Log analysis is an unobtrusive technique used to better understand search behavior and evaluate search systems. However, in contrast with open web search, in a vertical search system such as a digital library or media archive the collection is known and central to its purpose. This drives different, more collection-oriented questions when studying the logs. For example, whether users need different support in different parts of the collection. In a digital library, the collection is categorized using professionally curated metadata. We conjecture that using this metadata can improve and extend the methods and techniques for log analysis. We investigate how to identify different types of search behavior using the metadata explicitly, how to explain and predict user interactions for the different types of behavior found, and finally how to communicate our research results to domain experts.