Increasing pressure to reduce costs and improve service levels has forced companies to refine where they position inventory throughout their supply chains and at what level. Traditional supply chains followed a �push� system, holding inventory at the point of consumption. Because this system relies heavily on long-range forecasts, many companies shifted to a �pull� system, holding no inventory at all. But that approach had its own problems. Eventually, companies started to employ a hybrid approach, creating the push-pull supply chain paradigm. These trends, and the software that supports them, are having a profound effect on the parcel industry.
 
Push Supply Chains
In a push supply chain, production and distribution decisions are based on long-term forecasts. The manufacturer uses orders received from the retailer�s warehouses to forecast demand. Obviously, it takes some time for a push-based supply chain to react to a changing marketplace. Long-range forecasts are essential to a push system. There are some basic principles of all forecasts. These include the following:
 
The forecast is always wrong � It is impossible to forecast any supply and demand exactly. You can come close, but you will never be completely accurate.
 
The longer the forecast horizon, the worse the forecast � It is difficult to match supply and demand, and it is even more difficult if one needs to predict customer demand for a long period of time, such as the next 12 to 18 months.
 
Data updates lead to forecast updates, and aggregate forecasts are more accurate � As the firm receives more demand data, it uses the data to update its forecast, and therefore update inventory levels, safety stock and order quantities. However, these data updates cause traditional inventory management to experience the bullwhip effect, a phenomenon common in push supply chains. The bullwhip effect describes the fact that often, variability of orders received from retailers and warehouses is much larger than variability in customer demand (hence, a �bullwhip�). The longer the lead time, the larger the increase in variability. The increase in variability leads to excessive inventory due to the need for large safety stock; large and more variable production batches; unacceptable service levels; and inability to manage resources effectively since it is not clear whether the firm needs to plan resources based on average demand or based on peak demand. The problems of the bullwhip effect led many companies to go to the other extreme: a pure pull supply chain.
 
Pull Supply Chains
In a pull supply chain, production and distribution are demand driven and are based on actual customer demand rather than forecast. In a pure pull system, the firm does not hold any inventory and only produces to order. These systems are very attractive since they allow the firm to eliminate inventory, reduce the bullwhip effect, increase service levels and generally react nimbly to a changing market.
 
Unfortunately, there are many industries in which it is very difficult to implement a pull supply chain strategy. For example, production lead times of furniture are too long, making it is impractical to react to demand information. Furthermore, in a pull strategy, it is frequently more difficult to take advantage of economies of scale since production and distribution decisions are in response to specific customer demand. Batch production or efficient transportation modes, such as truckloads, are hard to achieve.
 
These advantages and disadvantages of push and pull supply chains have led companies to look for a new supply chain strategy that takes advantage of the best of both worlds � enter a hybrid of the two systems: push-pull supply chain systems.
 
Push-Pull Supply Chains
In a push-pull strategy, some stages of the supply chain are operated in a push-based manner, while others are operated in a pull-based strategy. The push system is typically in the initial stages, and the pull system occurs in the later stages. The interface between the push-based stages and the pull-based stages becomes the push-pull boundary.
 
A PC manufacturer may build to stock, and thus make all production and distribution decisions based on forecast. This is a typical push system. On the other hand, in a push-pull strategy, the manufacturer builds to order. The component inventory is managed based on forecast, but final assembly is in response to a specific customer request. Push is the part of the manufacturer�s supply chain prior to assembly, while pull is the part of the supply chain that starts with assembly and is performed based on actual customer demand.
 
In this case, the manufacturer takes advantage of the last principle of all forecasts: aggregate forecasts are more accurate. Demand for a component is actually the sum total of all finished products that use that component, and finished goods demand is easier to predict than component demand. This, of course, leads to safety stock reduction. Dell Computers is an excellent example of the impact the push-pull system has on supply chain performance. Dell keeps an inventory of components and assembles only when there is an actual order.
 
Postponement, or delayed differentiation, in product design is another excellent example of a push-pull strategy. In postponement, the firm designs the product and the manufacturing process so that decisions about exactly what product is being produced can be delayed as close to the final assembly as possible. The manufacturing process starts by producing a generic or family product, which eventually is modified to a specific end product as ordered. The portion of the supply chain prior to product differentiation is typically a push-based strategy. That is, the generic product is built and shipped according to long-term forecasts. Since demand for the generic product is an aggregation of demand for all its corresponding end products, forecasts are more accurate, and thus, inventory levels are reduced. In contrast, customer demand for a customized end product has a high level of uncertainty, so product differentiation occurs only in response to individual demand. The portion of the supply chain starting from the time of differentiation (customization) is a pull-based supply chain.
 
In a push-pull system, the push part is applied to the portion of the supply chain where long-term forecasts have small uncertainty and variability and can be reasonably predicted. On the other hand, the pull part is applied to the portion of the supply chain where uncertainty and variability are high, and therefore, decisions are made only in response to realized demand.
 
Inventory is a necessary evil. A push-pull supply chain is designed on the premise that inventory is essential in an effective supply chain strategy, but careful long- and short-term planning can reduce it to the minimum necessary to meet customer demand.
 
Impact on Transportation and Fulfillment
The new push-pull supply chain paradigm introduces a major shift in fulfillment strategies. Shipments move from cases and bulk shipments to single items and smaller size shipments and from shipping to a small number of stores to serving thousands of geographically dispersed customers. This shift has also increased the importance and the complexity of reverse logistics. Parcel shipping becomes paramount.
 
Table 1 summarizes the impact of the push-pull paradigm on fulfillment strategies. The new developments in supply chain strategies are very good news for the parcel and LTL industries. Both the pull and the push-pull systems rely heavily on individual (e.g., parcel) shipments rather than bulk shipments. This is especially true in the Internet B2C area, where a new term has been coined: e-fulfillment.
 
Another impact of e-fulfillment on the transportation industry is the significant increase in reverse logistics. Indeed, in the B2C arena, e-fulfillment usually means that the supplier needs to handle many returns, each of which consists of a small shipment. Online retailers must build customer trust, and a generous returns policy is one way to do that. Parcel shipping is already set up to handle these returns, a major issue in B2C and, in many cases, in B2B commerce. This is a real challenge for the LTL industry, which traditionally has not been involved in door-to-door services like the parcel industry.
 
E-fulfillment logistics requires the ability to reverse the flow easily from B2C to C2B, serve globally dispersed customers and do it with a short lead time. Only parcel shipping can do all that. Indeed, the parcel industry has the additional advantage of an excellent information infrastructure that enables real-time tracking. Thus the future looks promising for the parcel shipping industry and, in particular, for those carriers and consolidators who work to modify their own systems in order to integrate them with their customers� supply chains.
 
Identifying the Appropriate Supply Chain Strategy
New software has recently been introduced that can optimize inventory positioning over the entire supply chain. This type of software can help companies significantly reduce inventory costs while maintaining and sometimes significantly decreasing service times to the customers. The software can:
 
�           Identify the push-pull boundary; that is, identify supply chain stages that should operate in a make-to-stock fashion and distinguish them from those operating in a make-to-order fashion. Such information can push inventory to less costly locations in the supply chain.
 
�           Take advantage of the risk pooling concept. Demand for a component used by a number of finished products has smaller variability and uncertainty than that of the finished goods.
 
�           Replace traditional sequential supply chain strategies with globally optimized ones. In a sequential (or local) strategy, each stage tries to optimize its profit with very little regard for the impact of its decisions on other stages in the same supply chain. Being totally non-political, the software considers the entire supply chain and identifies strategies for each stage that will maximize supply chain performance.
 
The impact of push-pull strategy on reducing inventory costs and increasing service levels is best illustrated in Table 2. The line named �current� represents the traditional trade-off between committed service to customers and the cost of inventory required to support that commitment. By positioning inventory optimally across the supply chain, the firm is able to shift the tradeoff curve. (See the curve named �optimized.�) Locating the push-pull boundary at the optimal point in production reduces inventory cost. Similarly, optimally positioning inventory across the supply chain reduces response time to the customers without increasing inventory cost, which is a win-win situation for both the producer and the customer.
 
David Simchi-Levi is a professor of Engineering Systems at the Massachusetts Institute of Technology. He may be contacted by e-mail at dslevi@mit.edu. Edith Simchi-Levi is the vice president of Operations at Logictools, Inc. She may be contacted by e-mail at edith@logic-tools.com.