In countries at all levels of development, assessing the opinions of health service users is increasingly promoted as an integral part of quality evaluations. However, there has been much debate on how best to measure user opinions. This article discusses findings from a study in South Africa, which employed both closed-ended facility exit interviews (total 337) and open-ended community-based focus group discussions (total 14) to obtain users' views on the same set of primary care providers. We outline various difficulties encountered in the interpretation of the data. First, in the absence of explicit and universal standards, users evaluated providers against their experiences with other health care services available to them in their areas. Responses were thus highly context specific, dependent on the particular configuration of services in each site. Secondly, the focus group discussions provided a very different (generally more negative) picture of providers to the exit interviews, suggesting that where and how views of health services are elicited has a large bearing on the results obtained. Thirdly, the focus group discussions appeared to encourage dramatic representations of what was, on observation, a banal everyday reality. Both methods defied superficial reading, and each appeared to have limitations in establishing the 'truth' about people's opinions. We conclude that there is a need for greater recognition, in quality assessments and in quality assurance, that user and community opinion is a social rather than a technical phenomenon. As such it is dynamic, bound to particular contexts and difficult to capture in single, 'snap-shot' assessments, no matter how well designed. In the context of quality assurance programmes, time spent assessing user views may be better used in other ways such as training and supporting health care workers to engage directly in dialogue with communities around needs and expectations.