In the face of grand societal challenges, collective action is indispensable. While socially-evaluative interventions—those that provide feedback based on social comparisons—have shown promise in fostering pro-environmental behavior, their long-term effectiveness and psychological mechanisms remain insufficiently understood. Drawing on social identity and social influence theories, I argue that the success of such interventions should not only be measured by their impact on behavior but also by their influence on individuals' social identity. I conducted a longitudinal field experiment with 101 citizens in a medium-sized German city to investigate how social comparison feedback on residential heating behavior affects both energy conservation and group identification over time. Participants were assigned to either a control condition with individual energy-saving tips or one of two treatment conditions receiving real-time feedback comparing their room temperatures to the group average. Group identification and heating energy conservation behavior were assessed at three time points: before, during, and after the intervention. Results reveal that socially-evaluative feedback alone did not directly affect group identification but had an indirect effect via perceived group difference—moderated by behavioral similarity. Importantly, I find that group identification moderates the effect of feedback on behavioral change over time: high identifiers increased their conservation behavior, while low identifiers decreased it. These findings suggest that social identity plays a dynamic role in determining the trajectory and stability of behavioral change. Our study contributes to the debate on the effectiveness of social interventions by highlighting the relevance of identity dynamics in shaping long-term outcomes. It calls for a shift in focus—from merely changing behavior to also understanding and supporting social identity development in interventions designed to tackle grand challenges.