Classical control charts detect when a process loses its stability, but they do not give indications on the gravity of the instability and on the risk incurred if the process is not halted immediately. We suggest to use a statistical model and a window of data to evaluate the probability that the next item will not respect the specification limits. This measure of risk can then be used to decide whether the process should be halted immediately or if production could continue until it is possible to fix the process. The properties of this method are explored numerically and a case study is provided.