The reliability of several lenght structure-based population parameters depends on the quality of lenght frequency-distributions obtained during fishery sampling. Sampling commercial catches are relatively precise and cheap, however it may produce sistematic errors. Sampling strategy, in this case, is essential for the success of the study. Under this perspective, lenght sampling of the skipjack tuna, landed in Itajaí, Santa Catarina (southern Brazil), by the commercial pole-and-line fishery, was evaluated for robustness, precision and to detect possible bias. The estimation of standard errors and size class range of the samples revealed that a minimum of 100 fish are required to be measured of each series. Desagreements in lenght frequency distributions were observed within the same fish hold and between fish holds. Therefore more than one 100 fish series must be measured in each landing and preferentially from all filled fish holds. In addition temporal and spatial variations of the population structure, indicate that sampling effort each month must cover catches originated from the large number of fishing areas possible.