In most cases, in genetic breeding of Arabica coffee, plot size is defined in an empirical manner. It is often based only on the experience of the breeders and the availability of resources, potentially leading to a reduction in precision. Therefore, the aim of this study was to estimate the size of the experimental plot for evaluation of coffee yield. We evaluated two experiments for validation of cultivars with 12 treatments set up in a randomized complete block design with three replicates and plots composed of 50 plants. Each plant was considered as a basic unit. Estimates of ideal plot size were made by maximum curvature of the coefficient of variation, linear-plateau segmented model and by the resampling methods. We discussed the variations in the parameter estimates for different plot sizes. Divergence was seen among the plot sizes estimated by the different methodologies. Increasing the number of plants per plot led to a higher experimental precision to the point that the increase was no longer significant. The plot size recommended for evaluating coffee production is from seven to 19 plants.