scandiceae using 12 representative species. We used two bioinformatic pipelines, pyRAD and RADIS (based on STACKS), to assemble RAD seq data and we tested the influence of various combinations of parameters on the robustness of the inferred tree topologies. Although different data processing approaches produced alignments with various amounts of missing data, they converged to two well‐supported topologies, irrespective of the phylogenetic method applied. Highly supported trees showed Scandicinae as sister to all other clades and indicated that Daucinae and Torilidinae are sister groups, thus confirming the relationship inferred from morphology. We conclude that the RAD seq method can be successfully used to resolve deep relationships formed 20 Mya within Apiaceae. We provide recommendations for parameter settings in RADIS and pyRAD for the analysis of taxa that have accumulated considerable genomic divergence.