This dissertation theorizes how IS portfolio drift emerges, what its consequences are, and how it is managed. I define IS portfolio drift as locally-driven but centrally-unintended adaptations of an organization's collection of IT assets. These changes comprise workarounds of officially-mandated applications and the implementation of alternative systems. They occur mainly because of a poor fit of the officially-mandated technology with the task it is intended to support. Drift can be costly: the expenditure on developing, purchasing and supporting the officially-mandated applications is squandered, while additional funds are needed to maintain the shadow systems. Decision quality could also be reduced, because of uncertainty over data provenance. However, drift could also be beneficial to an organization, as it reflects the innovativeness of employees in adapting to changes in local environments or in using emerging technologies to enhance their performance. Traditionally, research on IS governance has taken a rational and agency-oriented perspective. I conceptualize IS governance from a practice perspective to highlight how portfolio drift is an ongoing, macro-level outcome of the micro-level actions of various agents. This analysis reveals that the process of governance is influenced by the relative allocation of power (capital) between users and managers, and between the operational functions and the IT unit, how this allocation came to be, and how it changes over time. This study uses case studies of ten units within two organizations to delineate a process theory of portfolio drift, and explain the situated practices used to steer the portfolio-in-use so that it meets the goals of the agents involved.