To effectively develop, implement, and scale social innovations, innovation orchestrators must select appropriate innovation strategies. We propose that (responsible) innovation strategies should be tailored to the complexity of the problem they aim to address to avoid causing harm to society and the environment and instead regenerate them. Yet, tailoring innovation strategies to problem complexity remains challenging. In this paper, we ask how orchestrators can differentiate innovation problems and adapt their strategies accordingly. Through a mixed-method approach, incorporating action learning and a quasi-experimental design with pre-and post-tests, we co-developed an innovation problem-classification framework alongside a team of seasoned innovation orchestrators. We subsequently tested the framework's efficacy across three distinct participant groups: (1) innovation orchestrators with over five years of experience, (2) innovation orchestrators with 0-5 years of experience, and (3) entrepreneurship students with no orchestration experience. Our findings demonstrate that the training has positive significant effects on the accuracy of classification of innovation problem typologies. We provide insights into innovation problem classification and education practices to enhance responsible social innovation. @en