Camera traps are widely used to collect data for wildlife management, but species-specific testing is crucial. We conducted three trials to optimise camera traps for detecting numbats (Myrmecobius fasciatus), a 500-700-g mammal. We compared detection rates from (1) Reconyx PC900 camera traps installed at heights ranging from 10-45 cm, and (2) Reconyx PC900, Swift 3C standard and wide-angle camera traps with differing detection zone widths. Finally, we compared elevated, downward-angled time-lapse cameras installed at heights ranging from 1-2 m to obtain dorsal images for individual numbat identification. Camera traps set at 25 cm had the highest detection rates but missed 40% of known events. During model comparison, Swift 3C wide-angle camera traps recorded 89%, Swift 3C standard 51%, and Reconyx PC900 37% of known events. The number of suitable images from elevated, downward-angled cameras, depicting dorsal fur patterns, increased with increasing camera height. The use of well regarded camera trap brands and generic recommendations for set-up techniques cannot replace rigorous, species-specific testing. For numbat detection, we recommend the Swift 3C wide-angle model installed at 25-cm height. For individual numbat identification, elevated, downward-angled time-lapse cameras were useful; however, more research is needed to optimise this technique.