An improved formulation for stochastic steady state optimizing control problems is presented. Owing to its different structure, it is applicable to a larger group of practical problems than an earlier version (Lin et al. 1989 a). Theoretical analysis is carried out to investigate the viability of the new formulation. A modified approach to stochastic steady state optimizing control problems under such a formulation is also presented. Here, a modified steady state identification technique is employed to estimate the mean derivative that is required in calculating the modifier. The dynamic information acquired from the real process between any two successive controller set point changes is used. The advantage of this approach is that it significantly reduces the required number of controller set point changes and is much less sensitive to noise than the modified two-step methods. Optimality and global-convergence conditions are provided. Consistency and asymptotic behaviour of the modified steady state identification method are also investigated. Comprehensive computer simulations are provided not only to show the effectiveness and reliability of the presented optimizing control algorithm, but also to demonstrate the accuracy of the steady state estimation. Comparisons are made between this approach and the modified two-step methods, showing that the presented approach is very efficient and reliable.