This study was devoted to the prediction of polymer material aging. The prediction of the shelf time of tire rubber is used as an example in this article. The main steps of the whole procedure are described. They are the design of the experiment, accelerated aging testing, the construction of a multiresponse mathematical model and parameter estimation, and the extrapolation of the model in real-life settings. The main pitfalls were deduced, and techniques to overcome these pitfalls are described. Novel methods of data modeling, such as evolutionary design of experiment and successive Bayesian estimation, were used. (C) 2005 Wiley Periodicals, Inc.