This paper discusses the concept of dynamic time warping as a tool for supervision and fault detection with particular reference to bioprocess applications. According to the complexity of the biological reactions and the usually time-varying behaviour of the living cells, the supervision process must be insensitive to time-axis fluctuations and initially unknown parameters. While other methods, like observer-based or statistical approaches, require either mathematical model descriptions or quite a lot of historical data sets for training, the knowledge in this approach is represented in what the user defines as typical or reference pattern in the data, from a whole fermentation, from particular phases or from single events. The procedure supports the detection of single events as well as the detection of growth phases or plant failures. Especially in research applications, where a structured model is lacking, the dynamic programming approach is proposed to be employed advantageously. Results are given for high-cell-density fed-batch cultivation of E. coli and diauxic growth in batch of S. cerevisiae.