The sampling of commercial marine fisheries for management purposes often displays a key weakness in the form of poor documentation of the scientific basis of sampling and estimation, the assumptions made, and the practical constraints. This paper reviews systematically the theoretical and practical options that can remedy this situation and recommends that decisions be archived in regularly updated ‘Sampling Approach and Modifications’ (SAM) documents. Defining the target population, the observable population (usually a subset of the target), and the assumed links between them is important, along with the distinction between design‐ and model‐based sampling approaches. Fleet‐targeted and stock‐targeted sampling strategies are contrasted, the latter being much harder to implement. Sampling protocols aimed at estimating quantities of fish landed and discarded, length–frequency distributions, length‐related variables such as age, weight and maturity, and ratio variables such as catch per unit of effort and the proportions of discards are discussed, together with the raising of estimates to fleet and/or stock levels. The ideas are summarized in the specific contexts of landings sampling, logbook schemes and sea‐going observer surveys. SAMs are commended for enhancing the scientific value of fishery sampling, and for encouraging methodological discussions among users and producers of the data.