In this paper, a real-time smart metering approach is proposed that reduces the reporting frequency for a given maximum error constraint. The proposed scheme employs an adaptive change detection filter that tracks the mean trend of the local power consumption, and load estimates are reported only when the estimation error exceeds a prescribed magnitude bound. The run-length of the proposed scheme is analyzed and expressions for the average time to a report are derived. Simulation results with both simulated and real-measurement data are provided to illustrate the tracking performance, and the reductions in the reporting rate. The results show that the proposed scheme can significantly reduce the reporting rate in exchange for a small bounded tracking error.