In the paper a sequential design rule for the regression problem based on heuristic considerations is proposed, and the effect of sequential design in the estimation problem for the linear regression model is demonstrated by means of numerical simulation. In is found that esspecially for a small number of experiments there may be attained considerable improvements in the A- and the D-criterion in relation to non-sequential experiments in the region of absolutely small values of parameters relatively to the variance of experimentation errors. The results presented in the paper make useful further investigations in obtaining sequential design policies.