The collaborative nature of sales and operations planning (S&OP) begs the question: Do we need a statistical forecast? Is the “best-fit” engine a dinosaur that ought to be relegated to the past? Many companies ask these questions as if there is just one answer—yes or no. Progressive companies understand that statistical forecasts add value if used in the right way.
Statistical forecasts serve as a starting point based upon non-biased historical patterns. Salespeople are great at adjusting forecasts based upon their market and customer knowledge. Asking a salesperson to create a forecast without a starting value builds in bias from the start and often creates frustration for the salesperson who wants to devote time to pursuing leads rather than creating sales projections. In an effective S&OP process, collaboration builds on the statistical foundation.
But not all statistical forecasts produce a great starting point. Stable historical demand patterns typically produce more reliable statistical forecasts. Conversely, highly volatile demand patterns may produce unreliable statistical forecasts. The key is to identify which will add value. Products can be grouped as being good or bad candidates for statistical forecasting based on volatility and importance. Planning strategies can be assigned based upon segmentation using volatility and importance criteria. Good candidates are statistically forecast while bad candidates use planning approaches such as like item curve fit modeling for new products and reorder point for low volume or erratic demand items.
Statistical forecasting still has a place. To learn more about when and how you should use statistical forecasting, check out our new videos:
How is your company using statistical forecasting? Let us know in the comments section.