While concepts of connectivity are increasingly used in determining locations for marine protected areas, they are much less applied in the management of fish stocks, which are assumed to be well-mixed populations. However, due to seascape structure and often asymmetrical dispersal, the stocks of many species are unlikely to be well mixed and there is potential to enhance management by utilising emerging ecological modelling approaches that incorporate functional connectivity. Here, we apply a new model, MerMADE, that couples biophysical modelling of dispersal with spatial population demography, to predict within-stock patterns of connectivity of sandeels in the North Sea. By deriving origin- and destination-centrality measures, we highlight a set of key origin sites within the area occupied by the stock that contribute immigrants to many other sites and also identify patches that are particularly isolated, unlikely to receive immigrants from elsewhere. We show that the connectivity characteristics of the stock have a strong impact on how rapidly it recovers following a major harvesting event that leads to a patch depletion. Furthermore, the recovery of a local population will depend on the demographic status of the sites from which it can obtain immigrants. Thus, sites that provide strong out-centrality (especially if they themselves have weak in-centrality) and sites that are especially isolated should be harvested less heavily. To reduce the potential for local or regional stock collapse, models incorporating both biophysical dispersal and local demography are needed to support spatially explicit management of commercial marine species.