How well can indicator groups, as sets of species with well-known distributions, represent the distribution of overall biodiversity to select networks of areas for conservation? In the literature, reliable indicator groups of complementarity have proven difficult to find, with some taxon-based indicator groups resulting in no more species represented than when areas are chosen at random. We rigorously test which quantifiable characteristics of species make them better components of indicator groups of complementarity in area networks. We find that even indicator groups comprised of randomly chosen, taxonomically unrelated species perform better than randomly chosen areas, and we demonstrate the improved efficiency of protected-area networks possible when using indicator groups chosen on the basis of species' characteristics.