|
Topic: KPIs |
pammy71
New Member
Total Posts: 1
Joined: May 2006
|
Posted: October 10 2007 4:17 AM
Hi
I need to set some KPIs for our forecasting team... I was interested to see what KPIS you work to in your area...
Thanks in advance
|
tmworthy
New Member
Total Posts: 3
Joined: February 2004
|
Posted: January 29 2008 1:46 PM
We look at sales forecast accuracy (SFA) measured by WMAPE at the following levels:
* Item-national (right product measure)
* Customer (promotion accuracy measure)
* Geography (right place measure)
We also look at forecast bias, to ensure we're not consistently over or under-calling the forecast.
|
mvgilliland
Member
Total Posts: 16
Joined: February 2004
|
Posted: March 19 2009 7:59 AM
tmworthy is correct to suggest using measures of both accuracy and bias, but be sure to measure down to the most relevant level of granularity. For example, if a manufacturer has only one distribution site, then Item/National is ok, but if you inventory and ship product from multiple distribution centers, then you'd want to know forecasting performance by Item/DC. (In the multiple DC situation, Item/National forecast accuracy is really an indication of bias. While it is great if your Item/National error is zero, that could be achieved by having terrible forecasts by Item/DC that just cancel each other out in the aggregation.
You should also measure Forecast Value Added (FVA), to make sure your forecasting process is doing better than a naive forecasting model such as a random walk. Forecasting performance should always be evaluated in terms of what a naive model would do. See this free recorded webcast http://www.sas.com/events/cm/176129/index.html and the accompanying white paper http://www.sas.com/apps/forms/index.jsp?id=wp&cid=6216 for details on how to do FVA analysis, and hear about companies that have used it.
|
GMCCDemand
New Member
Total Posts: 1
Joined: September 2009
|
Posted: June 23 2010 10:09 AM
I would like to know if others are using a BIAS measurement that is weighted by item (or other level of granularity). We are using a simple Fcst/Hist calculation (where 100% is ideal, and 100% + is overforecast and - 100% is underforecast). However, this is only done at total level (i.e. all items within a PNL/Merch group). I am looking for a calculation that will weight the items that comprise the total to give me an overall BIAS for the selected group.
Appreciated,
Glenn
|
Subscribe
Bookmark
Print
Email 
