Carbon offsets from the REDD+ (reducing emissions from deforestation and degradation) framework to protect forests are expected to see a 100-fold increase in market value by 2050. However, independent causal impact evaluations are scarce and only a few studies assess benefits to communities themselves, a core objective of REDD+. Following a pre-analysis plan, we use a before-after-control-intervention (BACI) framework to evaluate the impact of a large-scale voluntary REDD+ project in Sierra Leone—the Gola project. We use a panel of both satellite images and household surveys to provide causal evidence of the impact of the project on local deforestation rates and socioeconomic indicators over the first 5 yr of its implementation. We find that REDD+ slowed deforestation by 30% relative to control communities while not changing economic wellbeing and conservation attitudes. We find suggestive evidence that the programme increased the value of alternative income sources, by shifting labour away from forest-dependent farming activities. A cost-to-carbon calculation shows that REDD+ led to 340,000 tCO2 in avoided emissions per year, with an estimated cost of US$1.12 per averted tCO2. Our study contributes to developing an evidence base for voluntary REDD+ projects and offers a robust approach to carry out BACI assessments.