Millions of hectares of land worldwide is in urgent need of restoration. However, many past landscape and ecosystem restoration efforts are likely to have been less than effective. Some of the reasons for these problems include a lack of monitoring, or being subjected to poor quality monitoring practices, or that the restoration efforts are too often not linked with ecological concepts and/or theory. These deficiencies make it difficult to: (1) identify general principles for broader application beyond site-specific insights, (2) quantify success, and (3) identify general ways to improve restoration programs. In this paper, I outline some personal perspectives on how these deficiencies might be addressed using insights from two decades of work on restoration in the temperate woodlands of southeastern Australia. I argue problems associated with a lack of monitoring and limited connection to ecological theory need to be rectified if future landscape and ecosystem restoration programs are to be more successful than those in the past and are to deliver better ecological returns on investment. This should help enhance learning and boost the potential for continuous improvement in restoration efforts.