We develop a user-oriented reliability model, a simple yet effective approach, to address huge amounts of Web contents, and more importantly tackle users' dynamic navigation behaviors for a Web system. The users' intrinsic capabilities and behaviors, unlike programs' common recurring behaviors, call for a new approach for modeling the reliability of an organization's Web system. Our approach differs in the modeling scope of the Web system and the means of modeling user behaviors. We exploit the idea of cyclomatic complexity to cover the maximum number of independent paths, which overcome the less likely repetitive user navigations, and still maintain the dynamic nature of Web surfing. The modeling process takes Web servers' access, error, and referrer information as inputs, and translates them into a flow graph. A transition tree is then derived from the flow graph, based on depth-first traversal and dynamic programming. Accordingly, the reliability of the Web system can be computed by excluding those unreliable paths. This model is further applied to prioritize the error removal sequence to assist in allocating resources and effort.