Benchmarking the Forecasting Process

As I mentioned earlier the success of a forecasting function depends on people, process, technology and resources. Among them, the process is the most important component that connects people to technology and to resources. Therefore, it has to be well thought out and well structured. What other companies are doing can help you to set up the right process, and improve it if it is already in place. DEFINING FORECASTING PROCESS The process lays out in detail: What kind of forecasts and level of detail are needed? How far ahead forecasts will be prepared? How often they will be updated and revised? What data/information will be used to prepare them, and where that will come from and how. Where the forecasting function will reside? Who will participate in the process? What kind of infrastructure and technology will be needed? Which forecasting software/ system and models will be used? Which forecasting philosophy will be followed - one-number forecast or multiple-number forecasts? Which forecasting approach will be followed - bottom up, top-down or somewhere in the middle? Do we need a monthly consensus meeting to build a consensus? Who will have the authority to override forecasts, if needed? How forecast accuracy/error will be measured and documented? Here I will describe how different companies are dealing with such issues. WHERE FORECASTING FUNCTION RESIDES It makes a difference where the forecasting function is placed, because each function has a bias of its own. Production people tend to link forecasts with their production capacity. If the choice were given between over- and under-estimation, they would prefer over-estimation. With over-estimation, they would have less headaches resulting from shortages. Salespeople prefer under-estimation particularly at a time when forecasts will be used for setting quotas. Finance people in general are conservative, but their mindset changes when they report to Wall Street. For marketing people, it all depends. If the advertising budget is tied to a forecast, they will prefer over-estimation. The higher the forecast number, the more money they will get for advertisement. Larry Lapide, VP and GM of Benchmarking Services at AMR Research, says that it does not matter much where the forecasting function is placed as long as it is not within Planning. If it is placed within Planning, there is a danger of plans becoming the forecasts. No matter where you place the forecasting function, there will be some bias. One way to eliminate bias is to have a separate forecasting department. At present, only 9.15% of the companies surveyed have a separate forecasting department. In many companies, different departments do their own forecasting. With that, they don't need to defend their numbers because they won't be accountable to anyone other then to themselves. They will be forecasting numbers that meet their objectives. But this kind of process can create chaos within an enterprise. Since each plan will be based on different forecast numbers, demand and supply will not match, causing excess inventories and/or lost opportunities. From the survey, it appears, that no particular department has a hold on the forecasting function. In fact, it is scattered all over. All the industries combined, 19.63% of companies have the forecasting function within Marketing and another -20.47% within Operations/Production, the rest is thinly scattered over other functions. (See Figure 1) The same is true within industries. There is no department where a large percent of companies have placed this function. The highest is in Pharmaceuticals where 38.30% of companies have placed this function within Marketing. Looking at the results of 2001 and 2002, one interesting thing emerges, that is, companies seem to be moving the forecasting function away from Finance and into Logistics. When all industries are combined, the percentage of companies having this function in Finance declined from 13. …