供應鍊管理中資訊共享策略之研究

行政院國家科學委員會補助專題研究計畫成果報告

※※※※※※※※※※※※※※※※※※※※※※※※

※

※

※ 供應鍊管理中資訊共享策略之研究 ※

※

※

※※※※※※※※※※※※※※※※※※※※※※※※

計畫類別：þ個別型計畫

□整合型計畫

計畫編號：NSC 89－2416－H－002－107－

執行期間：89 年 08 月 01 日至 90 年 07 月 31 日

計畫主持人：蕭正平 副教授

共同主持人：陳靜枝 副教授

本成果報告包括以下應繳交之附件：

□赴國外出差或研習心得報告一份

□赴大陸地區出差或研習心得報告一份

□出席國際學術會議心得報告及發表之論文各一份

□國際合作研究計畫國外研究報告書一份

執行單位：國立臺灣大學資訊管理學系

中華民國 90 年 10 月 30 日

行政院國家科學委員會專題研究計畫成果報告

供應鍊管理中資訊共享策略之研究

Study of Infor mation Shar ing Str ategy in Integr ated Supply

Chain Management

計畫編號：NSC 89-2416-H-002-107

執行期限：89 年 08 月 01 日至 90 年 07 月 31 日

主持人：蕭正平

國立台灣大學資訊管理學系

共同主持人：陳靜枝

國立台灣大學資訊管理學系

v 中文摘要

由於全球經濟體系的成形，企業無論大小都感受 到來自全世界的競爭壓力。不僅如此，競爭壓力更進一 步帶來對降低成本及快速反應的要求。企業不僅必須加 強本身的生產技術與成本控制，更必須在整個供應鏈 （supply chain）管理中創造更高的價值及新的市場機 會，以期對環境作出更好的回應。此外，網際網路的蓬 勃發展，也建立了電子商務的新環境。在要求個人化、 大量客製化（mass customization）及快速反應的電子商 務時代中，供應鏈管理也益形重要。供應鏈管理具價值 之處，即為能將所有牽涉產品生產、配銷過程的個體及 其相互鏈結與影響均列入考慮。其中十分重要的一點， 即為供應鏈中資訊的共享。由於生產及運銷的實際成本 及速度，受到技術面的限制較多，因此在降低產品成本 並提高反應速率的努力中，企業必須由試圖由供應鏈中 獲得有效的資訊，進而制定其存貨及生產之原則與安 排，方能達成零延遲之生產與供銷，並提昇整體供應鏈 網路之效率。由於供應鏈管理十分複雜，因此，若能先 針對供應鏈之特性與結構加以分類，並對不同型態之供 應鏈採取不同之資訊共享策略，進而研究其不同績效， 將有助於進一步建立其數量化模型及最佳化之研究；更 可以幫助實務界及學術界更瞭解供應鏈之特性及其應 用。因此本研究分析供應鏈網路中資訊共享的模型，並 以模擬的方式，瞭解在不同型態之供應鏈網路中，是否 需要不同的資訊分享策略。同時在此過程中，亦能深入 認識不同資訊分享策略，瞭解不同策略之特性，作為進 一步制訂策略的依據。 關鍵詞：供應鍊網路、資源分配、半導體製造。

v Abstract

This paper studies the effects of information sharing strategies on the performance of a supply chain. We first consider four types of information sharing strategies: (1)order information sharing where every stage of the supply chain only knows the orders from its immediate downstream stage; (2)demand information sharing where every stage has full information about the market demand; (3)inventory information sharing where each stage shares its inventory levels and demand information with its immediate upstream stage; and (4)shipment information sharing where every stage shares its shipment data with its immediate upstream stage. Our results indicate that information sharing improves supply chain performance of overall inventory cost and fill rate when demand is relatively stable. Then we show that different forms of information sharing have their “signature” performance and may worsen some performance metrics when the variability of demand is high. Finally, we find that a hybrid information sharing strategy, which uses demand information sharing in the distribution network while using

inventory information sharing in the supplier network, is a better strategy to improve the overall performance of the supply chain when demand mix is volatile.

Keywor ds: Infor mation Shar ing; Supply Chain Management; Value of Infor mation; Stochastic Demand; Pr oduct Mix Var iability

v Introduction

More and more companies have recognized that there is a direct link between the performance of supply chains and the availability and quality of timely information. It is widely known that Wal-Mart and Proctor & Gamble (P&G) share information regarding the retail sales of P&G products at Wal-Mart stores. This information enables P&G to do a better job of managing its production of these products and provides Wal-Mart with greater “in store” availability. Furthermore, companies such as Dell and Cisco are sharing information with suppliers and customers to reduce working capital and inventories. The flow of information through the supply chain enables them to match supply closely to consumer demand and to anticipate changes in the marketplace. The wide use of advanced information technologies (e.g., EDI and Web technologies) in supply chains also suggests that companies have come to realize the importance of information sharing.

In fact, many supply-chain problems can be attributed to lack of information sharing between supply chain members. One important observation in supply chain management, prominently known as the bullwhip effect, suggests that demand variability be magnified, as it is further upstream in the supply chain. The bullwhip effect is an important concern in supply chain management for several reasons. First of all, the increased order variability requires each supply chain member to hold excessively high inventory levels in order to meet a boom-and-bust demand pattern. Secondly, despite the overall overstocking throughout the supply chain, the lack of synchronization between supply and demand could lead to complete stock out at certain times. Finally, the bullwhip effect increases not only the physical inventories but also operating costs. Poor demand forecasts based on the distorted orders result in erratic capacity planning and missed production schedule. Therefore, the bullwhip effect should be minimized. Information sharing among supply chain members can provide benefits in terms of

supply chain visibility and reduced order variability. For example, sharing of demand information enables each of supply chain partners to forecast accurately based on real demand.

Another important observation is that supply chains with expanding product varieties are faced with increasing uncertainty in demand mix. Many companies have customized their products to satisfy the requirements of different market segments. The demand mix of a customizable product may change widely while the total demand does not change very much. Previous research in supply chain management has mainly concentrated on reducing demand mix uncertainty through delayed product differentiation. In this paper, however, we will show that enhancing the value of information sharing among supply chain members can significantly reduce the uncertainty related to product variety.

The value of information sharing can be defined as the benefits realized from obtaining or sharing information minus the costs associated. The cost of an information sharing policy may include the additional information cost and coordination cost. The information cost may include information systems investment and other charges by either suppliers or customers for providing information, while the coordination cost may include communication costs and administration costs. Recent development in information technologies, such as Web technologies and Enterprise Resource Planning (ERP) systems, dramatically reduce both the information cost and the coordination cost. In this paper we consider the benefits of information sharing and ignore the technology and coordination costs involved.

Recently, academic researchers have showed a growing interest in the value of information sharing in supply chains. Lee, Padmanabhan and Whang (1997) found that sharing real demand information across the supply chain members reduces the bullwhip effect. Chen (1998) studied the relative benefits of echelon-stock policies over those of installation stock policies in a multi-echelon environment. The latter decisions at a given facility depend only on local inventory information as opposed to this information combined with information on all downstream facilities. Chen et al. (1999) quantified the bullwhip effect for multiple-stage supply chains with and without centralized demand information and demonstrated that centralizing demand information can significantly reduce but not completely eliminate the bullwhip effect. Gavirneni, Kapuscinski, and

Taylur (1999) analyzed information flow between a supplier and a retailer in a two-echelon model of capacitated supply chain. They studied the relationships between capacity, inventory, and information at the supplier level and how these relationships are affected by inventory system and demand distribution. Using a multi-agent simulation model, Tan (1999) tested the impact of information sharing on supply chain performance. She studied how various information sharing policies, i.e., no

information sharing, sharing of complete demand information, sharing of downstream customer’s shipment data, and sharing of downstream customer’s inventory information, behave under different supply chain structures and demand patterns. One of her interesting findings is that a hybrid information sharing policy improves supply chain performance under volatile demand. Lee et al. (2000) studies how to use shared information to improve the supplier’s order quantity decisions with a known autoregressive demand process. Cachon and Fisher (2000) investigated a supply chain model with one supplier, and N retailers, stochastic consumer demand, and batch ordering. They showed analytically how the manufacturer benefits from using information about the retailer’s inventory levels.

While the value of information sharing is widely recognized, we want to investigate how information sharing affects supply chain performance, what types of information supply chain members should share, and how they should share it. In this paper, we first study the four common information sharing models in a linear supply chain with N stages: order information sharing (Model 0), demand information sharing (Model 1), inventory information sharing (Model 2), and shipment information sharing (Model 3). In model 0, each stage of the supply chain does not know the status of its downstream stages and forecasts are based only on the orders from its immediate downstream stage. The Beer Game is probably the most famous case of Model 0 in a traditional supply chain. Even when the end consumer demand is relatively stable, the bullwhip effect intrinsic in the chain leads to poor forecasts and high inventories. Model 1 assumes total real demand visibility. Real-time demand information is transmitted from the end-consumer back through every stage in the supply chain. This means that any real change in demand can be known at all points in the supply chain. Demand information helps channel partners forecast demand more accurately, reduce safety stock, and anticipate customer needs. Direct sales model, sharing of POS data, and collaborative planning and optimization belong to this type of information sharing. In Model 2, each stage of the supply chain shares information about its inventory and actual demand with its supplier. By monitoring its downstream inventory levels, the supplier can synchronize its production and delivery schedule with the downstream customer’s demand and hence maintain a high level of availability. This strategy is currently common in the grocery and fashion retailing industry. Vendor managed inventory (VMI), schedule sharing window, and continuous replenishment belong to this type of information sharing. Model 3 assumes that each stage knows its downstream customer’s shipment data. For instance, in the computer industry, manufacturers, such as HP and IBM, request sell-through data on withdrawn stocks from their resellers’ central warehouse. Since the downstream customer deliveries can be viewed as a proxy for the downstream customer demand, each stage can use its

downstream customer’s shipment data to create more accurate forecasts and target inventory level. Unaware of possible downstream stock out, however, this strategy may underestimate downstream’ orders received and the needed buffer inventories and incur many stock-outs. To summarize, these models constitute a range of levels of information sharing: Model 0 is a no information sharing case, Model 1 is a full demand information sharing case, and Model 2 and 3 are one-stage information sharing cases.

v Model Formulation

Consider a linear supply chain with N stages. For a given stage k, stage k-1 is its customer and stage k+1 is its supplier. The end customer in the supply chain is called the consumer. Consumer demand arises at stage 1, stage 1 orders from stage 2, etc., and stage N orders from an outside supplier. This triggers material flows in the opposite direction. We first assume that each stage has a fixed lead-time and uses a periodic review inventory policy with a fixed review time, one period. When the demand in a period exceeds the on-hand inventory, the excess is backordered. We also assume that each stage maintains a high service level so that each of the N stages can control its inventory “locally.” Finally, we assume that each stage faces a normal demand and the demands are independent across periods. Such an assumption is reasonable when the service level at intermediate stage is very high, 95% or higher, and when the demand for the product has a small coefficient of variation. However, we do not assume that each stage knows its demand process. Instead, it must still forecast the mean and variance of the demand. The objective is to find out how information sharing affects the performance of the supply chain.

We make use of two subscripts and one superscript in the model. The first subscript refers to the echelon, the second to the time epoch and the superscript to the model number. Thus

X

1_kt will be the value of variable X at stage k in period t in Model 1. For stationary parameters, we omit the time period. For each stage and period, define:

L = lead time plus 1 (review period), H = unit holding cost rate,

D = real consumer demand with a mean of

ì

and a variance of 2 ó ,

S = the target inventory level, Q = the quantity of stock ordered,

ìˆ

= forecast demand,

ó

ˆ

_{= standard deviation of errors of forecasts,} β = average fill rate.

This paper treats decentralized environments in which each stage makes its own decisions based on what it knows about the supply chain. The assumption of complete information usually made in multi-echelon inventory theory may be unrealistic in real supply chains. Within each period, for a given stage k, the following sequence of events occurs: (1)the inventory decision is made with a target inventory level and an order is placed to its supplier; (2)demand is realized

outbound shipments are released; (3)inbound shipments are received and inventory and fill rate are assessed. A periodic inventory policy requires each stage of the supply chain to raise its inventory level up to a given target level in each period. One common form of this policy is to set the target inventory level in period t, Skt, as given by (Lee and Billington 1993):

Skt = Lk

ìˆ

_kt+ z

L

_k

óˆ

_kt, (2.1) where z is the safety-stock factor associated with the customer service level and is fixed for each of the N stages, and z L_k

óˆ

ktis an estimate of safety stock. In this paper, however, we use a simplified periodic review policy where the target inventory level is of the form:

Skt = (Lk+ SSk )

ìˆ

_kt. (2.2) Note that (2.2) is similar to (2.1) with the safety stock z

k

L

óˆ

ktreplaced by SSk

ìˆ

kt, i.e., the safety stock is expressed in units of the forecasted demand. If the mean and standard deviation faced by stage k are

k

ì

and

ó

_k, we have z

L

_k

ó

k= SSk

ì

k. Given the service level, a larger value of SSk signals a higher coefficient of the demand at stage k.

The following performance measurements are used to evaluate the information sharing models:

Inventor y cost. Let

ì

_k and

ó

_k, respectively, be the mean and standard deviation of the demand faced by stage k. The average inventory level per period is the sum of safety and average cycle stock, and is given by z

L

_k

ó

_k+

ì

_k/2. The average on-order inventory is Lk

ì

_k. Since we assume complete backordering, and hence no demand is lost in the system, so

ì

_kis equal to

ì

. The total supply chain inventory cost per period is given by

(

)

∑

=

+

+

N 1 k k k k k

L

ì

ì

/

2

z

ó

L

h

(2.3)

Fill r ate. The fill rate measures the proportion of

demands that are met from the inventory on hand. It is an important indicator of service level. The long-run relationship between the safety-stock factor and fill rate is βk = 1 −

L

_k

ó

kGu(z) /

ì

(Silver and Peterson, 1985). The average of the fill rates over all the stages can be expressed as

(

)

∑

=

−

N 1 k k k u

(

z

)

ó

L

/

ì

G

1

N

1

, (2.4)

where Gu(z) is the probability that a unit normal variable takes on a value of z or larger (Nahmias 1985).

As shown in equation (2.3) and (2.4), for a fixed service level, lower order variance allows each stage to carry less safety stock on average and thus reduces the total inventory cost; for a fixed service level, lower order variance increases the overall fill rate. Hence,

the performance of the supply chain squarely relies on demand uncertainty seen by each stage. Inventories are often used to protect the supply chain from uncertainties, but it is an expensive solution. We will demonstrate how information sharing can reduce order uncertainty at each stage of the supply chain and hence improve the performance of the supply chain.

v Model 0 -- Order Information Sharing

In Model 0 (Chen et al,1999), demand forecasts at each stage are based only on local “demand” information, i.e., the orders from the immediate downstream stage. We assume that each stage uses the simple moving average forecast method with n observations to estimate the mean of demand, i.e.,

t 1

ìˆ

= n _{D n} 1 i i t

∑

= − and

ìˆ

_kt = n Q n 1 i i t , 1 k

∑

₌ − − k= 2,… , N, (2.5)

where

Q

_k−1,t−iis the order placed by stage k-1 in period t-i.

Suppose that each stage, k, follows a period review policy where the target inventory level is given by (2.2) and the safety stock, SSk, is chosen to buffer

against the order variability from stage k-1. At stage k, we can determine the variance of

Q

_kt relative to the variance of its demand,

Q

_k−1,t. So we write

Q

_ktas

kt

Q

=

S

_kt – (

S

_k,t−1−

Q

k−1,t−1).

Note that

Q

_kt may be negative, in which case we assume that the excess inventory is returned without cost. Using (2.2) and (2.5), we can write the order quantity

Q

_kt as

kt

Q

= (Lk+ SSk )

ìˆ

kt− (Lk+ SSk )

ì

ˆ

k,t−1+

Q

k−1,t−1

=(1+(Lk+SSk)/n)

Q

k−1,t−1–((Lk+SSk)/n)

Q

k−1,t−n−1. Taking the variance of

Q

_kt, we get

( )

[

( ) ( )

]

( k1) 2 2 k k k k k 1 2L SS n 2 L SS n Var Q Q Var = + + + + − . (2.6) Hence we can deductively derive the following

expression for the variance of the orders placed by stage k for Model 0,

Q

_k0, relative to the variance of real demand ( )Q

_[

(L SS)n (L SS) n

_]

Var( )D Var k j j j j j k       + + + + = _∏=1 2 2 0 2 2 1 ,k=1,… ,N. (2.7) The increase in demand variability is an increasing function of Lk, the lead times, and SSk, the safety stocks,

and a decreasing function of n, the number of observations used in demand forecasting. This paper treats Lk and n as constants. So the variability of the

safety stock directly contributes to the increase in order variability. More importantly, the variance increases multiplicatively at each stage of the supply chain. This expression shows the bullwhip effect that demand variance increases quickly as it moves up a decentralized supply chain. Empirical evidence also shows that the orders placed by a retailer tend to be more variable than the consumer demand seen by that retailer.

Lemma 1. If each stage of the supply chain uses a simple moving average forecast with n periods, a periodic review policy defined in (2.2), and the demand process seen by each stage is an i.i.d. normal, the increased order variability results in a high supply chain inventory cost and a low fill rate.

v Model 1 -- Demand Information

Shar ing

On the other extreme, demand information sharing assumes that the first stage (i.e., the retailer) shares its real-time demand information with each of the upstream stages. Since each stage has real demand information, each stage will use the same estimate of the mean demand, i.e.,

n D ìˆ n 1 i i t t       =

∑

= − , (2.10)

Consider an echelon inventory policy where the target inventory level is given by

t k 1 i i i kt (L SS)ìˆ S       ₊ =

∑

= , k = 1,… , N, (2.11) Where the safety stock, SSi, is chosen to buffer against

the end demand uncertainty. According to (2.6), (2.10), and (2.11), we have the following expression for the variance of the orders placed by stage k,

Q

1_k, relative to the variance of retail demand.

(L SS) n 2 (L SS) n Var( )D 2 1 ) Q ( Var 2 2 k 1 j j j k 1 j j j 1 k             ₊ + + + = _{∑ ∑} = = ,k=1,… ,N.(2.12)

In comparison with (2.7), (2.12) demonstrates that the increase in demand variability at each stage of the supply chain is additive instead of multiplicative. Chen et al. (1999) showed that if demand information is shared, the increase in variability seen by each stage of the supply chain is the same whether the supply chain follows an echelon inventory policy or not. Hence demand information sharing can reduce the bullwhip effect and reduce the safety stock.

Theor em 1. Under the conditions of Lemma 1, sharing of demand information decreases supply chain inventory by reducing the bullwhip effect. However,

(a) it may improve supply chain fill rate under stationary demand,

(b) it worsens the fill rate under non-stationary demand.

v Model 2 -- Inventory Information

Shar ing

In this type of information sharing, each stage shares its inventory and demand information with its immediate upstream stage. Suppose stage k-1 shares its actual demand and inventory status with stage k period by period. For any period t, the following are defined for stage k-1 after an order is placed and demand occurs: on-hand inventory Ik-1,t; backorders,

Bk-1,t; on-order inventory, OIk-1,t; inventory position,

t , 1 k t , 1 k t , 1 k t , 1 k I B OI

IP₋ = ₋ − ₋ + ₋ ; and target inventory level,

Sk-1,t. By knowing these information, stage k can derive

the demand to stage k-1,

t , 1 k 1 t , 1 k 1 t , 2 k S IP Q_{− −} = _{− −} − ₋ , and

the demand of stage k-1, t , 1 k t , 1 k t , 1 k S IP Q ₋ = ₋ − ₋ . Two distinct characteristics of this relationship are the following. First, because stage k knows the demand to stage k-1, both stage k and stage k-1 can forecast the mean of demand based on the orders of stage k-2 in n periods.

n

D

ìˆ

ìˆ

n 1 i i t t 2 t 1

∑

= −

=

=

, k = 1, 2, and kt

ˆ

µ

= _ì n _{Q n} i i t k kt

∑

= − − = 1 , 2 ˆ , k = 3,… ,N, (2.15) where i t , 2 k

Q _{− −} is the order placed by stage k-2 and received by stage k-1 in period t.

Moreover, stage k can plan its safety stock based on the demand of stage k-1 in order to provide a given customer service level. Therefore, by knowing its downstream inventory information, stage k can implement the echelon-based inventory control. Hence we can derive the variance of the orders placed by stage k: ( )D Var n SS L n SS L Q Var k j j j k j j j k             + + + + = ∑ ∑ = = 2 2 1 1 2 ) ( 2 ) ( 2 1 ) ( , k=1,2, and ( )2 2 2 2 1 1 2_{) 1 2 ( ) 2 ( )} ( − = =             ₊ + + + = _{∑ ∑} k k j j j k j j j k L SS n L SS n VarQ Q Var ,k=3,… ,N, (2.16)

where the safety stock, SSj, is based on the variability

of demand to stage j.

In comparison with (2.7), (2.16) demonstrates that the increase in demand variability between stage k and stage k-1 is additive not multiplicative. Stage k uses the actual demand at stage k-1, which is less variable than the orders placed by stage k-1, to create more accurate forecasts. Thus, Model 2 eliminates one stage of information distortion, i.e., stage k-1, and consequently reduces some degree of the bullwhip effect. Moreover, by monitoring its downstream inventory levels, stage k synchronizes its production and/or delivery schedules with the downstream demand to ensure that products are consistently available to the customer. So each stage can maintain a stable service level.

Theor em 2. Under the conditions of Lemma 1, compared with Model 0, sharing of inventory information not only improves supply chain fill rate but also reduces certain degree of supply chain inventory; compared with Model 1, this policy increases inventory cost when end demand is volatile.

v Model 3 -- Shipment Information

Shar ing

In Model 3, each stage of the supply chain knows its downstream customer’s outbound shipment data. A shipment represents the amount of a product each stage immediately ships to its customer in response to a customer order after previous backorders are met. Suppose stage k-1 shares its shipment data with stage k.

Let Wk-1,t and Ik-1,t, respectively, be stage k-1’s

outbound shipment and on-hand inventory in period t. We have Wk-1,t = min{Ik-1,t,

Q

k−2,t}. If each stage

maintains a high fill rate, the downstream shipments are very close to the demand at the downstream stage. In this case, we can use the variance of downstream shipments to approximate that of the demand to the downstream stage. It follows from (2.16) that the variance of the orders placed by stage k can be expressed as: ( ) ( )3 1 2 2 1 1 3 ) ( 2 ) ( 2 1 L SS n L SS n VarW Q Var k j j j k j j j k             ₊ + + + ≈ ∑ ∑ = = ,k=1,2, and ( ) ( )3 1 2 2 1 1 3 _{1 2 ( ) 2 ( )} − = =               ₊ + + + ≈ ∑ ∑ k k j j j k j j j k L SS n L SS n VarW Q Var ,k=3,… ,N, (2.19)

where the safety stock, SSj, is based on the variability

of the shipments of stage j-1.

Like Model 2, this model eliminates one stage of distortion since the shipment data, unlike the downstream orders, are not subject to the bullwhip distortion. However, when stock out occurs at the downstream, the outbound shipments under- represent the demand and consequently underestimate the inventory needed to buffer against demand uncertainty. In the long run, the mean of the shipments of stage k-1 can be expressed by the product of the mean of the expected demand

ì

and the fill rate βk-1. Thus,

planned on the downstream shipments, the expected target inventory at stage k is given by:

3 k S =

( )

3 kt S E =(Lk+ SSk )βk-1E

( )

ì_k−1,_t =(Lk+Sk)βk-1µ, (2.20)

If the expected target inventory level Sk= (Lk+

SSk )

ì

gives a safety factor of z, it follows from (2.1)

and (2.20) that

S

3_k gives a safety factor of

[

âk−1−(1−âk−1)Lk SSk

]

z. Hence compared with

Model 2, this model decreases service level and consequently stock out occurrences increases and fill rate decreases. Therefore, this policy is more sensitive to the order variability and may lead to a high backlog and a low fill rate under volatile demand.

Theor em 3.Under the conditions of Lemma 1, if each stage of the supply chain has a really high customer service level, then sharing of one-stage shipment data helps improve supply chain performance. Otherwise, it may underestimate the safety stock to buffer against the demand uncertainty and result in a low fill rate under unstable demand.

v Conclusion

Companies have long been aware of the value of information sharing in supply chains; however, there has not been much research on what kind of information supply chain members should share and how to share it in order to achieve the appropriate supply chain performance. Our study offers the following insights into these long-standing concerns. First, information sharing helps counter the phenomenon of demand variability amplification, mainly caused by the time lag between channel partners and the safety stocks in the supply chain. In traditional supply chains, orders are the only information channel members exchange. These orders are usually highly distorted by the lead times and the

high safety stocks at the downstream stages and convey very limited portion of real demand information. This information distortion is further magnified as it is further upstream in the supply chain and consequently the increased demand variability leads to overstocking throughout the system. There may also be a high backlog and a low fill rate since a boom-and-bust order pattern results in a very high inventory at some times and complete stock out at other times. The demand information sharing policy reduces the bullwhip effect to a large extent while the inventory information sharing policy and the shipment information sharing policy reduce at least one level of information distortion.

Second, the impact of information sharing on supply chain performance largely depends on the underlying demand process and the supply chain structure. There is no information sharing policy that is uniformly superior to the others because each supply chain has its unique characteristics. In this paper, we first consider a generic supply chain of a single product in order to study how the benefits of information sharing policies are influenced by demand patterns. We find out that various information sharing schemes unanimously improve supply chain performance under relatively stable demand. Under volatile demand, however, different information sharing strategies affect different performance measures differently. The demand information sharing policy can significantly reduce the bullwhip effect and thus the safety stocks. But the minimum safety stock at each stage causes an increase in backlog and a drop in fill rate when the end-demand variability is high. Although the inventory information sharing policy does not perform as well as the demand information sharing policy in terms of inventory savings, this strategy is able to maintain a high fill rate by maintaining sufficient inventory. Like the inventory information sharing policy, the shipment information sharing policy can reduce one level of information distortion and thus supply chain inventory. By not considering backlog, it underestimates the downstream demand and may lead to a high backlog and a low fill rate when demand is highly unpredictable. Therefore, corporations often need to trade-off gains in some performance metrics against losses in other measures.

More studies are needed to look into matching the product, demand process, production and distribution process, and supply chain structure with the right information sharing strategies. This paper has only examined two information sharing schemes (one-stage inventory and demand sharing, and one-stage shipment information sharing) between the two extremes of no information sharing and real-time demand information sharing, and only one hybrid sharing approach. Other forms of sharing schemes and hybrid combinations could be developed and applied to various demand and supply situations.

v References

[1] Cachon, G. and M. Fisher, 2000, “Supply chain inventory management and the value of shared

information,” Management Sci., 45(8), pp. 1032— 1048.

[2] Chen, F.Y., Z. Drezner, J.K. Ryan, and D. Simchi-Levi, 1999, “The bullwhip effect: managerial insights on the impact of forecasting and information on variation in a supply chain,” Quantitative models for supply chain management, Tayur, S. eds.Kluwer Academic Publishers, Norwell, MA, pp. 417— 439.

[3] Chen, F.Y., Z. Drezner, J.K. Ryan, and D. Simchi-Levi, 2000, “Quantifying the bullwhip effect in a simple supply chain: the impact of forecasting, lead times, and information,” Management Sci., 46, pp. 436— 443. [4] Chen, F., 1998, “Echelon reorder points,

installation reorder points, and the value of centralized demand information,” Management Sci., 44, pp. S221— S234.

[5] Gavirneni, S., R. Kapuscinski, and S. Tayur, 1999, “Value of information in capacitated supply chains,” Management Sci., 45, pp. 14— 24. [6] Lee, H.L., C. Billington, 1993, “Material

management in decentralized supply chain,” Operations Research, 41, pp. 835— 847.

[7] Lee, H.L., P. Padmanabhan, and S. Whang, 1997, “Information distortion in a supply chain: The bullwhip effect,” Management Sci., 43, pp. 546— 558.

[8] Lee, H.L., K.C. So, and C.S. Tang, 2000, “The Value of Information sharing in a two-level supply chain,” Management Sci., 46 pp. 626— 643. [9] Li, L., 1992, “The role of inventory in

delivery-time competition,” Management Sci., 38, pp. 182— 197.

[10] Nahmias, S., 1989, “Production and Operations Analysis.”

[11] Richard Irwin, Homewood, IL.Silver, E.A. and R. Peterson, 1985, “Decision Systems for Inventory Management and Production Planning,” John Wiley, New York.

[12] Tan, G.W., 1999, “The impact of demand information sharing on supply chain network,” PhD thesis, Department of Business

Administration, University of Illinois at Urbana-Champaign, Urbana, IL.