IT service providers are faced with a dilemma when trying to ensure proper function and effective operation of their clients' infrastructure. On one hand, frequent changes to the IT infrastructure are required to ensure smooth operation; on the other hand, studies show that changes are responsible for 80% of all incidents that result in client outages. This paper proposes a novel methodology for investigating the role of change in incident prevention. We provide a detailed analysis of the change-incident space, offer algorithms on linking incidents to changes that caused them, and show how such data can be effectively used to build predictive models for incident prevention. We conclude by presenting our methodology applied to a real-world dataset and use cases.