In collaboration with two communities living in, and on the edge of, Gorongosa National Park (GNP), Mozambique, we researched the importance of different landscape units to these communities and used the information to develop a management plan for GNP. We conceived the importance of a landscape to local people as a ratio of the benefits they derive from it and the costs of accessing or using those benefits. To test this expectation, we developed Bayesian belief models, for which the parameters were the relative preference weightings derived from community members (the relative preferences for benefits and relative expectations of costs). We then collected field data to confront the models for each of the two sites. In a parallel process, we conducted a vegetation survey to generate a map of the vegetation types, as well as an index of biodiversity importance for each vegetation type of the two 20-km2 sites. For each site, we simplified and converted the benefit:cost model into a local community importance surface, or map, and then overlaid a conservation importance surface on it in order to identify locations that were of high importance to both conservation groups and the local community. Such areas would require careful management attention. This paper discusses the implications of the research for the planning of GNP, as well as the strengths and weaknesses of the approach.