Reliable and comparable estimates of biodiversity are the foundation for understanding ecological systems and informing policy and decision-making, especially in an era of massive anthropogenic impacts on biodiversity. Environmental DNA (eDNA) metabarcoding is at the forefront of technological advances in biodiversity monitoring, and the last few years have seen major progress and solutions to technical challenges from the laboratory to bioinformatics. Water eDNA has been shown to allow the fast and efficient recovery of biodiversity signals, but the rapid pace of technological development has meant that some important principles regarding sampling design, which are well established in traditional biodiversity inventories, have been neglected. Using a spatially explicit river flow model, we illustrate how sampling must be adjusted to the size of the watercourse to increase the quality of the biodiversity signal recovered. We additionally investigate the effect of sampling parameters (volume, number of sites, sequencing depth) on detection probability in an empirical data set. Based on traditional sampling principles, we propose that aquatic eDNA sampling replication and volume must be scaled to match the organisms' and ecosystems' properties to provide reliable biodiversity estimates. We present a generalizable conceptual equation describing sampling features as a function of the size of the ecosystem monitored, the abundance of target organisms, and the properties of the sequencing procedure. The aim of this formalization is to enhance the standardization of critical steps in the design of biodiversity inventory studies using eDNA. More robust sampling standards will generate more comparable biodiversity data from eDNA, which is necessary for the method's long-term plausibility and comparability.