Berliner (1987) discussed the issue of controlling the output (response) towards the specified value by choosing the values for independent variables in a regression mixture model, taking it as a Bayesian Decision Problem. The quantification of the potential loss was done with the help of quadratic loss function, which was a symmetric loss function. We have tried to quantify this loss with the help of Precautionary Loss Function and Modified Squared Error Loss Function, in linear Scheffé mixture model and comparison is established between these loss function. Results are improved as compared to Berliner (1987).