The exactitude of the prognoses generated by systems automatic it can vary obviously. Aplicativos developed in leaves of calculation of Excel, rarely offer high exactitude, while yes they do the systems of prognosis specialized that use methods of proven series of time statistically, which use models like exponential smoothing, Box-Jenkins, Croston and others. Empirical studies have shown substantial differences of the exactitude between commercial tools of prognosis that use the same types of models. These differences are caused by several reasons including the design and development for the selection algorithms, also by as the parameters calculate and as the models are optimized and initialized. The best form to evaluate a system of generation of prognoses, is to model the passed scenes, to generate the results through system and of comparing them with the presented/displayed reality.

This will allow him to evaluate of better way the trustworthiness than a system of prognoses can offer its process of planning. When implementing an automated system of prognoses, the user must have present that an automatic algorithm sees its data as a series of numbers and selects a model with base to statistical parameters. It is probable that sometimes the glider has a very great experience with respect to the behavior of the demand of products and the market, which can generate applying adjustments to the prognoses presented/displayed in the system or to even reject the projections presented/displayed by the system. There are many cases in which definitively the judgment of experts is the best option, weakening the results of an automatic system. The series of time work catching landlords within the historical data and extrapolating the model in the future. The methods of series of time are appropriate when the user has a reasonable amount of data and is a continuity enters the past and the future (regularity). They are very appropriate methods when we foretold in the short term.