Impact of Lead-time Distribution on the Bullwhip Effect and Supply Chain Performance 

Abstract

In this paper, we simulate an extendable multi-agent linear supply chain to evaluate the impact of different lead-time distributions on bullwhip effect and supply chain performance under centralized and decentralized information sharing strategies. Given a fixed total lead-time across the supply chain, centralized information sharing and disintermediation improve the supply chain performance. A skewed lead-time distribution also reduces the bullwhip effect under decentralized information sharing strategy. Although the bullwhip effect remains unchanged, different lead-time distributions will lead to different supply chain performance. These insights can help practitioners in the re-configuration of the supply chain.

Keywords: Lead-time Distribution, Bullwhip Effect, Supply Chain Performance, Simulation, Information Sharing

1. Introduction 

Today’s global market has become more and more time-sensitive and time competitive. Shortening product life cycles, heightened expectation from customers, dangers of being dependent on a long forecast horizon in a volatile marketplace etc. have led both enterprises’ and academic researchers’ attention to the study of lead-time (Martin, 1998). Lead-time refers to the time lag between placing an order and receiving it (Li, 2000). It is one of the most important causes of bullwhip effect. 

Reasons to reduce lead-time have been grouped as: (1) improvement of the ability to quickly fill customer orders that cannot be filled from stock; (2) reduction in the bullwhip effect; (3) more accurate forecasts due to a decreased forecast horizon; (4) reduction in finished goods inventory levels (Simchi-Levi and Kaminsky, 2000). Fisher and Raman (1996) model a Quick Response System to shorten the lead-time. Chen (2000) quantifies the impact of lead-time on bullwhip effect. His model is closely related to our paper. 

In industrial practice, many strategies have been introduced to reduce lead-time. Wal-Mart uses cross-docking with most of its suppliers and shares its retail sales data (POS data) with Proctor and Gamble (P&G) to reduce lead-time. EDI and other advanced web technology have been used to shorten the order processing and communication time. To the suppliers who fail to recognize time as a competitive variable or whose systems cannot meet the needs for fast-changing market, the cost can be considerable. Compaq estimated that it had lost $500-$1bin in sales in 1994 because of stock-outs on its laptop and desktop computers (Martin, 1998)

Although lead-time reduction is well studied, there is relatively little research on lead-time distribution. We define lead-time distribution as varying the lead-time between each tier while keeping the total lead-time of the whole supply chain constant. There are two limitations in the previous research on the lead-time reduction: first, it only focuses on one or two tiers and does not have an overall view of the whole supply chain; and second, it does not include other performance metrics such as fill rate, total cost and inventory cost etc. into consideration. Our motivation for this research is to gain a deeper understanding of the behavior of supply chain network (SCN) under different lead-time distributions. In particular, given a fixed total lead-time, how the different distributions of this lead-time will affect the bullwhip effect and supply chain performance under both centralized and decentralized information sharing strategies. We also evaluate the effect of the number of tiers. The insights from our simulation results can help improving supply chain performance and point out direction for further optimization research.  

We have organized this paper as follows: Section 2 describes the bullwhip effect and information sharing strategies in the supply chain context; then introduces Chen’s quantification model of lead-time’s impact on bullwhip effect. Section 3 describes the simulation model and performance measures used in this model. Section 4 reports and analyses the simulation results. Section 5 summarizes our findings and suggests future research directions.

AMCIS final.rar