Unpredictability in environmental dynamics may be the greatest challenge of our time. However, sustainability initiatives are designed with the assumption that people’s access to natural resources is relatively predictable. This assumption is based on the principle that users must have the right to exclude free riders through clear boundaries to use resources sustainably. Lack of boundaries is linked to the tragedy of the commons. However, clear boundaries are not always possible, especially in areas with unpredictable ecosystem dynamics. How to achieve sustainability in unpredictable systems is still poorly explored. To answer this question we evaluated 11 SES through interviews.For each SES identified, we selected at least three experts to assess the level of predictability faced by the human groups in each SES. We used a three-step interactive process. First experts were asked to define a measure of a good day for the specific SES. We defined “good day” as a day or hunting/gathering/fishing excursion for which a household of that specific SES would feel satisfied with the quantity of resources they had accessed. The “good day” measure was determined by consensus among all experts in each SES.Following the consensus on what would be a good day in each SES, we asked experts to estimate the level of predictability of the human group in achieving the “good day” target. We used a structured expert elicitation protocol (the Investigate, Discuss, Estimate, and Aggregate (IDEA) protocol) to establish this accurately. The IDEA protocol comprises two phases. Initially, experts provided an average score, a confidence interval, and the likelihood that the confidence interval accurately represents the real value (ranging from 50% to 100%). The question we asked was about the number of good days/trips out of 10 that a household in that specific SES would have (not necessarily 10 consecutive days, but rather 10 consecutive attempts to access and use resources).Following individual consultations with experts from each SES we normalized all answers to 80% confidence interval. Then, we showed all answers to the experts of the same SES to potentially adjust their responses based on aggregated findings. This iterative process enabled experts to compare and refine their assessments. The final score was the average number of good days, and average confidence interval of all experts’ answers.We repeated the same process regarding the impact of implementing clear boundaries in that specific SES. We asked them to measure the quantity of “good days” in case strict boundaries were implemented in that specific SES (questionnaires in supplementary material)The number of experts interviewed for each case study ranged from three to six, totaling 48 entries from 35 experts. Each expert was interviewed at least three times. Interviews were carried out between June 2022 and December 2023. Interviews were conducted in Portuguese, English, and Spanish. Ethical approval was obtained from the Ethics Committee at Imperial College London (approval number 21IC7307).