Translocations to restore populations of endangered species are an important conservation tool, but a reliable diagnosis is needed to assess their success. We used capture-recapture modeling to analyze the adult apparent survival of released and resident tortoises in two translocation projects in Spain monitored for 14 and 29years. We tested if long-term survival rates differ between released and resident individuals, if survival was lower during the phase of establishment (i.e. release cost), how long acclimation lasts and if increased density due to releases affects survival. We found lower survival of released tortoises during the phase of establishment (1 to 3years) when residents were already present. After establishment, survival was very high and unaffected by density-dependence. Body condition before release was similar between recaptured and dead/missing tortoises, and did not predict establishment survival. Stochastic population viability analysis showed that success when releasing small numbers of individuals strongly depends upon adult long-term survival. Release of small second batches of tortoises was not sensitive to a growing population, regardless of its release timing. Our results highlight long-term survival as crucial in translocation projects of long-lived species, invalidating short-term (first year) survival assessment, when survival release cost does not match long-term survival. A release cost of different duration should be included in model estimation before modeling predictions. Releasing tortoises (for welfare of captive individuals or for mitigating human negative impacts) in an already established population is not recommended under most circumstances. Acclimation cost is followed by survival approaching wild counterparts. If this milestone is not achieved, the project needs to be carefully assessed to adopt other management options or should be stopped altogether.