Resumo. Studies on sampling techniques appropriate and accurate for measuring the yield of native species of Cerrado are still very rare, constituting a gap in the formulation of Sustainable Management Plans for the Biome. This study was conducted in an area of cerrado sensu stricto, located in the Experimental Ecological Reserve of the University of Brasilia, Federal District, with the objective of evaluate the efficiency of randomized branch sampling to quantify the yield of fruits of Caryocar brasiliense Camb. (Pequi). After a floristic inventory in the area of cerrado sensu stricto, it was selected one tree of Caryocar brasiliense in the fruiting period for application of randomized branch sampling. The production of fruits of Pequi was estimated by randomized branch sampling, tested by two techniques of selection of branches: uniform probability and probability proportional to the diameter of the branch. The validation of the method was made by comparison with the actual production, defined by the total count of fruits. The sampling error for a sampling intensity of ten paths was equal to 99.92%, using the technique of uniform probability and equal to 61.45%, using the technique of probability proportional to the diameter of the branch. It was concluded that the randomized branch sampling, using the techniques of uniform probability and probability proportional to its diameter, was imprecise and inaccurate for the study case. For this reason, it is necessary to carry out further studies and some modification on the technique that allows generate estimates of the fruit production of Caryocar brasiliense with small errors and without tendency.