Using a “Value Added” Approach in Supply Chain Forecasting
By IDB Guest Blogger: Michael Gilliland, SAS
Materiel and supplies are critical to the men and women of the U.S. Armed Forces, at home and abroad. Whether it’s ammunition, weapons, IT, or toilet paper, if they need it, someone is in charge of getting it and making sure there is sufficient inventory. The Government Accountability Office has identified this mission-critical role – supply chain management – as an area for improvement for the U.S. Department of Defense (DoD), particularly inventory management through better forecasting. As DoD works to improve its supply chain forecasting capabilities, this is an area where DoD can learn from the mistakes of the private sector.
The reality is that forecasting can be a huge waste of management time. This is not an indictment of the practice of forecasting as a whole, but rather of how organizations usually approach and apply forecasting incorrectly.
The problem is not that forecasting is pointless, irrelevant, or unnecessary. Rather, the problem is supply chain leaders squandering too much time and too many resources on forecasting with a myriad of bad practices.
Tools and Methods
In some cases, the technologies organizations use for demand planning and forecasting are outdated or simply misapplied. By relying on outdated tools or methodologies, these forecasters miss the progress made in recent years to improve accuracy, reduce bias, and minimize the cost of forecasting through large-scale automation. Even heroic efforts on their part are likely to deliver underwhelming results, whereas an unpoliticized and unbiased forecast can lead to cost savings and, more important, can save lives.
The goal of forecasting is to obtain an objective, dispassionate number that is as accurate as can reasonably be expected given the nature of whatever you happen to be forecasting. Rather than working from this perspective, however, many managers and forecasters have unrealistic expectations for the level of accuracy achievable. They rely too heavily on the current “fit” of models to history when their job is to forecast the future. Almost invariably, the forecast will be less (often much less) accurate than the fit to history. (It is always much easier to explain the past than to predict the future.)
Don’t Trust the Process
Perhaps the most blatant example of waste in the forecasting process is “forecasting by committee.” This is where a forecast is passed through so many different stages of approval and has been tweaked by so many collaborators that its integrity is actually degraded. It is easy to see how this approach could occur in any bureaucratic environment – including the military – where each participant has a personal agenda they express with their adjustment.
The problem is that this kind of elaborate review process ends up being extremely costly – in two ways. First, because the process is unnecessarily consuming everyone’s time. And second, because your outfit may actually be in a worse position than if you had not attempted to incorporate so much “management intelligence” into the forecast in the first place. In a study of eight commercial supply chain companies, Steve Morlidge (author of the book Future Ready) found less than half their forecasts were more accurate than the “naïve forecast” – i.e., the forecast you get by doing nothing and simply using the last available data point as your future prediction.
Know Your Limits
Even if you use the proper tools and methods to forecast future supply chain needs, it is important to realize that overall forecastability still limits the maximum possible accuracy of forecasts. Supply chain leaders need to avoid demanding a level of forecast accuracy that is simply impossible to obtain because they have not considered the nature of what they are trying to forecast.
To illustrate this concept, suppose your job is to forecast Heads or Tails each day in the tossing of a fair coin. While you may have some lucky streaks and forecast correctly several days in a row, over the long haul your forecast will be correct just 50% of the time. It doesn’t matter if your ranking officer demands 60 percent accuracy or higher – you are limited to a 50 percent accuracy ceiling by the nature of the behavior you are trying to forecast. Not even bigger computers and more sophisticated software will help – there is nothing anyone can do to achieve 60% accuracy. Unachievable objectives motivate forecasters to simply give up, or find a way to cheat and game the system.
As seen in this scatterplot of 5000 items being forecast by a consumer goods manufacturer, the variability or “volatility” of a demand pattern has a big impact on how well we can expect to forecast it. Smooth, stable, repeating patterns can be forecast quite accurately with simple methods. But wild, volatile, erratic patterns may never be forecast accurately, no matter how many resources we commit to forecast them. To the extent that we can control and limit volatility, we are likely to achieve more accurate forecasts.
Solution: Knowledge is Power
Private sector companies are increasingly using a method called “Forecast Value Added (FVA) Analysis” to improve performance of their forecasting processes. FVA is the application of basic science to evaluate a forecasting process – measuring each step in the process, and determining whether it is “adding value” by making the forecast more accurate and less biased. Here is an example of an FVA report (in this case, the Analyst Override step is just making the forecast worse!):
Forecast Value Added analysis is about rooting out the waste and inefficiency from forecasting efforts (so is consistent with a lean approach to supply chain management). It allows organizations to streamline their process and redirect the non-value-adding efforts into more productive activities that will be more beneficial to the mission at hand. Understanding cautionary tales of corporate sector mistakes and challenges can help the DoD create real improvements in defense supply chain management. By avoiding wasteful steps and procedures, defense organizations have the opportunity to achieve better forecasts with less effort, and less cost.
|Michael Gilliland is an IDB guest blogger with more than 15 years of forecasting and supply chain management experience in the food, apparel, and consumer electronics industries. Gilliland has been featured in the LOGTECH Advanced Program in Logistics and Technology. He is author of The Business Forecasting Deal, and has published articles in Supply Chain Management Review, Foresight: The International Journal of Applied Forecasting, Journal of Business Forecasting, Analytics, and APICS magazine. He is also an editor for Foresight. Gilliland holds a BA in philosophy from Michigan State University and master’s degrees in philosophy and mathematical sciences from Johns Hopkins University. He writes The Business Forecasting Deal Blog at blogs.sas.com/content/forecasting. For more than a decade, SAS and the IDB have worked collaboratively to raise the level of awareness of the value analytics to defense leaders.|