Nearly 10% of the world's total forest area is formally owned by communities and indigenous groups, yet knowledge of the effects of decentralized forest management approaches on conservation (and livelihood) impacts remains elusive. In this paper, the conservation impact of decentralized forest management on two forests in Tanzania was evaluated using a mixed method approach. Current forest condition, forest increment and forest use patterns were assessed through forest inventories, and changes in forest disturbance levels before and after the implementation of decentralized forest management were assessed on the basis of analyses of Landsat images. This biophysical evidence was then linked to changes in actual management practices, assessed through records, interviews and participatory observations, to provide a measure of the conservation impact of the policy change. Both forests in the study were found to be in good condition, and extraction was lower than overall forest increment. Divergent changes in forest disturbance levels were in evidence following the implementation of decentralized forest management. The evidence from records, interviews and participatory observations indicated that decentralized management had led to increased control of forest use and the observed divergence in forest disturbance levels appeared to be linked to differences in the way that village-level forest managers prioritized conservation objectives and forest-based livelihood strategies. The study illustrates that a mixed methods approach comprises a valid and promising way to evaluate impacts of conservation policies, even in the absence of control sites. By carefully linking policy outcomes to policy outputs, such an approach not only identifies whether such policies work as intended, but also potential mechanisms.