Soft independent modelling of class analogy (SIMCA) is applied to identify near-infrared (NIR) spectra of ten excipients used in the pharmaceutical industry. For each class at least 15 excipient samples were collected for the data base, considering different batches and occasionally various suppliers. Therefore the data of the classes are not always homogeneous. The performance of the original SIMCA method, which is usually described in the literature and also applied by the users, carried out at two confidence levels, 95 and 99%, on original data, SNV (standard normal variate transformation) and second derivative pre-processed data, is discussed. Reasons for the rejection rates are given. No objects were assigned to a wrong class using SIMCA.