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 interactions may improve 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 this conceptual model with field experiments manipulating transplant density and configuration across dune elevations 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, transplants survived best in facilitation-maximizing (high density, clumped) arrays in high stress conditions, but these benefits did not translate to higher transplant growth or resilience. In a parallel experiment in Massachusetts where dune grasses experience frequent saltwater inundation, fewer transplants survived, suggesting that there are thresholds above which intraspecific facilitation cannot overcome local stressors. These results suggest that ecological theory can be utilized to guide restoration strategies based on local stress regimes, maximizing potential restoration success and return-on-investment of future efforts.