The success of marine stocking programs hinges on releasing hatchery-produced juvenile animals into the optimal marine habitat. This study sought to identify optimal microhabitat features of coastal seagrass meadows for juvenile sea cucumbers Holothuria scabra, a species cultured widely for stock restoration, sea ranching and sea farming. Groups of 25 juveniles were released into 30 replicate 1 m² open sea pens embedded into sediments in a coastal bay of New Caledonia at sites with varying biotic and physical features. Survival after 8 to 10 d ranged from 0 to 100% and averaged 77%. Boosted regression tree analyses found that different variables affected survival, growth and burying behaviour. Survival was significantly higher at shallow depths with intermediate seagrass cover (~42%). Growth rate was significantly higher at comparable intermediate seagrass density (~34%) and for smaller juveniles, presumably displaying compensatory growth. Burying frequency of juveniles was largely explained by habitat variables, notably a high organic carbon content of sediments and shallow seawater depths. Juveniles survived better where they buried more frequently, providing empirical evidence of a predator-avoidance mechanism. Our findings reveal that marine animals can display non-linear responses to habitat features such as seagrass cover. These marine invertebrates should be released in habitats that optimize survival, growth and behaviours and be spread among multiple sites to mitigate against stochastic mortality events. This experimental approach offers clear advantages over factorial designs for identifying optimal habitats for captive-release programs.