Sampling designs for collection of shark size composition data are needed to fill one of the primary knowledge gaps hampering management of shark fisheries in Central America. These designs need to be tailored towards multi-species fisheries with highly structured landings, which are driven by buyer demand in port. In addition, due to the difficulty of maintaining catch quality outdoors in a tropical environment, the designs ideally would have minimal impact on the often-rapid catch landing process. In this study, simulations based on data from exhaustively sampled landings of Costa Rican mediana escala (78 landings) and avanzada (21 landings) commercial longline vessels were used to test single-cluster systematic sampling designs for estimating size (commercial weight category) composition of retained catches of shark species, as well as of other taxa present in landings with sharks. The wide range in numbers of sharks per landing led to sampling designs for which the frequency of within-landing sampling depended on the number of sharks to be landed. Designs for shark-only sampling and sampling of all taxa were tested. For the mediana escala fleet component, these designs performed well in terms of absolute relative error (ARE) for common species in the landings, including the silky shark, the dominant shark species, and swordfish, striped marlin and yellowfin tuna. There was less than a 5% chance that the ARE would exceed 0.2 for most weight categories of these species, even at the lowest sampling frequency. For all but the large weight category of striped marlin, there was only a 2% chance (or less) that the ARE would exceed 0.3. These results suggest that a single sampling design can be implemented for this fleet component to obtain estimates of the size composition of common shark and non-shark species, alike, in landings with sharks. For relatively uncommon shark species in the landings, such as hammerhead and the pelagic thresher sharks, the same sampling designs generally resulted in higher ARE, except for the dominant weight categories (small hammerhead sharks and large pelagic thresher sharks). However, the level of error itself associated with these designs may still be acceptable, depending on management needs; ARE can provide an overly pessimistic evaluation of performance for uncommon species. For landings of the avanzada fleet component, the performance of the shark-only sampling scenarios for the silky shark was generally similar to the results obtained for the mediana escala fleet component, except for the large weight category, although performance for this weight category, as measured by error, might still be acceptable. However, differences in performance between the shark-only and the all-fish scenarios for the silky shark raise the concern that the number of landings of avanzada vessels available for this analysis may be too small to make a definitive statement about sampling design performance for this fleet component. Development of sampling designs for the avanzada fleet component would benefit from an increase in sample size for analysis and a more detailed study of landing characteristics.