Among recent developments in the field of software reuse has been the increasing reuse of coarse-grained components, and it has been proved that granularity has great impact on component's reuse performance. However, previous studies have ignored rigorous and effective methods to support coarse-grained component identification and design, particularly granularity optimization design. In this paper, a stability-based component identification method, STCIM, is presented to resolve this problem. First a feature-oriented component model and the corresponding component granularity metrics are briefly presented. By establishing mappings between business model space and component space, component design process may be regarded as the process of decomposition, abstraction and composition of business model elements, with four different mapping strategies discussed to obtain dynamic component granularities. Furthermore, it is thought that component granularity is closely correlative to the stability of business models: the more stable the business model, the larger the corresponding component granularity may be. A metrics for model stability with three factors, i.e., number of isomers, stability entropy and isomer similarity, is presented, and the corresponding component identification algorithm based on <i>Most Stable Set</i> is discussed in details. Finally a practical case is described to validate the method in this paper.