Reducing Emissions from Deforestation and Forest Degradation (REDD+) has been piloted in several developing countries. Limited funding available for REDD+ suggests that there is a need to adopt a targeted approach (i.e., targeting selected groups or regions) to make REDD+ projects more effective. However, there is no clear understanding of how targeting could be done based on households’ preferences for various design features of a REDD+ policy. Using choice experiment data obtained from two groups of households (project participants and outsiders) belonging to three types of forest management regimes (private, government and community) in Indonesia, this paper aims to identify classes of households that have similar preferences towards REDD+ design features. The scale adjusted latent class analysis indicates that there are four classes of households: (1) supporters emphasising household benefits, (2) supporters emphasising community projects, (3) indifferent but objecting restrictions, and (4) sceptics demanding monetary benefits. We also found that forest management regime is a key determinant of household classes. Our results suggest that REDD+ projects are likely to be more accepted by households in the community- and government-managed forest regimes. Such information will be useful to develop more targeted REDD+ projects.