پیش بینی روند برای ثبات در زنجیره تامین

This paper revisits the use of trend forecasting to determine ordering policy in supply chains by viewing it as a part of the control process for making the supply responsive to demand. Trend forecasting is often used to assess demand — a tracked variable in the control context, which drives supply — a tracking variable. Used in this way, it is often observed to increase instability creating the so-called bullwhip effect. Trend is used on the other hand with reliability to increase stability in controller control, but with the difference that a trend of a tracking variable is used to drive correction. While both processes involve use of trend to determine policies for achieving reliable performance, the outcomes of the former are variable while those of the later can create improvement in control with certainty. The similarities and differences between the two processes are discussed and guidelines developed for applying trend forecasting to enhance stability in supply chains.

Demand forecasting has become an integral part of supply chain management and many sophisticated tools are now available to project demand either directly or indirectly through projecting its sources (Wheelwright and Makridakis, 1985). In fact, demand forecasting has come to be used as a matter of routine rather than being focused on addressing any specific problems of supply chains it is applied to (Gung et al., 2002). Use of trends is an important part of demand forecasting and many tools and methods have been suggested to improve the accuracy of demand forecasts (Bermudez et al., 2006). Albeit, trend forecasting as a discipline has often been viewed with reservation as it ignores its own impact on systems whose structure it often does not recognize, which creates much variability in its performance (Van Vught, 1987). Use of demand forecasts in driving supply chains is also observed to increase what has come to be known as bullwhip effect, which manifests in amplification of inventory cycles as one moves farther from demand in the supply chain. While it is widely recognized that distortions of information created by this effect can lead to tremendous inefficiencies in terms of inventory investment, poor customer service, lost revenues and misguided capacity planning (Lee et al., 1997), a clear guideline for intelligent use of forecasting does not exist and it is often used as an end rather than as a means with the expectation to improve the performance of the supply chain. Indeed, the importance of understanding complex production processes is quite widely advocated for effectively managing supply chains (Lee et al., 1997 and Khurana, 1999), but how these are influenced by the widespread use of forecasting is not well understood and, driven by widespread software support, trend forecasting continues to be used without focus on problem solving. Taking my inspiration from the use of trend in servomechanisms to create what is called derivative control, I propose in this paper that trend forecasting can be directed to reinforcing control processes also in a supply chain that should reduce instability. To accomplish this, however, the existing control processes in the supply chain must be identified and the trend of an appropriate internal corporate variable used to create derivative control to supplement them. System dynamics modeling is used to construct the control systems I experiment with pertaining both to servomechanisms and supply chains.

There seem to be similarities in policies created by use of derivative control and forecasting. Both use trend information about the variable in the system to drive an error correction process. However, while the former explicitly creates a control process driven by the trend of error in a tracking stock, the latter may often use trends in tracked variables to create remedial processes with often dysfunctional consequences, since their underlying feedback structure may counter their intended goals. Using complex forecasts of many tracked variables may further complicate the process creating further unforeseen consequences, although demand forecasting software encourages this. It is also observed that while derivative control is cognizant of the feedback process it creates, forecasting in social systems may accidentally create dysfunctional feedback structure since forecasting is done without any intent to create a control process. There is, however, no reason why forecast-related information cannot be used to improve performance of a policy since derivative control can use similar information to improve performance with great reliability. To accomplish this, we need to carefully identify the structure of the policy, the feedback loops it creates and the variables forecast. Tracking variables seem to be better candidates for forecasting than the widely-used tracked variables. Complex forecasts involving many variables might only increase the uncertainly of the outcomes. Derivative control in engineering seems to offer a good model for designing policies using trend information. Using this model, we can use forecasting reliably in improving system performance. Further research should aim at understanding control processes in complex supply chains involving cascaded delays and attempting to create the variety of control processes using the types of control widely used in servomechanisms. Other contexts in which the control model can be applied include financial and national planning and environmental remediation.