Although research alliances (RAs) have long been seen as mutualistic and reciprocal, RAs face numerous obstacles navigating stakeholders' differing goals, incentives, and information. This longitudinal, comparative case study of two RAs uses principal-agent theory to analyze these interdependent challenges and their relationship to RA strategy and design. Findings suggest that while some RAs may be better designed to balance the competing interests of various stakeholders, increasingly contested definitions of RA effectiveness among those stakeholders have muddled RA identities. As a result, RA researchers are now often held to expectations that their organizations were not originally designed to meet. We argue that this has implications for how RAs are funded, designed, and, ultimately, evaluated.