An alternative model for an on-line monitoring process for variables is presented in this paper. The characteristic of interest follows a normal distribution and changes of the distribution parameters are described by a two-state Markov chain. Items are produced independently and a single item is examined at every m manufactured items similar to an on-line quality procedure to monitor process for variables previously presented by Taguchi et al. [1989. Quality Engineering in Production Systems. McGraw-Hill, New York]. The objective of this study is to determine the optimal sampling interval m and the optimal control limit d that minimizes the expected cost function. A program using S-PLUS was developed to compare Taguchi’s and the alternative proposal. The alternative model provides a better performance than Taguchi’s, since it yields a lower cost and feasible parameters (control limits and sampling interval).