While the principles of sustainable development are widely accepted, considering these principles effectively during implementation planning and performance measurement remains a challenge. We argue that predominantly-used results-based approaches, which monitor performance against predefined targets and indicators, are ill-suited to performance management of complex systems, such as sustainable land management. These approaches tend to cause distortions and may even constrain performance. As an alternative, we propose a decision analysis framework, Stochastic Impact Evaluation (SIE), that considers multiple goals and trade-offs. This framework produces decision recommendations that are based on the current state of knowledge and designs measurements to reduce decision uncertainty. We describe how SIE can be used to prioritise measurements through value of information analysis and to evaluate impact and improve adaptive management. Using synthetic case studies, we illustrate how SIE could be applied to guide countries and organizations towards meeting their commitments to restore millions of hectares of degraded land. Adoption of SIE would help countries evaluate intervention alternatives against multiple outcomes, minimize implementation risks, and measure performance in terms of overall return on investment. We evaluate the widely promoted United Nations Land Degradation Neutrality framework and its Target Setting Programme, which has been adopted for monitoring Sustainable Development Goal Target 15.3, and indicate how SIE could overcome many of its shortcomings. We recommend that performance evaluation of land restoration initiatives should focus on decision quality and adaptive learning rather than only on final results against targets. Finally, we suggest actions to increase adoption of decision analysis for development.