Salmon hatcheries are management tools intended to stabilize declining abundance in salmon populations and sustain salmon fisheries. One key area of uncertainty is how hatchery release practices influence juvenile-to-adult survival. Data quality and quantity vary considerably among hatcheries making it difficult to assess the role of release practices. Using a Bayesian hierarchical approach, we analyzed releases and recoveries of Chinook ( Oncorhynchus tshawytscha; 21 hatcheries) and coho ( Oncorhynchus kisutch; 16 hatcheries) salmon in British Columbia from 1972 to 2017. Higher survival rates were associated with increasing weights-at-release, earlier releases of Chinook salmon, and later releases of coho salmon. The addition of environmental (sea surface temperature, Pacific Decadal Oscillation), and predator (harbour seals, killer whales) covariates did not improve model performance relative to models that used year effects to account for declines in survival and interannual variability in ocean conditions. Optimizing release practices could increase returns by 6%–245% for Chinook salmon and 5%–160% for coho salmon based on median posterior estimates for each hatchery. With these results, a large-scale adaptive management approach could be implemented to test our understanding of release practice–survival relationships and evaluate cost–benefit trade-offs.