Modelling an Atlantic salmon (Salmo salar) population as three phenotypically homogeneous sub-stocks showed different sub-stock dynamics and time-trends, a finding which was robust to three quite extreme variations of marine-mortality scenarios. A process-centred demographic analysis of a phenotypically mixed population (river North Esk, eastern Scotland) separated the life-cycle into two portions: density-dependent (spawner to smolt), represented by a stock-recruitment relationship; density-independent at sea and in both the estuary and river during return to spawning sites. Extensive data spanning three decades was judiciously used to parameterise the model for three phenotypically distinct sub-stocks (early-running multi-sea-winter; late-running multi-sea-winter; and one-sea-winter). Quantification of different population parameters was achieved for the putative sub-stocks, illuminating historical population-trends and underlying mechanisms. The approach forms a template for similar model decompositions at other rivers with similar data. Broad implications for salmon management more widely included the enhanced benefits of a process-based life-cycle model that described more phenotypically homogeneous sub-stocks (than a single-stock formulation) for the transport of population parameter values (or derived Biological Reference Points) from well documented (parameter-donor) catchments to data-sparse (parameter-recipient) catchments.