Data from sensor based analytical instruments (e.g. electronic nose) can be arranged in three dimensions as sample×time×sensor. Due partly to limitations in software, partly to general practice, the dimensions of the data are often reduced in the time mode and traditional two-way chemometric models (e.g. principal component analysis (PCA)) are used for the exploration of the data. Thus the internal relationship between the different modes (samples, time and sensors) is destroyed and the result can be an incomplete data analysis.A new approach to handle the time information is introduced combining new and more advanced multi-way chemometric models with traditional pre-processing techniques as an alternative to PCA-like methods. As the pre-processing is an essential but often time-consuming part of the data analysis, a semi-automated approach has been used to make comparison of multiple analyses simple.The main feature of this new approach is the exploration of the total time profiles such that potentially relevant information is not discarded by feature extraction before the actual data analysis. Multi-way modelling is evaluated and for sensor based data, it is shown that the so called PARAFAC2 multi-way model that handles shifted profiles offers some advantages compared to PCA and alternative multi-way methods.The principles are exemplified with an example from the licorice industry. An algorithm and graphical user interface has been made available at http://www.models.kvl.dk to ease testing this new approach.