The allocation of development tasks to sites is one of the most important activities in the management of global software development projects. Its various influences on the risks and benefits of distributed projects require careful consideration of multiple allocation criteria in a systematic way. In practice, however, work is often allocated based on only one single criterion such as cost, and defined processes or algorithms for task allocation are typically not used.Existing research approaches mainly focus on selected aspects such as the minimization of crosssite communication and are difficult to adapt to specific environments. This article presents a customizable multi-criteria model for task allocation in global software development projects.Based on an analysis of the state of the practice, a set of requirements was derived and used for evaluating existing task allocation models from different domains. The Bokhari algorithm was identified as a suitable starting point and modified with respect to the fulfillment of the requirements. The modification includes the development of mechanisms for customization, the incorporation of cause-effect relationships, and the use of probabilistic modeling of uncertainty with Bayesian networks. The application of the model is demonstrated in different scenarios that represent typical hypothetical and real distribution decision problems in industrial contexts.Experience from applying the model to such problems has shown, for instance, that depending on the weight of different criteria, very different task distributions will result. This demonstrates, in consequence, the need for systematic multi-criteria task allocation support in global software development.