Replanting native vegetation is a broadly accepted method for restoring degraded landscapes. Traditionally, seed used for restoration has been locally sourced to avoid introducing maladapted plants and to minimize the risk of outbreeding depression. However local adaptation is not universal and is disrupted by, for example, climate change and habitat fragmentation. We established a common garden experiment of ca. 1500 seedlings sourced from one local and two non-local provenances of Eucalyptus leucoxylon to test whether local provenancing was appropriate. The three provenances spanned an aridity gradient, with the local provenance sourced from the most mesic area. We explored the effect of provenance on four fitness proxies after 15 months, including survival, above-ground height, susceptibility to insect herbivory, and pathogen related stress. The local provenance had the highest mortality and grew least. The local provenance also suffered most from invertebrate herbivory and pathogen related stress. These results provide evidence that no advantage would be gained during the establishment of Eucalyptus leucoxylon at this site by using only the local provenance from within the range we sampled. Our results suggest that incorporating more diverse seed mixes from across the aridity gradient during the restoration of Eucalyptus leucoxylon open woodlands would provide quantifiable benefits to restoration (e.g. 6–10% greater survival, 20–25% greater plant height, 16–45% more pathogen resistance during establishment). We demonstrated these restoration gains by embedding a common garden experiments into a restoration project, and we recommend this approach be more widely adopted because it provides an effective way to facilitate adaptive management options for restoration stakeholders based on empirical evidence.