Abstract Restoring woody vegetation on degraded agricultural land is a widespread and common ecological restoration practice. However, highly variable plant survival and growth limit outcomes for many projects. Inconsistent reporting and monitoring of projects mean that an assessment of the relative importance of community‐assembly processes is limited, particularly over longer timescales. We use 7 years of monitoring data of nearly 2000 native trees and shrubs in a restoration project on ex‐agricultural land in south‐western Australia to test the potential effects of facilitation or competition from neighbouring plants, as well as look for patterns in their interaction with the attributes of individuals and species traits. Overall, plant size was the strongest single predictor of survival and incremental growth. Individual plants in neighbourhoods with higher inter‐generic basal area were more likely to survive, with this effect strongest in smaller individuals. When plants were larger, they were less likely to grow when in neighbourhoods with high intra‐generic basal area. Taller‐growing plants (higher species maximum height) were more likely to survive when individuals were small (basal area of 1–10 cm2), compared with shorter growing plants. Growth was also more likely in taller‐growing plants, and this relationship increased with the size of the individual. Recruitment was very low, with just 148 new recruits recorded across the 42 plots over 7 years. Maximizing the growth of plants in restorations in the early stages may promote survival and growth in the longer term. We also demonstrate that increased levels of inter‐generic neighbouring plants may improve individual plant survival in the restoration of ex‐agricultural land. As a result, we suggest tailoring direct‐seeding methods to minimize clustering of congeneric individuals. We also highlight the need to find means of promoting recruitment for the long‐term sustainability of restoration efforts.