We analyze in a two-phase model the optimal mix of ex ante experimentation and ex post learning for the dynamic adoption of activities with uncertain payoffs. In a first pre-introduction phase, costly experimentation is undertaken to decide whether to adopt the activity or abandon...