In this paper, it is assumed that the mean of a normal process
is monitored by a CUSUM control chart. When the control chart triggers a
signal and declares that the process has gone out of control, a search process
is started to find the time of change and the causes of going the process out
of control. Several methods (plans) for finding the true (real) change point
is proposed. It is shown that the plans which are based on the likelihood of
the points in time perform better.