How a Base Stock Policy Using "Stale" Forecasts Provides Supply Chain Benefits

Abstract

Managers often engage in forecast updating with the expectation that forecast updating reduces expected shortage and inventory costs. One undesirable effect of forecast updating is that it may lead to the bullwhip effect, a phenomenon where the variability of demand increases as one moves up the supply chain. The bullwhip effect can be undesirable for the supplier because more volatile orders from the downstream stage can be very costly to the supplier. It can make it more difficult for the supplier to forecast demand, and it can lead to large fluctuations in supplier production levels from period to period. Using "stale" or old forecasts may sound foolish, but their judicious use in a two-stage supply chain can improve fulfillment from the upstream stage to the downstream stage and reduce the fluctuations in production levels. We study a two-stage supply chain where the demand process is nonstationary and both stages use an adaptive base stock policy. We propose a policy that uses old forecasts to set base stock levels at the downstream stage while using current forecasts to communicate upcoming orders from the downstream stage to the upstream stage. We study a decentralized supply chain setting, and we find that our policy can reduce the expected supply chain inventory and shortage costs and significantly reduce the fluctuations in production levels compared to that of using current information. We also study a cooperative supply chain setting and, surprisingly, we find in numerical examples that our proposed policy results in very small increases in the expected systemwide inventory and shortage costs compared to a systemwide optimal policy, while reducing the fluctuations in production levels.inventory control, forecasting, forecast revision, bullwhip effect, production smoothing

Information Sharing and Order Variability Control Under a Generalized Demand Model

“…The MMFE demand model described thus far generalizes many commonly used demand models in the literature, such as the i.i.d. normal demand model, the AR(1) model (Lee et al 2000, Raghunathan 2001), the IMA 0 1 1 model (Graves 1999, Miyaoka andHausman 2004), the general ARMA model (Gilbert 2005, Gaur et al 2005, the linear state-space model (Aviv 2003), and the advance demand information model (Gallego and Ozer 2001).…”

“…Aviv (2003) used a general time-series framework to study the benefits of information sharing in supply chain management. Miyaoka and Hausman (2004) showed that under the IMA 0 1 1 demand model, using "stale" forecasts can reduce the bullwhip effect and improve fulfillment from the supplier to the retailer.…”

“…The advance demand information model by Gallego and Ozer (2001) is essentially an MMFE model. Most time-series demand models can be interpreted as a special case of the MMFE model, such as the AR(1) model (Lee et al 2000, Raghunathan 2001), the IMA 0 1 1 model (Graves 1999, Miyaoka andHausman 2004), the general ARMA model (Gilbert 2005, Gaur et al 2005, and the linear state-space demand model (Aviv 2003). Other papers that make use of the MMFE demand model include Gullu (1996Gullu ( , 1997, Toktay and Wein (2001), Iida and Zipkin (2006), and Lu et al (2006).…”

Measuring endogenous supply chain volatility: Beyond the bullwhip effect

“…Several authors have investigated possible ways to minimize the bullwhip effect. One approach has focused on the relative value of past demand information (Cachon, 1999;Miyaoka and Hausman, 2004;Balakrishnan et al, 2004;Daganzo, 2004); others focused on reducing lead time variability (Chatfield et al, 2004;Kim et al, 2006). Yet other authors have investigated the broader cost implications of the bullwhip effect (Metters, 1997;Chen and Samroengraja, 2004).…”

On the replenishment policy when the market demand information is lagged

“…For example, Zhang (2005) and Miyaoka and Hausman (2004) assume that the order-up-to level is updated every time period by using old forecast information and quantify the impact of such nonoptimum forecasting on a supply chain cost. In their model, however, the latest market demand information is exploited when the order quantity is determined.…”