Theraison d'êtreof maximum-demand metering is first discussed and its history briefly touched on. The anomalies involved in the existing definitions and practice are pointed out. The effect of the timing of the integration period is considered, first hypothetically and then practically in the case of six actual load curves. The effect of the length of the period is then considered in the same two ways. The results are plotted and tabulated so that the effect of changes in length, etc., can be roughly predicted. The paper concludes with an experimental section embodying tests carried out by the British Electrical and Allied Industries Research Association. A number of maximum-demand meters were tested against loaded plant over certain of the load cycles, and the readings compared with the temperature rises.