The emergence and implementation of edge-computing devices is one example of the challenges presented to HAZOP's “traditional” application. Legacy sensors are replaced by Internet of Things (IoT) devices that take sensor readings, pre-process them to accommodate non-linear properties, and present them for transmission over a network to a consumer. As these collections of edge devices become more commonplace, these clusters will also attract safety responsibilities. New technologies often rely on hidden interrelationships and interdependencies. Communications infrastructures taken for granted can give rise to many failure modes, including cybersecurity-related, such as “data insertion”. These failures will contribute to extended causal chains. Therefore, HAZOP should be adapted to provide structure and modularity to analyses of these large-scope, complex and data-dependent systems. This paper begins with an example of the current HAZOP methodology, developed primarily for analogue, electrical, electronic, and centralised digital systems, such as a programmable logic control (PLC). Through case studies, we establish the requirements for edge-based systems, addressing complexity through boundaries, interfaces, modularity, and, crucially, repeatability. The tenets of HAZOP remain the same. Changes are required to manage the complexities of data-driven, data-centric edge systems effectively.