Bullwhip effect

Bullwhip effect was first observed by Procter and Gamble (P G) in the sale of baby diapers (Lee et al., 1997). Even though diaper sales at the retailers were stable over time, wholesale orders to factories surged up and down, swinging widely over time. Thus, there was very little imcertainty in the end-customer demand, but a very high uncertainty in the orders to factories and suppliers. Similar phenomenon was also observed by HP, apparel manufacturers, and the grocery industry. 

---

In 1958, Forrester started studies on an effect which is nowadays often referred to as the bullwhip effect. The bullwhip effect describes the amplification of temporal variations of the orders in a supply chain the more one moves away from the retail customer. Forrester showed that small changes in consumer demand result in large variations of orders placed upstream [4, 5]. It is interesting that this effect occurs even if the demand of final products is almost stable. For his studies, he assumed that some time delay exists between placing an order and the realization of this order (production). Furthermore, he assumed that each part of the supply chain plans its production and places its orders upstream taking into account only the information about the demands of its direct customer. 

The beer game demonstrates the value of sharing information across the various supply chain components. In practice, supply chains are usually more complex and much harder to manage. Current research has investigated that in practice the bullwhip effect is due to the following reasons ... 

The identification of these reasons led to recommendations how to avoid the bullwhip effect. Some of these recommendations are ... 

Whang, S. (1997) The bullwhip effect in supply chains. MIT Sloan Manag Rev, 38(3), 93-102. 

The bullwhip effect motivated research and practice to focus on crosscompany supply chain optimization of information and material flows between companies. Several authors specify a set of objectives related to cross-company supply chain optimization ... 

Another key challenge at the tactical level is to take into account the dynamics of the supply chain. Indeed, in recent years many suppliers and retailers have observed that while customer demand for specific products does not vary much, inventory and back-order levels fluctuate considerably across their supply chain. For instance, examining the demand for Pampers disposal diapers, executives at Procter Gamble noticed an interesting phenomenon. As expected, retail sMes of the product were fairly uniform there is no particular day or month in which the demand is significantly smaller or larger than any other. However, the distributors placed orders to the factory that fluctuated much more than retail sales. In addition, P G s orders to its supphers fluctuated even more. This increase in variability as we travel up in the supply chain is referred to as the Bullwhip effect. For more on this effect, see Simchi-Levi et al. (1999). 

Note that the effect of the information and delivery lags between the stages is to increase a customer demand increase from 4 to 8 units in period 11 to a wholesale order change from 4 to 27.04 in period 13 to a distributor order increase from 4 to 59.296 in period 15 and a factory increase in its brewed cases from 4 to 114.592. Each decision reflects a rational choice given the parameters. This increase in variance or orders as we go up a supply chain is referred to as the buUwhip effect. Demand updating is only one possible reason for the bullwhip effect. 

As inventory is maintained at various warehouses as well as the manufacturer, a large amount of material can be sitting in inventory in the supply chain. As replacement orders pass down the supply chain, the orders are pulled from the inventory that is available. If we look at a simple example of a retail store to a warehouse to the manufacturer, the order quantities will be smaller from the retail store to the warehouse, which will try to cover the orders quickly with inventory. When the inventory reaches a low level at the warehouse, a larger replacement order is placed with the manufacturer. The retail store may order 10 items per week, and the warehouse has an inventory of 100 items. For nine weeks there are no orders placed to the manufacturer and in the tenth week an order for 100 items is placed. If the manufacturer had 100 items in inventory they would have been held for more than nine weeks before they could be shipped to the customer. This adds storage costs to the manufacturer and can disrupt the production schedule when the large order arrives. This reaction in the supply chain is called the bullwhip effect. ... 

In order to overcome the bullwhip effect in the supply chain, better forecasting techniques need to be used to predict the demand for the material for the next period as the produetion sehedules are produeed. 

Multiagent Methodology to Reduce the Bullwhip Effect in a Supply Chain... 

Keywords Bullwhip effect Supply chain management Multiagent system Time series forecasting... 

Likewise, the Bullwhip Effect generated at each step can be defined as the ratio of the variance in orders sent to the upper node of the supply chain, and the variance in orders received from the bottom node of the supply chain. 

The results presented in Table 1 show, broadly speaking, the huge efficiency of the multiagent model developed in this paper versus one-one method. In all cases, the achieved results, in terms of Bullwhip Effect, improve the performance of the one-one model in several orders of magnitude. 

Multiagent Methodology to Reduce the Bullwhip Effect in a Supply Chain Table 2. Optimal Policy for each level of the supply chain in test A-1. 

The obtained results again demonstrate the effectiveness of multiagent model in reducing Bullwhip Effect generated along the supply chain. In all cases, the results generated by the one-one model are improved, although the difference is more relevant in some cases than in other ones. 

This situation evidences again that the use of simple forecasting methods, coordinated through a multiagent system allows a great improvement, in terms of Bullwhip Effect, comparing to the results of the one-one model. There is not clear proportionality between the result provided by the multiagent system and the result provided when all... 

When analyzing the results, it is more appropriate to do it from a relative point of view that from an absolute one. When considering a larger number of data, and since the series in some cases have definite trends, the values of the Bullwhip Effect are significantly lower than in the cases analyzed with random demands. 

A different situation is the one for the time series ALl 1 (Fig. 13). Multiagent model also allows a significant reduction in the Bullwhip Effect, as it is divided by 2.29, but not so large as in the previous case. Figures 14, 15 and 16 show the variations of purchase orders made by the four levels of the supply chain in the different situations. With these data, the multiagent system is not able to produce such a high improvement... 

The results obtained, and shown in Table 4, also show that close negotiation and collaboration in the supply chain between Factory and Wholesaler, on the one hand, and Shop Retailer and Retailer, on the other, is a very appropriate strategy for the reduction of the Bullwhip Effect. Collaboration significantly improves the performance... 

The results presented in this section show that the use of advanced forecasting methods leads to the reduction of Bullwhip Effect. Thus, the inclusion of ARIMA models at the lowest level of the supply chain provides very interesting results, and it can significantly reduce, in many cases, the Bullwhip Effect. In these circumstances, we... 

---