As part of an international empirical study we have developed a high-level theory of the structure and evolution of inter-organizational information systems (IOIS), but face two issues in testing this theory; the very large set of factors possibly influencing IOIS and need to bound a complex research object. We describe a novel solution involving an explicitly theory-driven coding of our raw empirical data to produce a new interpreted “meso-level” ground upon which to derive and test predictions. To justify this approach, we first give a pragmatic analysis which traces the problem to the traditional choice between top-down (deductive) and bottom-up (inductive) methods for linking theory and data. We then assess the validity of our approach against standard philosophical positions. We find that a hard Empiricist view commonly espoused in information systems would not admit our approach but that Critical Realism does and suggests a conceptual interpretation of our meso-level.