In Forecasting, Part I, I pointed out that accurate forecasting is a complex and challenging problem but that it is generally still possible to create good forecasts through the proper combination of judgmental and statistical methodologies. Specifically, manually adjusted statistical forecasts by an expert who has “inside” information, is aware of “one-time” events, and / or who is responsive to the latest environmental changes can often (dramatically) improve forecast accuracy, provided human bias does not creep in. (Thus, only practitioners with domain knowledge should adjust statistical forecasts using a structured process and only do so when there are known changes in the environment that the statistical model wasn’t really built to handle.)
Then in Forecasting, Part II, I pointed out an article in Purchasing that noted that when it comes to commodity forecasting, judgmental forecasts by experts have the best accuracy on record, demonstrating that expert human judgment applied to good statistical models with solid historical data that also take into account market intelligence and global economic trends are the way to go.
Now I am going to draw your attention to a recent white paper, sponsored by Supply Chain Consultants, by Tom Wallace and Bob Stahl titled Forecast Less and Get Better Results that demonstrates that the conventional wisdom that companies need to project forecasts and plans far into the future at a highly granular level is not necessarily right. Specifically, it points out that detailed forecasts and plans are normally only needed inside of what the authors call the Planning Time Fence or the point in the future when the cumulative lead time to acquire the material and build the product is only a short time away. They argue that outside of this planning time fence, you should only be concerned with aggregate volumes.
Specifically, they argue that up until it’s time to plan a production run, you should only be concerned about forecasting the aggregate volumes required for raw materials beyond the average planning time fence. After all, if you’re a large fast food chain, chances are you can predict with a fairly high degree of accuracy how many burger patties you are going to need over the next year, even if you can’t predict exactly how many Big Burgers, Bob Burgers, or Bo Burgers you are going to sell in any specific week. And if you are a toy manufacturer, chances are you can predict roughly how much plastic you are going to require over the next quarter, even if you don’t know precisely how many units of Dolly House or Trixie Truck you are going to be asked to manufacture. Attempting to forecast to a granular level too far in advance will just mean you’ll constantly be revising your forecasts and wasting time and resources, instead of focussing on what’s truly important for sourcing – the raw material aggregate volume, since that’s where your leverage is.
I’m sorry, but “I’m not omnipotent” just doesn’t cut it anymore!