Highly automated and unmanned manufacturing requires process monitoring and in-process control to prevent damage to the workpiece or machine tool due to tool failure. The positioning of sensors close to the process is crucial to the success of such monitoring. One way of achieving this in machining applications is to equip toolholders with sensor systems. The Institute of Production Engineering and Photonic Technologies (IFT) has developed a sensory tool holder based on MEMS acceleration sensors that measures radial vibrations. The sensory tool holder system can be used to monitor production processes such as milling, drilling or tapping. In order to effectively use the signals from the sensory toolholder system for closed-loop control, it is necessary to convert these signals into characteristic values. This paper shows that wavelet decomposition of process-related acceleration signals is suitable for generating such a characteristic value for wear monitoring of end mills. Long-term roughing and finishing data from a real production process were analysed for this purpose.