This article investigates the determinants of research organizations’ performance. A theoretical construct is built based on the traditional “environment–conduct–performance” framework applied to industrial organizations and operationalized by using information from a recent extensive survey of research organizations. From an empirical point of view, using hierarchical clustering techniques, we first underline the diversity of research organizations in terms of structure and conduct parameters (in order to better identify their “behavioral patterns”) and then investigate the link between these classifications and institute performance by means of binomial regression models. From a managerial point of view, this research could help to formulate policy and management recommendations for improving performance. This article addresses two critical issues: 1) Performance measurement of research organizations remains a challenge despite multiple efforts made by academics, research organizations, management, and policymakers. The broad range of impacts resulting from research institutes can be tacit or codified, direct or indirect, among other characteristics. In many cases, these impacts are related to humans, which makes them a) difficult to measure as such, and b) involves challenges in terms of the causalities of impacts. 2) The determinants of research organization performance remain a partial black box.