Talk:Forecasting

Archives (Index)

2012 2013 2017

This page is archived by ClueBot III.

Proposal to include "Cost of Forecast Error" in Forecasting Accuracy section

Hello editors,

I see the "Forecasting accuracy" section currently lists standard statistical metrics (MAE, MSE, RMSE). I would like to suggest expanding this to include metrics that address the financial value of accuracy, which is a critical component in applied business forecasting.

In many commercial contexts (supply chain, retail), forecast error is asymmetric, the cost of "under-forecasting" (stockouts) is often different from "over-forecasting" (holding costs). Therefore, a model with the lowest statistical error (MAPE) is not always the most profitable model.

I am an author of two peer-reviewed papers that propose and validate the Cost of Forecast Error (CFE) as a complementary metric.

Proposed Sources:

1. Catt, P. (2007). "Assessing the Cost of Forecast Error (CFE)." Foresight: The International Journal of Applied Forecasting, Issue 7, pp. 5-10.
2. Catt, P., Barbour, R.H., & Robb, D.J. (2008). "Assessing forecast model performance in an ERP environment." Industrial Management & Data Systems, 108(5), pp. 677-698.

Suggested Text for Consideration: (Ideally placed after the list of standard error metrics)

"In supply chain and business applications, statistical accuracy does not always equate to financial performance. Because the cost of errors is often asymmetric (e.g., lost sales vs. inventory holding costs), researchers have proposed alternative metrics such as the Cost of Forecast Error (CFE). This approach evaluates models based on the financial risk they generate rather than raw statistical deviation (Catt, 2007; Catt, Barbour & Robb, 2008)."

COI Declaration: I am Peter Catt, an author of the cited papers. I invite independent editors to review if this addition improves the article's coverage of applied forecasting.

Thank you. AsymmetricLoss (talk) 01:43, 10 December 2025 (UTC)