AbstractThis paper uses data on 23 process development projects in pharmaceuticals to explore the broader issue of how organizations create, implement, and replicate new routines. A framework is presented which links approaches to experimentation and the structure of underlying knowledge. Although the concept of learning‐by‐doing is well accepted in the literature, the framework here suggests that where underlying scientific knowledge is sufficiently strong, effective learning may take place outside the final use environment in laboratories (i.e., ‘learning‐before‐doing’). This proposition is tested by comparing how an emphasis on laboratory experimentation impacts process development lead times in two different technological environments: traditional chemical‐based pharmaceuticals and new biotechnology‐based pharmaceuticals. The data indicate that in chemical‐based pharmaceuticals—an environment characterized by deep theoretical and practical knowledge of the process technology—more emphasis on laboratory experimentation (learning‐before‐doing) is associated with more rapid development. In contrast, in biotechnology‐based pharmaceuticals—an environment in which process technology is often characterized as being more of an ‘art’ than a science—a greater emphasis on laboratory experimentation does not seem to shorten process development lead times. These results suggest that there is no one best way to learn, but that different approaches may be required