This paper provides an in-depth analysis of the event-based metering strategy proposed by Simonov et al. This strategy is an alternative to the traditional periodic (time-based) metering where the power demand is averaged in fixed time periods (e.g. every 15 minutes). The event-based approach considers two thresholds that trigger an event, one related to the (instantaneous) power demanded, other to the accumulated energy consumed. The original work assumed these thresholds fixed for the measurements. Our present contribution relaxes this assumption by proposing a method to set the thresholds from the percentage of the peak power consumption over the period under analysis. This approach, in contrast to the time-based and the fixed thresholds, better captures the actual power demanded when different households with diverse power demand profiles are studied. In this sense, our method provides a more efficient way to store electricity demand data while maintaining the estimation error (in relation to the real-time power demand) under acceptable values. Numerical examples are presented to illustrate the advantage and possible drawbacks of the proposed method.