scenarioForecasting as a need
For every company, the ability to forecast the future trend of numerical variables, such as sales or the demand for goods and services, has always represented a crucial and strategic element for planning and programming, in the financial, production, marketing and logistics areas. The effectiveness of the plans and strategies defined in the future depends on the goodness of the forecasts, thus determining the achievement of the expected results. Yet even today, this activity is often carried out with inadequate tools, if not manually, thus limiting the number of forecasts that can be carried out each year given the considerable commitment of time and resources required.
The reason for the lack of acceptance and diffusion of forecasting technologies in companies lies in the lack of integration with business systems and processes, but also and above all in the 'obscurity' of the proposed forecasts.
Forecasting algorithms generate numbers on which it is difficult to imagine a business strategy without proper justifications that make them understandable and critically analyzable.
systemsForecasting & Big Data
Management transactional data, data from sensor and plant networks, as well as data from the web, mobile devices and open data sources, the computational capacity to process it in ever shorter time frames, and parallel systems available in the cloud.
All of this is creating new possibilities for forecasting systems today, especially in the ability to explain, correct, and evaluate forecasts by taking advantage of the vast amounts of information available.
BEAM Mantis: Enterprise Forecasting
BEAM Mantis, an integrated system dedicated to eXplainable forecasting, is part of this context.
The system, born from the experience of forecasting in large supermarket chains, allows the parallel forecasting on thousands of time-series, organized in categories and hierarchies.
The forecast is decomposed by the system into separate components: seasonality, trend, impact of external explanatory elements and automatically validated on historical data.
Some of the key features of the system:
- Uses variables from internal or external sources, to evaluate their impact on the series to be forecasted and thus improve the forecasts by correcting them with new information
- It allows the user to understand the 'number' provided by the algorithm, through a rationale related to each component of the prediction.
- It realizes forecasts in a short time even on thousands of series and keeps it continuously updated thanks to parallel computing infrastructures.
- Make multiple predictions, testing several hypotheses simultaneously.
- Mantis is a Cloud Native application and exploits all the potential of the Cloud in terms of parallel processing and on-demand allocation of resources: the costs are related only to the actual computation required to the system.
- Mantis is easily integrated with enterprise systems, allowing for the automatic feeding of time series and the reacquisition of forecast results in the usual applications.