Service-oriented architecture (SOA) is designed to enable integration of heterogeneous business components, and thus is argued to improve enterprise evolution and agility. In order to identify service-oriented system evolution taxonomy, it is suggested to build on complex adaptive systems (CAS) theory that interprets phenomenon of evolution and emergence of new properties. This required a survey of service-oriented system design and development techniques and principles, and their mapping to CAS evolutionary principles. Based on this analysis, the paper identifies four major categories of service-oriented system evolution a service structural properties, e.g., flexibility of service, loose coupling, re-usability, etc. b enterprise enabling factors, e.g., enterprise architecture, organisation structure and management style c evolution dynamics, e.g., interactions within the enterprise and with users, prototyping, iterative and incremental development in SOA realisation, etc. d controlling factors, e.g., continuous reflection, learning and adjustment supported by formalised standards in SOA programming, SOA quality attributes and SOA governance techniques.This led to deriving a process model utilising these four categories to enable service-oriented system evolution. Metrics are suggested to benchmark system structural properties and controlling factors, including SOA flexibility, SOA quality attributes and SOA governance to ensure sustainable evolution and avoiding disordered results in response to change.