A frequency response function (FRF)-based sensitivity analysis is practically useful and advantageous for both model updating and structural dynamic modification. However, in the case of a large system, the updating process involves much computational effort due to the full system matrices. Another limitation is that FRF sensitivities have a non-monotonous character as a function of the input parameters, and thus divergence is possible during the iterative estimating process. In this paper, a new FRF sensitivity method focused on a substructuring approach is developed. This method, called the component receptance sensitivity method, has a computational advantage by avoiding the iterative computation of unmodified components. In addition, a stable estimating technique which minimises the possibility of divergence caused by a non-monotonous, non-linear character of receptance sensitivity is proposed. Finally, the suggested method is applied to the estimation of spring stiffness values in a spring-supported plate model. As a result, the high accuracy and efficiency of the method is verified.