Empirical studies have shown substantial accuracy differences between commercial forecasting tools that use the same types of models. These differences are caused by several reasons including the design and development of algorithms of selection, also as the parameters are calculated and as models are optimized and initialized. The best way to evaluate a system of generation of forecasts, is modeling the past scenarios, generate results through the system and compare with the presented reality. This will allow you to better assess the reliability of a forecasting system can provide to your planning process. To implement an automated system forecasts, the user should be aware that an automatic algorithm sees their data as a series of numbers and select a model based on statistical parameters. It is likely that sometimes the glider has a very large experience with regard to the behaviour of the products and the market demand, which can generate apply adjustments to the forecasts presented in the system or even reject the projections presented by the system.
There are many cases in which the judgment of experts is definitely the best choice, distorting the results of an automatic system. Time series of work capturing patterns in historical data and extrapolating the model in the future. Time series methods are appropriate when the user has a reasonable amount of data and there is a continuity between the past and the future (regularity). They are very appropriate methods when we forecast in the short term. This is due to the assumption that indicates that the model and the trends will remain constant. There are many situations where the methods of time series may not be the best choice. In these scenarios are new products (due to its short history), the prognosis for products subject to permanent events (promotions and/or interruptions of business that generate constant volumes deviations) and prognosis of products whose behavior depends on many variables besides the time as prices, variables of the market, etc.