How Search in Science Impacts on the Value of Inventions at Early and Late Stages in the R&D cycle

The literatures on innovation and organizational learning have identified search in prior science and technology as critical inputs to industrial R&D. Efforts to distinguish the contributions - separately and combined - of these two search
orientations are scarce, and quantitative estimates offer contradictory results. The contributions to R&D from science are particularly elusive. To achieve some transparency on these issues we study R&D in biotech drug discovery, where the
role of science is pervasive and structured into a recurrent sequence of inventions required to build a drug candidate. The patents filed on these inventions offer, through their citations to prior art, a fine-grained view of the role of science along the R&D cycle. Applying a unique text-mining algorithm we categorize a set of 1,058 patens from Scandinavian drug discovery firms into six types of drug-related inventions. Tests confirm a significantly decreasing presence of science vis a vis technology over the R&D cycle. Effects of the composition of search on the value of single inventions show notable differences. In early R&D increasing predominance of search in science detracts from invention value. In late R&D, inventions increase in value when search includes a scientific orientation. Science-based R&D is cognitively heterogeneous and builds value in forms requiring sophisticated R&D management. Our results add to the theoretical understanding of search and the role of science in innovations. They also explain why an aggregate view, as presentedin prior literature, on the value produced by science in R&D leads to contradictory or insignificant findings.