Consumers are increasingly using Internet portals when searching for relevant health information. Despite the broad range of health information portals (HIPs) available, usage problems with such portals are still widely recognized and reported. In this study, we analyzed usage data from an operational health information portal and identified ways in which these problems can be addressed. While previous usage data and log analysis research has focused more on user behaviors, query structures, and human-computer interaction issues, this study covers more comprehensive issues such as content. We describe a taxonomy of usage issues derived from a literature analysis. We describe how we validated and refined the taxonomy based on analyzing the usage data from an operational health portal. Findings from the usage data indicate that a range of content issues exist that lead to unsuccessful searches. The analysis also highlights that users’ ineffective information seeking strategies are not well supported by the system’s design. We use this taxonomy to propose a usage-driven, consumer-centered approach for dynamic improvements of HIPs. We also discuss the study’s limitations and directions for future research.