A key determinant of any group’s performance in such contexts as varied as product development, consulting, and craft manufacturing is its transactive memory system (TMS): that is, its shared, tacit memory system for managing and communicating information relevant to the group. Using the methodology of system dynamics, we model the relationship between TMS and productivity by leveraging the theory of learning-by-doing at both the group and individual levels. We also incorporate into the model the concepts of “group forgetting,” in which employee turnover reduces group knowledge. We also include the effects of specialization, overspecialization, and knowledge obsolescence. We then simulate the impact of each of these refinements and perform sensitivity analyses on them. Finally, we discuss several implications of this model for future research. One implication is that representing group learning processes by a single, traditional, power-law learning curve may be in many cases inadequate. Another is that the very development of a TMS may create excessive individual specialization that is detrimental to future productivity levels.