This paper investigates the effectiveness of Bayesian updating (i.e. the process of improving initial probability estimates by incorporating data from real operation) in complex and dynamic systems. A mathematical model including various types of dynamic input (i.e. variable time-dependent failure probability) was developed in order to test whether decision making based on Bayesian updating would provide better performance, by means of lower failure probabilities and/or lower cost. This investigation showed that using Bayesian updating (with the assumptions of uniform probability distribution and independent events) does not lead to better results, on the contrary in many cases in can lead to a much inferior performance, which is a result of certain deficiencies of this process in dynamic systems. (C) 2010 Elsevier Ltd. All rights reserved.