Population models, using empirical survival rates estimates for different life stages, can help managers explore whether various management options could stabilize a declining population or restore it to former levels of abundance. Here we used two decades of data on five life stages of the Cedar River, USA Sockeye Salmon, Oncorhynchus nerka, population to create and parameterize a life-cycle model. This formerly large but unproductive population is now in steep decline, despite hatchery enhancement. We gathered population-specific data on survival during five stages: 1) egg-to-fry, 2) fry-to-presmolt, 3) presmolt-to-adult return from the ocean, 4) adult en route from the ocean to the spawning grounds, and 5) reproduction. We ground-truthed the model to ensure its fit to the data, and then we modified survival and other parameters during various stages to examine future scenarios. Our analyses revealed that low survival of juveniles in Lake Washington (stage 2: averaging only 3% over the last 20 years), survival of adults returning to fresh water to spawn (stage 4), and survival of adults on spawning grounds to reproduce (stage 5) are likely limiting factors. Combined increases in these stages and others (specifically, the proportion of fish taken into the hatchery to be spawned) might also recover the population. As in other integrated hatchery populations, managers must weigh options relating to balancing the fraction of natural- and hatchery-origin fish, and our results showed that increasing the fraction of fish taken into the hatchery alone will not recover the population. Our model brings together population-specific data to help managers weigh conservation strategies and understand which stages and habitats are most limiting and how much survival must increase to achieve recovery targets. By extension, our analyses also reveal the utility of such models in other cases where stage-specific data are available.