Benchmarking Forecasting Practices in America

It is a common practice to benchmark a company against the industry practices to see how it's doing. If your company is performing below the industry norms, then the mission is clear. You have to do something to bring up your company to the industry level. If you are at par or above the industry norms, then you may like to raise the bar and do something to improve further the performance of your company. This is our fifth update on the forecasting practices in corporate America. Ever since we started, which was 2002, the response from our readers has been overwhelming, so we decided to continue this practice. Benchmarks are merely averagesaverage forecast error, average salary in the forecasting profession, background of an average forecaster, etc. Benchmarks given in this issue are based on the survey conducted by the Institute of Business Forecasting (IBF) at all our forecasting conferences and tutorials held in 2006 in the United States, which numbered five in total. Since the participants attended forecasting conferences and tutorials, they are either forecasters or plan to become ones. Most of the participants come from large companies. In fact, 65% of them came from companies with sales revenue of $500 million and over. It is difficult to cover all the benchmarks in one issue. What we have covered here are the ones that are most pressing - the benchmarks of the forecasting process, forecasting models, forecasting error, new product forecasting practices, forecasting software and systems, salary of forecasters, and background of forecasters. FORECASTING PROCESS The forecasting process plays an important role in forecasting, though forecasting models-not the process-is what matters most to many academic forecasters. All the forecasting books available in the market talk about models with little or no discussion about the process. It appears that the key to improving forecasts is to find the right model. However, the process is as important, if not more, as using the right model for best results. The process deals with, among other things, how forecasts are prepared and performance is measured and monitored. Take the example of Nike, which in 2000 implemented the i2 Technology forecasting expert system to prepare forecasts. (The expert system is the one that automatically tests different models, and then selects and prepares forecasts with the best one.) Nine months later, Nike declared that it lost $400 million because of inaccurate forecasts. It over-forecasted certain shoes and underforecasted others. Whom would you blame for that-forecasting software or forecasting process or both? Maybe both. Maybe the expert system was not up to the mark and thus could not select the right model. Maybe process was not in place to monitor forecasts. If a forecasting software under- or over-forecasted a given product or made a huge error two to three months in a row, someone would have noticed it had there been a process to monitor them. Since the problem was not recognized until it was too late and no action was taken to correct it, one would conclude that the process was not in place. The forecasting software was used as a black box. The forecasting process helps in two important ways. One, it is well recognized that statistical forecasts are nothing more than baseline forecasts. A judgmental overlay tends to improve the quality of statistical forecasts. It is a common practice inmany companies for a forecaster to prepare forecasts and then present them in a consensus meeting, which is attended by all the functions involved, including Sales, Marketing, Finance, and Production. The committee members review the forecasts and, if necessary, overlay judgment on them, which, by and large improves the forecasts. Two, the process forces the forecaster to regularly monitor forecast performances. If a model selected consistently yields large errors or consistently over- or under-forecasts, it means something is wrong somewhere which needs to be investigated and corrected. …