Evaluating collaborative systems remains a significant challenge. Most evaluation methods approach the problem from one of two extremes: focused evaluation of specific system features, or broad ethnographic investigations of system use in context. In this paper, we develop and demonstrate a middle ground for evaluation: explicit reflections on scenarios of system use coupled with analysis of the consequences of these use scenarios, represented as claims. Extending prior work in scenario-based design and claims analysis, we develop a framework for a multiperspective, multi-level evaluation of collaborative systems called SWIMs: scenario walkthrough and inspection methods. This approach is centered on the capture, aggregation, and analysis of users' reflections on system support for specific scenarios. We argue that this approach not only provides the contextual sensitivity and use centricity of ethnographic techniques, but also sufficient structure for method replication, which is common to more feature-based evaluation techniques. We demonstrate with an extensive field study how SWIMs can be used for summative assessment of system performance and organizational contributions as well as formative assessment to guide system and feature re-design. Results from the field study provide preliminary indications of the method's effectiveness and suggest directions for future research.