Occupancy monitoring techniques raise questions regarding trade-offs between anonymity and accuracy. Building on a rich multi-dimensional dataset, this study explores the correlations between occupancy and energy consumption at different levels, from whole rooms to individual plugs. Data aggregation of individual plugs in zones, either according to their location or their specific functional use, has been evaluated. Confirming previous studies, room level presence proved to be strongly correlated with, in order, total plugs consumption, whole room power consumption, and lighting use. A binary presence classification using Random Forest on energy features was conducted also at room level, and showed good results. The same algorithm for a three-classes categorization of room occupancy showed acceptable results despite lower accuracy. Overall, the analysis proved that a level of granularity preserving anonymity might be sufficient and should be considered depending on the needs for occupancy information in order to reduce energy consumption.
