Answers To Your Forecasting Questions

ANSWERS TO YOUR FORECASTING QUESTIONS Q. Regarding the “Benchmarking Forecasting Errors” published in the Winter 0607 issue of the Journal, are they goodnessof- fi t (model fit) errors or did they come from the holdout periods? If it is not confidential, may I ask about the number of companies included in the analysis? Also, is it possible to see errors at other levels, if available—something with time dimension of week and month as well as category? A. The errors we publish are the errors they actually experienced. They are not errors of holding periods as is defined in textbooks, but they are equivalent. The number of observations varies from question to question. For example, the number of people who responded to the forecast error of one-month ahead in 2006 is 98; 58 people responded to the forecast error of one-quarter ahead. In every IBF conference/tutorial, we ask attendees to respond to a number of questions. Everyone does not respond to every question. We have the forecast errors by one-month ahead, two-months ahead, one- quarter ahead, and one-year ahead. Also, the errors we publish are broken down by SKU, Category, and Aggregate. We don’t publish errors ...