Experimental evidence on the role of the built environment for promoting physical activity is important for informing how to create cities that promote active living. Parks provide opportunities for physical activity; however, there is little robust evidence on the impact of park refurbishment. Government agencies often modify parks, providing opportunities for natural experiment studies of these interventions. Such an opportunity was provided by the modification of a large park in Victoria, Australia in 2013 when the Recording and EValuating Activity in a Modified Park (REVAMP) study was established. Based on insights from the REVAMP study, this paper discusses challenges involved in conducting natural experiments in park settings, focussing on issues that may help design more effective future evaluations of the impact of park refurbishment. Natural experiments offer unique opportunities to evaluate the impact of large-scale changes to the built environment. They provide valuable data that might not otherwise be possible to gather, because of the costs associated with modification of the environment. However, factors beyond the control of the study team contribute to the complexity of both organising and conducting natural experiments, with potential flow-on effects to the quality of data. Therefore many extraneous factors need to be considered when designing, costing and conducting natural experiments; studies should identify opportunities to include key partners from the inception of the project, be flexible yet robust, and allow sufficient funding to accommodate unexpected changes in the research protocol.