Talk:Calculating Demand Forecast Accuracy

From Wikipedia, the free encyclopedia

This page should be deleted as measures of forecast accuracy, not limited to demand forecasting, are provided (in common notation) under Forecast - http://en.wikipedia.org/wiki/Forecast. Also, MAPE is not considered the primary demand forecasting error by statisticians and/or practitioners, MAE (MAD) is the primary measure, see Sanders, N. R. (1997). Measuring forecast accuracy: Some practical suggestions. Production and Inventory Management Journal, 38(1), 43-46 and Makridakis, S., Wheelwright, S. C., & Hyndman, R. J. (1998). Forecasting: Methods and applications (3rd ed.). New Jersey: John Wiley & Sons. Peter Catt 02:07, 15 December 2006 (UTC)

I agree that the content of this page is just a duplicate of the http://en.wikipedia.org/wiki/Forecast page. --Joannes Vermorel 14:37, 14 November 2007 (UTC)

I respectfully disagree. The forecast page is a primer on all kinds of forecasting, where as demand forecast accuracy page resolves confusions on calculating forecast error as used by the supply chain practitioners. I know this for a fact due to the many queries received from business practitioners across the world. Should the Wiki be just the home of the academic elite?? Markchockal (talk) 05:59, 27 November 2007 (UTC)

[edit] MAPE discussion

There has been considerable confusion on the issue of forecast accuracy. While academics use MAD and MAPE, there is little practical relevance the way MAPE is used by academics in the industry. The discussion on this page is perhaps one of a few on the web that brings to light the criticisms leveled against the use of MAPE as aruged by the academics.

While the industry practitioners have seen the rise of the discipline of demand planning, there is little recognition of this discipline among academic forecasters. The wiki should have a fair discussion of both the academic as well as the real world concepts. Markchockal 17:21, 14 March 2007 (UTC)

Shouldn't the formula for "deviation of the forecasted quantity from the Actual" be (forecast minus actual) over actual? I am not a mathematician, just a practitioner. My formula makes sense to me. The one on this page does not. For my purposes, anyway. —Preceding unsigned comment added by Dbrvnk (talk • contribs) 18:25, 14 February 2008 (UTC)

When given the choice if dividing the error by either actual or forecast, should you divide by the forecast you are far better off to always over forecast by about 10%. This way, you are dividing by a larger number, resulting in a smaller error.

Personally I don't like this particular measurement, but if you have to use it, divide by actual.

A much better approach is to use standard deviation of forecast error. It doesn't create a nice number for business to understand, but it is clearly your goal to reduce this number.