It is claimed that any data, especially time-sequenced data are of use only if they are homogeneous that is, only if they demonstrate control chart stability (CCS). The requirement of CCS is not always there for either predictions or baseline estimates for process improvement.
The two types of predictions known are interpolation and extrapolation. Control charts have a role in predictions, but the role is restricted. When making predictions requiring CCS, the moment the chart shows instability the predictions must stop. To make any further predications again the process must be first stabilized. CCS implies that we can have a prediction made from stable processes within a scope of mistake defined as the control limits of the chart. Each CCS period erroneously predicts when instability occurs and can only predict for a restricted time.
Control charts limits do not intrinsically offer predictive utility. A process can be stable with very extensive limits. Also, control charts limits can even have values that cannot occur in practice. When all the data for a period is used, CCS can recognize special-cause variation signals barring when a sample of rational subgroups is used.The idea of analytic studies is to understand the causal relationship between the conditions and the desired results. As knowledge enhances, the ground for predictions have to change from patterns to correlations to causalities.
The order of studies for process improvement is enumerative-analytic-enumerative. Enumerative is to institute baselines and analytic to find reasons. After a process change, a new baseline is deliberate. CCS does not hold good for determining baselines or making comparisons when the process exhibits instability.
CCS is accepted, preferred and immensely promoted for there are benefits to stable processes. A stable process needs less attention. The use of control charts is considered beneficial for monitoring processes that are preferred to be stable. It would still be better controlling process to have the desired effect and aiming process improvement.
See full story on isixsigma.com