Scientists and innovators can operate by developing and framing their own problem statements and solutions, but they can also become problem solvers to externally or pre-defined problems. Building on insights from open innovation we argue for a match between problem and solution types, such that externally defined problems benefit from boundary-spanning solutions that involve external actors, while internally defined problems do not. Using unique data on more than 2000 research grant applications to the Engineering and Physical Science Research Council (EPSRC), the largest strategic funding agency in the UK, we estimate a regression discontinuity design (RDD) model that compares near-misses with near-wins. Our findings show that scientists on investigator-initiated projects produce the highest impact research and have more funding success if they do not involve non-academic partners, while on the other side scientists on funder-initiated ('targeted') grants stimulate future research if they partner with industry. This suggests complementarity between problem formulation and solution types, such that externally defined problems benefit from boundary-spanning solutions, while internally defined problems have the highest impact when applying local solutions. These findings are of relevance to innovators and funders as they inform on the problem-solution match for impactful research and innovation.
