In this paper we describe a novel approach for searching large data sets from a mobile phone. Existing interfaces for mobile search require keyword text entry and are not suited for browsing. Our alternative uses a hybrid model to de-emphasize tedious keyword entry in favor of iterative data filtering. We propose navigation and selection of hierarchical metadata (facet navigation), with incremental text entry to further narrow the results. We conducted a formative evaluation to understand the relative advantages of keyword entry versus facet navigation for both browse and search tasks on the phone. We found keyword entry to be more powerful when the name of the search target is known, while facet navigation is otherwise more effective and strongly preferred.