Existing sampling methods in language typology strive to control for areal biases in typological datasets as a means to avoid contact effects in the distribution of linguistic structure. However, none of these methods provide ways to directly compare contact scenarios from a typological perspective. This paper addresses this gap by introducing a sampling procedure for worldwide comparisons of language contact scenarios. The sampling unit consists of sets of three languages. The Focus Language is the language whose structures we examine in search for contact effects; the Neighbor Language is genealogically unrelated to the Focus Language, and counts as the potential source of contact influence on the Focus Language; the Benchmark Language is a relative of the Focus Language neither in contact with the Focus nor with the Neighbor language, and is used for disentangling contact effects from genealogical inheritance in the Focus Language. Through this design, we compiled a sample of 49 three-language sets (147 languages in total), which we present here. By switching the focus of typological sampling from individual languages to contact relations between languages, our method has the potential of uncovering patterns in the diffusion of language structures, and how they vary and change.