Today, a variety of user interfaces exists for navigating information spaces, including, for example, tag clouds, breadcrumbs, subcategories and others. However, such navigational user interfaces are only useful to the extent that they expose the underlying topology---or network structure---of the information space. Yet, little is known about which topological clues should be integrated in navigational user interfaces. In detail, the aim of this paper is to identify what kind of and how much topological information needs to be included in user interfaces to facilitate efficient navigation. We model navigation as a variation of a decentralized search process with partial information and study its sensitivity to the quality and amount of the structural information used for navigation. We experiment with two strategies for node selection (quality of structural information provided to the user) and different amount of information (amount of structural information provided to the user). Our experiments on four datasets from different domains show that efficient navigation depends on the kind of structural information utilized. Additionally, node properties differ in their quality for augmenting navigation and intelligent pre-selection of which nodes to present in the interface to the user can improve navigational efficiency. This suggests that only a limited amount of high quality structural information needs to be exposed through the navigational user interface.