Demand planning takes the past sales data, such as invoiced sales quantities or sales revenue, and uses forecasts to update it for the future. To do this, it can use statistical forecasting techniques, such as the constant, trend, and seasonal models with exponential smoothing or linear regression.
The complexity and competitive nature of today's business environment requires organizations to consider many variables when developing a sales and operations plan:
- Multiple sources of demand plan data. For example, the manufacturer's forecast is based on a distributor's past sales and point of sales directly from the retailer.
Factors influencing demand. For example, the size of the sales force, advertising expenditures, price, promotions, and seasonality.
Demand plan data can be exchanged with sales organizations, customers, and suppliers.
Collaborative planning involves comparing your own forecast results with the ones of your customers. Composite forecasting involves combining several forecasting techniques to provide the forecast results.
Composite forecasting goes beyond the idea of pick-the-best and combines two or more different methods to get the final forecast for the short- and mid-term horizons.
It is possible to use a weighted average between two or more different forecast algorithms in one forecast model to define the best final forecast model.
Calculate Weighted Average Forecast: If you choose this method, you can assign weights to the forecasting algorithms you have added to the model. The system calculates the average forecast after multiplying each result by the weight assigned to the algorithm that was used for its calculation. Sliders make the relative weights of individual algorithms more transparent.
It is also possible to let the system run an automatic forecast execution. To do this, you have to create an application job and assign two or more different forecast models. In this case, forecast models must use different planning horizons in the configuration.
For example: The first model uses the single exponential smoothing and double exponential smoothing algorithms for the horizon between Week 1 and Week 12 with MAPE as forecast error measure, and the second model uses the weighted average algorithm for the horizon between Week 13 and Week 52 with MAPE as forecast error measure.