This paper discusses a method for warping spectral batch data. This method is a modification of a procedure proposed by Kassidas et al. [AIChE Journal 44 (1998) 864; Journal of Process Control 8 (1998) 381]. This iterative procedure is based on the dynamic time warping (DTW) algorithm. The symmetric DTW algorithm is discussed in this paper. Kassidas defined a certain weight that is received by every process variable in the DTW algorithm. However, high weights are received by process variables that contain no warping information. Therefore, a new definition of these weights is presented. These new weights take into account the amount of warping information of every process variable. The DTW algorithm using the new weights is compared to the procedure suggested by Kassidas. Furthermore, some aspects of this algorithm are optimized for speech recognition, but seem to be not necessary for warping batches. This concerns the normalization of the distance function. This step can therefore be omitted for warping batch data.