An integral part of global supply chains is the selection by international buyers of trading partners in developing countries. However, our understanding of how buyers find a suitable long term supplier is limited. I use unique buyer-seller customs data to directly observe experimentation activity in a large market - the "fast fashion" industry in Bangladesh. I study how buyers of ready-made garments conduct trials of suppliers at the order-product level before settling into sustained sourcing relationships. To illustrate this process, I use a model of idiosyncratic search costs where the buyer’s costs of testing a manufacturer are determined by the heterogeneity of potential suppliers. The model shows that (1) higher supplier heterogeneity is associated with lower experimentation, (2) as heterogeneity increases, search activity falls more markedly for larger buyers than for their smaller counterparts, and (3) while buyer-seller matches are positively assortative, more heterogeneous settings see all buyers -and more markedly, large buyers- willing to accept relationships with (weakly) worse suppliers. These implications are strongly supported by the data, and hold in terms of within-buyer, cross-market differences in experimentation behavior. Finally I show that these information frictions, rooted in supplier heterogeneity, matter for the distribution of rents in these relationships: price-cost margins for suppliers are positively related to the degree of heterogeneity in the environment.