Statistical Forecasting: what is it?
The importance of an effective tool
Statistical Forecasting is a set of analytical tools devoted to extracting available information in order to provide a forecast of future demand, as accurate as possible.
Forget traditional methods, we outperform them.
Your products are not all the same: our decision system detects different behaviours and use the more suitable forecasting model.
Dealing with uncertainty: statistical forecasting is a rigorous science, but in any forecast there is uncertainty. Our software provides you with the confidence intervals. What can be stated is that statistical forecasting provides the best possible guess of future demand given the available data.
How does it work?
Traditionally, statistical forecasting was based on extrapolating historical data from past sales. The statistical techniques employed in that context are already extremely sophisticated and can extract trends and multiple seasonality patterns or they can take into account special events. In recent years however the available tools have considerably expanded, thanks to machine learning and data mining.
Aggregate your forecasts!
Detailed forecast at the SKU level is usually what you look for: forecasting how many shoes of that kind, size, color you will be able to sell in a specific region. Our sales forecasting software provides you with aggregated forecasts at any level giving you a more understandable picture of future trends, differently from SKU-level forecasts which inevitably show an up-and-down pattern.
(*) We tested our predictive models on a dataset containing more than 5k time series of various behaviour and compared them to some standard naive forecasting methods with respect to common error metrics. We computed the perfomance in terms of winning cases percentage averaged on all considered error metrics. We can also state that our methods are outperformed by standard methods only when time series have few historical data and a very high percentage of zero values.