Ecuador is worldwide considered as one of the main natural flower producers and exporters -being roses the most salient ones. Such a fact has naturally led the emergence of small and medium sized companies devoted to the production of quality roses in the Ecuadorian highlands, which intrinsically entails resource usage optimization. One of the first steps towards optimizing the use of resources is to forecast demand, since it enables a fair perspective of the future, in such a manner that the in-advance raw materials supply can be previewed against eventualities, resources usage can be properly planned, as well as the misuse can be avoided. Within this approach, the problem of forecasting the supply of roses was solved into two phases: the first phase consists of the macro-forecast of the total amount to be exported by the Ecuadorian flower sector by the year 2020, using multi-layer neural networks. In the second phase, the monthly demand for the main rose varieties offered by the study company was micro-forecasted by testing seven models. In addition, a Bayesian network model is designed, which takes into consideration macroeconomic aspects, the level of employability in Ecuador and weather-related aspects. This Bayesian network provided satisfactory results without the need for a large amount of historical data and at a lowcomputational cost.