Smart sensor technologies to monitor temperature and moisture conditions in building components can be used to give automatic warnings for abnormal high levels of moisture. Flat compact roofs are of particular interest in this context, especially due to their vulnerability to rain leakages through the roofing membrane. Installing moisture sensors in such roofs measuring relative humidity (RH) and free water may give an early warning of rain leakages or condensation owing to air leakages from indoors before damage occurs. One challenge is, however, that normal moisture redistribution in the insulation layer over the season or a day may give very high levels of RH in the top or bottom part of the insulation. A sensor system must be able to distinguish between these normal RH levels and leakage events. Together with a sensor technology and control system producer, a semi-quantitative system that defines typical or normal RH levels in the insulation layer of the roof during the year has been developed. To investigate the applicability of such a system, measurements from the compact roofs of two pilot buildings located in Norway have been analysed. The objective has been to identify connections between measured data, occurring climate conditions and seasonal variations. Results may be used to further develop and improve the algorithms of the leakage warning system.