We use a sensor network of 173 sensors to monitor the energy consumption, the indoor environment quality and local meteorological conditions of three collective residential buildings composed of 62 dwellings. In particular, 146 sensors were deployed in 8 apartments and 27 sensors in the buildings common areas. The collected data are used to assess the performances of the three buildings which have recently been undergoing heavy retrofit actions. It also aims to accurately characterize some of the occupants behaviors which directly impact the buildings energy consumption and the indoor environment quality such as the occupancy patterns, the windows opening for natural ventilation or the temperature set point. We observe different occupancy patterns depending on the number of people in apartments and their schedule. As expected, we also observe a strong seasonal variation of windows opening rates and consequently in natural ventilation. Averaged indoor temperatures over the heating season are much higher than the values used in regulatory simulation scenarios. Furthermore, besides the recent buildings energy retrofit there is a total absence of time-of-day dependent heating load control which may explain a large amount of the buildings energy performance gap.