By Jonathan Thatcher, CSCP
Adventures in Demand Forecasting
On a quest to continuousimprovement
Reader F.S. writes, “My company receives different forecasts from our enterprise resources planning (ERP) system, the sales department, and suppliers—and none are very accurate. How can we improveour demand forecasting?”
Nothing can provide a perfect demand forecast. All tools contain a certain degree of error. However, reducing forecast error over time is possible if you learn from yournumbers and your people and apply continuous improvement methodologies. Begin by examining some pre-forecast considerations every organization should make. The Pareto principle. Assign all items to beforecast into one of three categories. The 20 percent of items that reflect 80 percent of demand are known as the A items, while those that comprise the rest of demand are the B and C items. As A itemshave the greatest impact, devote more effort to improving their forecast accuracy. Group versus individual forecasting. Forecast accuracy generally is higher for groups of items than for individualitems. Random forecast errors tend to cancel each other out when averaged across the group, and the same effect applies to other activities, including inventory pooling and insurance. Dependent versusindependent demand. Dependent-demand items need not be forecast; their demands can be calculated based on a bill of material or other known quantities. In contrast, independent-demand items do requiredemand forecasts. Typical items with independent demand are finished goods. Note that some items fall into both categories. While these tasks do not create a demand forecast on their own, they helpfocus on using a variety of demand forecast techniques. Consider also the following strategies, which involve both your numbers and your people. Intrinsic forecasting (moving averages). Moving...