Reengineering the Forecasting Process with Predictive Models
Forecasts usually start as historical performance which is then reviewed and adjusted based upon anticipated events. More mature forecasting may also incorporate generally accepted business rules “programmed in” to help “drive” the data into a forecast.
Over the years I’ve deployed uncountable applications which provide the ability to routinely generate “adjustable” forecast scenarios seeded from source system actuals and driven by business rules.
The way these systems work is:
- Establish historic performance for the forecast period as a base line (or starting point)
- If there is a business rule than can be used to effect or “drive” historical performance into a more accurate forecast, apply it.
- If the analyst has specific knowledge that will affect performance in the period, the forecast is adjusted.
- Review prior forecast results – the variance between the forecast and the actual performance – and set future objectives.
Predictive Forecasting
Forecasting applications now utilize predictive modeling to improve this process. These systems use a predictive model generated using accepted statistical disciplines and scored against historical performance.
Predictive Analytics offers the ability to generate very accurate forecasting models that can be automated to continually absorb in real time new performance data as it becomes available as well as “adjust itself” based upon the results of prior periods forecasting accuracy.
Case Studies are beginning to emerge reporting the latest success stories of organizations that have adopted this discipline. Tools in the IBM suite such as Cognos TM1 and SPSS Statistics offer a clear advantage when designing, developing and deploying these types of systems.
Clear Opportunity
This is a clear opportunity to reengineer your exiting forecasting process. The possibilities are endless!
