Abstract: Brand alliance is one of the effective strategies for new ventures to overcome their growth disadvantages. In order to improve the success rate of brand alliance partner selection for new ventures, new ventures' brand partner selection mechanism is explored using computational experiments method.Firstly, based on the complex adaptive systems (CAS) theory, process model of brand alliance partner selection is built. Secondly, the agent behavior conceptual model of new venture is constructed as the core agent in brand alliance system, and its interaction multi-agent model with other partnersand environmental agents. Thirdly, the external competitive environment and the agents' key characteristics which effect the formation of the alliance are focused on, such as the partner venture's type, brand equity, relationship commitment and brand alliance fit, argues that the brand alliance partner selection is a selection and discovery process of rules. Finally, the brand partner selection mechanism'computational experiment is computed. The results show that as the external environment changes dynamically, more and more new ventures will choose brand alliance strategy, but the choice between mature enterprises and new venture as a joint partner is not affected by the change of external environment. When a new venture chooses a partnership, its characteristics do not affect whether it adopts an alliance strategy. Regardless of its characteristics, the core venture prefers to the mature company as a partner; at the same time, the partnership commitment has the most significant impact on the partner selection, and the brand equity is the second one. The effect of brand alliance fit is not significant. When responding to the alliance request of the core enterprises, the candidate agents mainly depend on the degree of the evaluation by the core enterprises, followed by the brand equity. The research results have significant value for the management practice of new ventures.

Key words: new venture, brand alliance, partner selection, CAS theory, computational experiment