Biological wastewater treatment plants (WWTP) are complex systems to assess. Many parameters are recorded daily in WWTP to monitor and control the treatment process, providing huge amounts of registered data. A combined approach of extracting information from the WWTP databases by statistical methods and from the sludge physico-chemical characterization was used here for a better understanding of the WWTP operation. The monitored parameters were analysed by multivariate statistical methods: Principal Components Analysis and multiple partial linear regression. The WWTP operational conditions determine the sludge characteristics. The bacterial activity of the sludge in terms of extra-cellular polymeric substances (EPS) production was assessed using size exclusion chromatography and the internal structure of sludge flocs was observed by confocal laser scanning microscopy. The diagnosis of three paper mill WWTP enabled the identification of an important EPS production, the presence of the nitrification process and the presence of PO(4)(3-) nutrient in WWTP-A. These three main characteristics of WWTP-A were related with a systematically good sludge settling. In WWTP-B and C with bad settling, the bacterial activity was weak.