Cooling towers are part of industrial technical building services and one of the main drivers for water and energy demands in industry. For planning and operation, the management of cooling tower systems has to focus on several objectives from strategical, tactical and operational perspective. These objectives relate to operational reliability, energy and resource efficiency as well as impacts on health and environment. Due to dynamic interdependencies within connected production systems and with the local environment, cooling tower management can be regarded as a complex task. Thus, a software-based support for decision makers involved in cooling tower management is presented basing upon the new concept of cyber-physical production systems. Enabled by a systematic data acquisition procedure, this approach combines comprehensive methodologies using white-box and black-box models. For a thorough system analysis and performance prediction, three modules have been developed within the cyber system: The system inventory and monitoring module allows to monitor the dynamic cooling tower system behavior and to calculate key performance indicator related to its operational performance. In the simulation module, a white-box model is used to test different alternatives for operational improvements in terms of energy and water demands. By means of the data analytics module, a black-box approach is applied to predict the system behavior with high accuracy basing on historical data. The full potential of the concept comes to bear through a continuous and consistent data basis as well as the synergetic application of the methodologies. As result, the approach provides support for decision makers at every level of cooling tower management. The approach was successfully applied on an industrial cooling tower system in an automotive manufacturing plant located in Germany, proving the advantages of the complementary methodologies. Thereby, the simulation revealed significant reduction potentials of water demands (−7%) and energy demands (−27%) through adapted operational strategies. As a further result, dynamic energy demand profiles could be predicted with a high accuracy (R2 = 0.98) by using multivariate regression analyses.