Acoustical comfort inside open-plan offices is necessary for optimal work performance. In fact, it is well known that productivity is closely related to the acoustic conditions of the working environment. Noise inside offices can be due to various sources: HVAC devices, office equipment (printer, phones, etc.) and anthropic activities (human noise). The combined effect of these noise sources may play a key role in privacy metrics as the spatial decay of speech intelligibility. Starting from sound level measurements, noise sources are identified using statistical methods. The probability distribution of each source can be highlighted, conferring a different sound pressure level to each one. In order to investigate the dynamic behaviour of privacy criteria, these analyses are carried out for each octave band frequency. This work presents the results of different statistical algorithms, applied to data coming from several measurement campaigns in different offices.