This paper presents a data driven methodology using symbolic time series analysis to detect anomalies in an electronic product and predict its future health. Due to the complexity of the system under analysis, a multivariate Mahalanobis Distance approach is used to reduce the dimensionality of the problem as well as to capture correlations between monitored performance parameters. A non-linear dynamic Markov model is developed from the symbolic representation of system dynamics to distinguish healthy from unhealthy system states. Novel anomaly measures are used to detect existing and emerging faults and failures. A case study is presented to demonstrate the capability of the proposed methodology in real time system health monitoring. In the case study, notebook computers are exposed to a set of environmental conditions representative of the extremes of their life cycle profiles. The performance parameters monitored in-situ during the experiments are used to define a baseline for healthy systems and to identify specific parameter behavior. The baseline of healthy systems is used to differentiate unhealthy systems from healthy ones. The proposed anomaly detection methodology is verified by injecting an artificial fault into the system. Results from the study demonstrate the potential of the approach for system diagnostics and prognostics.