There are many reasons why a new Statistical Forecasting Solution is implemented. Perhaps it is part of a suite of applications being deployed in an ERP upgrade, or maybe a specific legacy solution is being replaced to enhance planning system capabilities. Acquisitions or mergers can promote the usage of a superior tool across your business or maybe the usage of Excel has reached the stage where the size of files, collaboration and security concerns have become a critical risk to the business.
Although Information Technology frequently provides the opportunity for forecasting change, the impetus and choice of planning solution should be driven by the people who run the forecasting processes in conjunction with the goals that management need to achieve. Sophisticated tools that provide slicker integration, more accurate forecasts, exceptions management and superior reporting that will, in turn, provide improved demand signals thereby reducing inventory, increasing margin and delighting customers.
Of course, usually many of these factors are combined to create a powerful argument for changing from Legacy Systems and Spreadsheets to a shiny new statistical solution. However a fundamental aspect that can sometimes be overlooked is to determine if statistical forecasting is the right planning method. Should you be forecasting statistically? If your data & structures are not suitable for statistical generation the "previous forecast" method or last years actuals with a simple % change could be more effective and considerably less expensive to implement and maintain.
Key decision drivers are often IT related such as:
- Licensing Costs
- Software Version / Patching / Integration issues
- New ERP Source
- New Planning Suite
or people and process related such as:
- User Maturity growth
- Cycle Time reduction
- S&OP / IBP implementation
or business imperatives derived from accuracy improvement such as:
- Inventory reduction
- Revenue and Margin growth
- Customer satisfaction
- Competitor competition
Use Segmentation analysis to determine how appropriate your data, dimensions & hierarchies are for statistical forecasting. Analyse the data that will be used to generate the statistical forecast and identify the proportions of data that are Smooth, Intermittent, Lumpy and Chaotic. Lumpy and chaotic data will generate forecasts of dubious quality and will require effort to validate and override. What proportion of your data is suitable for statistical forecasting?