Online service reputation measurement aggregates user feedback and related data to construct service reputations, helping users make informed choices in the absence of historical experience or complete information. However, existing reputation measurements often overlook user group satisfaction, which may lead to certain services that fail to meet the expectations of the majority being assigned disproportionately high rankings. Furthermore, by focusing only on a single time step, these methods disrupt the temporal continuity of reputation evolution, neglecting the dynamic interplay between user feedback and service reputation over time. To tackle this problem, we propose a novel dynamic reputation measurement that, for the first time, centers on maximizing user group satisfaction over the full time horizon. Specifically, we construct a Dynamic Service–User Evaluation model to capture the temporal feedback loop in which user ratings are influenced by historical service reputations, thereby preserving the temporal continuity of reputation. Building on this, the online service reputation measurement problem is formulated as a multi-agent cooperative task aimed at maximizing user group satisfaction, and a multi-agent reinforcement learning-based reputation measurement is proposed to optimize service reputations over the full time horizon. Extensive experiments demonstrate that the proposed method achieves the highest full-horizon User Group Satisfaction across all tested service scales, outperforming six representative baselines.