Finding a Cure for Forecasting Problems - Developing the Process
(
Page 2 of 3 )
Developing the Process
Our new forecasting process was developed over the course of a year. We first selected several company estimators, trade marketing experts and other professional staffers to form a team of 15 people who are now called forecasters. The team defined the forecasting model, which the IT staff then implemented in SAS. In the past, upward of 340 people had been involved at some level in forecasting sales. By limiting input to a group of people with specific skill sets, we were able to greatly improve our forecasts.
Data entry has been streamlined to prevent “multiple versions of the truth.” And having a smaller group of forecasting experts prevents confusion caused by too many staffers using different spreadsheets with differing estimates.
Our forecasters now work with monthly sales information to create forecasts three months out. The SAS forecasting tool works with Excel spreadsheets, and the navigation system is similar to a Pivot table, so it is familiar to users.
The process begins with estimators who work on forecasts for the following one to three months during the first nine to 12 days of the preceding month. The evaluation matrix is structured so that each estimator can find information for a specific group of customers with similar purchasing behaviors and food assortments. The estimators can look at the prior year’s data for that month or data from the past three months. All the data is linked to both total volume and promotional volume.
When the estimators finish this process, project leaders adjust and analyze the results with the help of trade marketing directors and sales directors. The validated results are sent directly to stock managers responsible for production planning. Our error rate is currently about 15 percent, down from 18 percent, and we want to reduce it to around 5 percent.
