Two plant fossil-bearing beds from the middle Barremian of Belgium were analysed to ascertain how experimental designs affect conclusions regarding palaeodiversity at a local scale. We analysed eight lateral samples per bed taken regularly every 3 m using an exhaustive sub-sampling method. The Clench equation was used to evaluate the completeness of the taxonomic inventory of the samples and the sampling effort needed to obtain a reliable representation of diversity. The number of replicates needed to obtain the same representation of diversity from different nearby lateral samples of the same bed ranged from 5 to 19. Richness (S), Evenness (J) and the number of equiprobable taxa (2H’) greatly varied between samples from the same bed, even over short distances. Only one of the studied samples was representative of the taxonomic inventory of its bed. Our study shows that 1) the selection bias of the sampling area is reduced by increasing the number of lateral samples taken in a bed, enabling more reliable conclusions about local-scale diversity; 2) intense sub-sampling methods are needed to account for statistically independent observations of detailed lateral variation; and 3) sampling methods in palaeodiversity analyses must look for a similar degree of representativeness in samples rather than a homogeneous sample size. Using a sampling effort analysis provides evidence for the completeness of the data set, adjusting the amount of work required. Implementing the Clench equation in palaeodiversity analyses improves the performance of data acquisition in palaeoecological studies and provides a quality test of the data sets derived from them.