The state-of-art machine tools incorporate a wide variety of sensors and associated signals that are used within the control system or as a process monitoring variables. Machine tool canalso be equipped with additional sensors required by customer or manufacturer with relatively no limitation. Therefore, the key issue is in “separating the wheat from the chaff”. Only those data that can be linked to machine tool failures, unintended customers’ behaviour, or (exceeding) machine loading, are suitable for further implementation in machine tool condition monitoring system. This paper uses the methods formerly known from system safety and reliability analysis – namely Failure Modes and Effects Analyses (FMEA) and its Diagnostics extension (FMEDA) – to identify such data and physical quantities. The outlined approach is supported by a practical case study on machine tool spindle condition monitoring. The proposed spindle monitoring is based on noise intensity and indirect cutting force measurement.