Restoration efforts have been escalating worldwide in response to widespread habitat degradation. However, coastal restoration attempts notoriously vary in their ability to establish resilient, high-functioning ecosystems. Conventional restoration attempts disperse transplants in competition-minimizing arrays, yet recent studies suggest that clumping transplants to maximize facilitative, intraspecific interactions improves restoration success. Here, we modify the Stress Gradient Hypothesis to generate predictions about where each restoration design will perform best across environmental stress gradients. We then test the model by combining measurements of physical stress with a field experiment manipulating transplant density and configuration across coastal dune elevational zones and latitudes. In hurricane-damaged Georgia (USA) dunes, grass transplanted in competition-minimizing (low-density, dispersed) arrays exhibited the highest growth, resilience to disturbance, and dune formation in low stress conditions. In contrast, facilitation-maximizing (high-density, clumped) arrays exhibited the highest survivorship in high stress conditions. Transplant survival was significantly lower on Massachusetts compared to Georgia dunes, suggesting there are thresholds above which intraspecific facilitation cannot overcome local stressors. Thus, switching from competition-minimizing to facilitation-maximizing transplant designs with increasing environmental stress can improve multiple metrics of restoration success, suggesting our modified Stress Gradient Hypothesis offers a roadmap for how to rebuild resilient, high-functioning coastal ecosystems.