It is rarely cost-effective to survey invertebrates for use in systematic conservation planning activities. The efficiency of sampling methods needs to be improved, and this is especially important at landscape and regional scales. We investigated two methods that could be used to improve regional scale sampling efficiency using a case study of ants, beetles, flies, bugs, spiders and wasps from the semi-arid Pilbara region of Western Australia. First, Generalised Dissimilarity Models (GDMs) were used to divide the region into landscapes with relatively homogeneous communities and environmental conditions. We found that some of these landscapes were large, and a low sampling density could be employed in these areas due to the low spatial turnover in species. Other landscapes were 1-2 orders of magnitude smaller, and a higher sampling density should be employed to capture the high species turnover and unique species in these areas. Variation of sampling density based on landscape dimensions could vastly improve survey efficiency. Second, we investigated whether one large sample or five small samples were a more efficient method to estimate the species composition of each landscape. We found that five small samples captured a higher proportion of landscape scale species richness for a fixed sampling effort, and was therefore a more efficient method to determine the species composition of the landscape. Combining five small samples also resulted in less sample variability than one large sample, which increases statistical power to detect changes. We concluded that GDM was an effective method to increase sampling efficiency, because it allowed sampling density to vary according to the spatial turnover in species. Using many small samples is a more efficient method to capture the species composition of landscapes than a single large sample with an equivalent sample size.