As the pace of business change increases, service-oriented (SO) solutions should facilitate easier maintainability as underlying business logic and rules change. To date, little effort has been dedicated to considering how the structural properties of coupling and cohesion may impact on the maintainability of SO software products. Moreover, due to the unique design characteristics of Service-Oriented Computing (SOC), existing Procedural and Object-Oriented (OO) software metrics are not sufficient for the accurate measurement of service-oriented design structures. This thesis makes a contribution to the field of SOC, and Software Engineering in general, by proposing and evaluating a suite of design-level coupling and cohesion metrics for predicting the maintainability of service-oriented software products early in the Software Development LifeCycle (SDLC). The proposed metrics can provide the following benefits: i) facilitate design decisions that could lead to the specification of quality SO designs that can be maintained more easily; ii) identify design problems that can potentially have a negative effect on the maintainability of existing service-oriented design structures; and iii) support more effective control of maintainability in the earlier stages of SDLC. More specifically, the following research was conducted as part of this thesis: - A formal mathematical model covering the structural and behavioural properties of service-oriented system design was specified. - Software metrics were defined in a precise, unambiguous, and formal manner using the above model. - The metrics were theoretically validated and empirically evaluated in order to determine the success of this thesis as follows: a. Theoretical validation was based on the property-based software engineering measurement framework. All the proposed metrics were deemed as theoretically valid. b. Empirical evaluation employed a controlled experimental study involving ten participants who performed a range of maintenance tasks on two SO systems developed (and measured using the proposed metrics) specifically for this study. The majority of the experimental outcomes compared favourably with our expectations and hypotheses. More specifically, the results indicated that most of the proposed metrics can be used to predict the maintainability of service-oriented software products early in the SDLC, thereby providing evidence for the validity and potential usefulness of the derived metrics. Nevertheless, a broader range of industrial scale experiments and analyses are required to fully demonstrate the practical applicability of the metrics. This has been left to future work.