In this paper, we propose a model for managing tasks of R&D projects. We assume that di®erent amounts of resources may be allocated to a task, leading to different costs, and di®erent average execution speeds. The advancement of the task will be stochastic, and the manager may change the allocated amount of resources according to the way the task is progressing. The revenues will depend on the time to the completion of the task, and their expected value will follow a stochastic process. We consider that a strategy for completing the task will consist on a set of rules that de¯ne the level of resources to be chosen at each moment, according to the values of several state variables. We discuss the evaluation for completing the task, and we brie°y address the problem of finding the optimal strategy