Abstract In process control systems (PCS) both the direction and magnitude of controller setpoint drift (SPD) are stochastic variables. In the conventional failure analysis of PCS only the direction of setpoint drift is under consideration (i.e. setpoint drifts high, setpoint drifts low) and failure model construction is often based on a static description of system operation (e.g. digraphs). Dynamic failure models are used to investigate the relevance of the magnitude of SPD to the failure characteristics of two simple level control systems with: (a) a single proportional controller, and (b) two on/off controllers. The failure models represent the probabilistic system behaviour in continuous time and continuous state space of the controlled variables. Model construction is based on the description of PCS operation by first-order differential equations, with the failed system states regarded as absorbingg states. Analytical expressions for MTTF and asymptotic values of Cdf for failure in different modes are obtained and compared to those obtained from the conventional approach to PCS failure analysis. The results show that: (a) PCS failure characteristics can significantly vary by the changing magnitude of SPD, (b) SPD effects also depend on the type of controller used, controller parameters, system failure margins and dynamics, (c) the convectional approach may overestimate PCS failure due to SPD, and (d) a static description of system operation may not be appropriate for PCS failure analysis, especially if design choices are based on the predicted failure characteristics. Dynamic failure model construction for more complicated types of PCS is discussed.