How Risky is our Data?
Hey, forecasters! From now on, every month we’re going to bring you powerful techniques to bring your statistical knowledge and forecasting process to the next level.
Now the fact is, not every data set is statistically forecastable. As forecasters, we need to be able to figure out what items forecast and what items don’t, and then determine whether those unforecastable items are important or not. We can get a summary of this with John Galt’s statistics. Most forecasters use MAPE and R-Squared for this, which do a great job of telling us how accurate our forecast line might be. However, they don’t really tell us how variable our data is – in other words, how costly an error might end up being. To get a clear picture of this, we need a third statistic:Coefficient of Variation (COV). You can turn on COV from the Statistics tab of the ForecastX Wizard under Analysis.
COV is a very simple calculation: Standard Deviation/Mean. The chart below shows the calculation at work – 37 divided by 49 comes to 0.75.
What does this mean for you? Basically, a lower COV makes it harder to miss on the forecast by any significant amount. Even if your forecasting method has almost no explanatory value, a data set without a lot of variation will look like it has a reasonable MAPE, and it will be easier to cover the variation with safety stock. If you have a higher COV and your forecasting methods aren’t working, your only recourse is to increase inventory levels or reduce customer service.
If this sort of problem has been on your mind, John Galt’s forecastability analysis service can help you get to the bottom of your data to figure out how big your target is. From there, you’ll be able to spend your time on items that need your help and make a bottom-line impact. Call us at 312-701-9026 to learn more!
Tradeshow season is here!
All you knowledge-hungry planners are likely already planning your trips to this year’s fall tradeshows. John Galt will be at both the IBF conference, Oct 17-18 in San Francisco, CA, and the APICS Conference & Expo, Oct 23-24 in Pittsburgh. Call us to arrange a meeting – we’ll be happy to talk through your planning challenges, show you the latest versions of Atlas Planning or ForecastX Wizard, or just answer your general questions about forecasting. 312-701-9026 is the number – come see us!
Schedule your upgrade today
» Learn more
» Learn more
» Learn more
Webinar featuring Noha Tohamy of AMR Research
» Request presentation
|Tel: (312) 701-9026
Fax: (312) 701-9033