Curbing deforestation in developing countries may be a cost-effective way to reduce carbon emissions and address climate change. But how to prevent the deforestation and degradation of forest lands in areas where landowners depend, for their livelihood, on slash-and-burn agriculture and extensive cattle ranching, two primary drivers of deforestation? In Brazil, the implementation of command-and-control measures, the expansion of protected areas, and interventions in the soy and beef supply chains, such as the Soy Moratorium established in 2006, have significantly curbed deforestation in the Amazon between 2005 and 2013. Despite this overall improvement, deforestation rates have continued at 5,000-7,000 km2 per year since 2009 and have increased again thereafter. Today, it is often argued that new mechanisms targeting small landowners are required to achieve further reductions in deforestation in the Amazon. Payment for Ecosystem Services (PES) contract may be an option but we need more evidence about which PES contracts work on the ground. In particular, we need to know which programs are most effective, in which context, and whether they work equally for all targeted participants. The objective of this experiment is to test the impact of different contracts of payments for environmental services (PES) on the deforestation decision of the Brazilian Amazonian cattle breeders. The experiment is based on the implementation of a pilot conservation program, in which participants are randomly offered a PES in exchange for not cutting their forest for an entire year. We randomly offer two types of contracts, which differ in their specifications, one being more flexible than the other. We evaluate whether one contract outperform the other, looking at forest loss, as measured by satellite imagery using SAR Sentinel-1 images. In each arm of the RCT, PES contracts is offered to participants following a Becker-DeGroot-Marschak (BDM) procedure. This design has many advantages: it is incentive compatible, since revealing the true price remains the best strategy for the participant; it allows to infer the causal effect of compensation payments on forest cover; and it allows estimating heterogeneous treatment effects.