An experimental wild harvest of Cycas arnhemica K.D. Hill, an understorey plant in tropical eucalypt savannas, was conducted in central Arnhem Land, Australia. Replicated harvest treatments were monitored over 2 years with tagged individuals. A range of a priori candidate statistical models were compared to determine the effect of wild harvest and environmental factors such as fire frequency and disturbance from feral animals on survival, recruitment and stem growth. The mean density of C. arnhemica was 1630 ha^–1 and ranged from 550 to 2250 ha^–1. Harvesting intensity in the 0.04-ha quadrats varied between 10 and 32% of all cycad stems (excluding seedlings). There was no clear effect of harvest treatment on stem survival, seedling abundance and stem growth on the remaining C. arnhemica 2 years after the harvest treatments were applied. Survival of woody stems was higher than that of seedlings but there was considerable overlap among the larger size classes. Seedling abundance was generally low and variable across the three sites, and seedlings were entirely absent from quadrats burnt twice during the study period. Stem growth varied considerably across the three sites and was similar between small and large stems. We suggest that the wild harvest of this abundant Cycas species in a remote region of northern Australia will have minimal impact on wild populations if focused on juvenile stems and return time is extended to 15–40 years at harvested locations. Future harvest management of this and other Cycas species in northern Australia will benefit from further research on stochastic population models to determine the most suitable harvest strategies, particularly for smaller Cycas populations.