ARTICLE IN PRESS Int. J. Production Economics 120 (2009) 162–175
Contents lists available at ScienceDirect
Int. J. Production Economics journal homepage: www.elsevier.com/locate/ijpe
A study on evaluation of demand information-sharing methods in supply chain Seung-Jin Ryu a,, Takahiro Tsukishima b, Hisashi Onari a a b
Waseda University, 51-14-07, 3-4-1 Okubo, Shinjuku, Tokyo 169-8555, Japan Hitachi, Ltd., Japan
a r t i c l e i n f o
abstract
Article history: Received 1 September 2007 Accepted 1 July 2008 Available online 18 October 2008
With the increasing globalization, there are many sources of uncertainty across the entire high-tech electronics equipment supply chain. These include demand uncertainty and supply uncertainty due to the use of unskilled labor, as well as the sudden breakdown of production facilities of upstream players. These cause many problems, including excess inventory and poor service, to name but two. Information sharing is a very important tool in dealing with these sort of problems because it helps to eradicate potential uncertainties related to various corporate behaviors. Although there has been much research on this subject since the 1990s, there have not been sufficient efforts to identify the control characteristics, or the responsiveness of various information-sharing methods. Our objective here is to clarify the control characteristics of informationsharing methods closely related to the supply chain community strategy. In this study, we evaluate the supply chain performance of two different types of information-sharing methods. One is the planned demand transferring method (PDTM), and the other is the forecasted demand distributing method (FDDM). We analyze supply chain performance for both methods in terms of throughput, inventory level, and service level. & 2008 Elsevier B.V. All rights reserved.
Keywords: Supply chain Information sharing Community strategy Demand information management Control characteristics Responsiveness Planned demand transferring method (PDTM) Forecasted demand distributing method (FDDM)
1. Introduction Supply chain (SC) collaboration has been strongly advocated by consultants and academics alike since the mid-1990s, under the banner of concepts such as vendor managed inventory (VMI), collaborative forecasting planning and replenishment (CPFR), and continuous replenishment (CR). It is widely accepted that creating a seamless, synchronized SC leads to increased responsiveness and lower inventory costs. Reducing uncertainty via the transparency of information flow is a major objective in external SC collaboration. Unpredictable or
Corresponding author. Tel./fax: +81 3 5286 3296.
E-mail addresses:
[email protected] (S.-J. Ryu),
[email protected] (T. Tsukishima),
[email protected] (H. Onari). 0925-5273/$ - see front matter & 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.ijpe.2008.07.030
non-transparent demand patterns have been found to cause artificial demand amplification in a range of settings. This leads to poor service level, high inventories, and frequent stock-outs (Holweg et al., 2005). Many companies, however, are reluctant to share information with their trading partners, afraid that the information will be used unfairly to the partners’ advantage. In order to motivate these companies to share information, they need to be aware of the benefits that information sharing systems can bring (Zhao et al., 2002). Back in the 1990s, research aimed at taking advantage of visibility for sales trends at that time. Bourland et al. (1996) and Chen (1998) showed the benefits of information sharing through offering sales information. On the other hand, Aviv (2001) and Cachon and Lariviere (2001) and suggested demand forecast sharing to address problems arising from the bullwhip effect. Moreover, Gavirneni et al. (1999) showed the benefits of inventory
ARTICLE IN PRESS S.-J. Ryu et al. / Int. J. Production Economics 120 (2009) 162–175
information sharing. However, there have still not been enough efforts to identify the control characteristics and the responsiveness of information-sharing methods. Our objective is to clarify the control characteristics of two different information-sharing methods and to verify the specific conditions in which the information sharing performs well. In this study, we examine SC community strategies and the types of demand information management closely related to information-sharing methods. Following this, we define two different information-sharing methods according to types of shared information and sharing procedures. Next, we evaluate the SC performance of two informationsharing methods in terms of throughput, inventory level, and service level. Finally, we show the results of the feasibility study on the high-tech electronics industry.
2. Objectives of research With the increasing globalization of the high-tech electronics equipment SC, there are many sources of uncertainty across the entire SC. These include demand uncertainty and supply uncertainty due to the use of unskilled labor, as well as the sudden breakdown of production facilities of upstream players, to name but two Tang (2006). These force downstream companies to maintain more inventory than they can handle with such uncertainties. Otherwise, they would be harmed by lowered service level. It is obvious that utilizing timely and accurate information about demand and adjacent players is the best way to cope with the various uncertainties of the SC. However, we have little idea about how to obtain this information from adjacent SC players or independent third party organizations. Furthermore, it is very important to know which types of information should be obtained to perform better. It is also important to understand how effectively the information obtained works on the operations of SC players. Our primary objective here is to define the appropriate types of shared information related to inventory status, and to suggest detailed procedures for information sharing which are easily linked to current planning processes. Moreover, we will evaluate information-sharing methods and verify which method is best at capturing timely and accurate demand information. We will go on to explain the simulation results in a way designed to be easily understood by practitioners. The work done here follows on from that of Bourland et al. (1996), who identified the benefits of realized-demand information sharing using electric data interchange (EDI). They demonstrated that the inventory related benefits of information sharing are sensitive to demand variability, safe stock level and the length of the order cycle. Lee et al. (1997) analyzed the demand variability amplification along a SC from retailer to distributors, and named this amplification effect the bullwhip effect. They made a significant contribution by identifying four causes for the bullwhip effect and developed strategies to mitigate it. Chen (1998) studied inventory policy and assessed the
163
value of the sales information known as centralized demand information (CDI). He verified that when there are increases in the number of stages, lead-time and batch sizes the value of information sharing also tends to increase. Higher demand variability decreases the value of information sharing. Chen (1999) also showed that the system becomes much more robust when the upstream managers are able to respond to end customer demands instead of orders from downstream players. Making accurate customer demand information available to the upstream members of the SC also seems to make the system more robust. Gavirneni et al. (1999) studied the benefits of inventory information sharing such as inventory policy parameters and inventory levels. He found that when the end-item demand variance is moderate and the value of D ¼ Ss (the difference between parameters of ordering policy) is not extreme, the supplier attains a great cost reduction. Lee et al. (2000) quantified the benefits of sales information sharing and identified the drivers that have significant impacts. They verified that a manufacturer obtains larger inventory/cost (inventory holding and shortage) reduction when demand is highly correlated over time, highly variable, or when the lead-time is long. Aviv (2001) studied the interaction between inventory and demand forecasting. He verified that collaborative forecasting tends to be more successful in the presence of high diversification of forecasting capabilities across the SC. The absolute and the marginal benefits of collaborative forecasting are larger when the lead-time is smaller. Cachon and Lariviere (2001) studied contracts that allow the SC to share demand forecast credibly. They showed that under forced compliance, the manufacturer is nevertheless able to share its information credibly and still expropriate all SC profits. Under voluntary compliance, the manufacturer is still able to share its forecast credibly, although not for free. Chen et al. (2006) proposed a risk sharing contract that requests the retailer to partially compensate for the maker’s loss in the case of a long lead-time and demand information updating. Li (2002) identified the incentives for firms to share information vertically in the presence of horizontal competition. They found that the direct effect always discourages retailers from sharing their information. The leakage effect discourages the retailers from sharing their demand information while encouraging them to share their cost information. Yao et al. (2008) analyzed the interaction between revenue sharing and quality of order fulfillment in terms of incentives. Kulp et al. (2004) developed a conceptual framework for information integration of information sharing and collaboration and empirically examined the framework through survey. Dejonckheere et al. (2004) examined the beneficial impact of information sharing in a multi-echelon SC. For the SC with information enrichment, the increase of the bullwhip effect will generally be of a linear nature with the level in the chain instead of geometrical. Jain and Moinzadeh (2005) analyzed the implications of reverse information sharing. Here, the retailer takes advantage of reverse information to reduce its inventory cost. Watson and Zheng (2005) showed the benefits of
ARTICLE IN PRESS 164
S.-J. Ryu et al. / Int. J. Production Economics 120 (2009) 162–175
sharing real-time sales data across all stages of SC by eliminating information delays and mitigating costs. Gaur et al. (2005) studied how the time-series structure of the demand process affects the value of sales information sharing. Chen and Yu (2005) demonstrated the potential value of delivery lead-time information, which is significant when the lead-time distribution exhibits high variability. Yao et al. (2005) analyzed how the manufacturer designed its return policy when it knew the demand ratio forecasts. Yue and Liu (2006) analyzed the value of demand forecast sharing in a direct channel SC. Li et al. (2006) studied the value of timely supply information and quantified the impact of upstream disruption on SC. Supply information sharing enhances the agility of firms while improving the stability and performance of the whole SC. Byrne and Heavey (2006) showed a case study results from real industrial example. And Welker et al. (2008) investigated the influences of the type and the level of SC integration using multi-case study among SMEs. 3. SC community strategy and types of demand information In this section we introduce SC community strategies and three types of demand information. 3.1. SC community strategy SC community strategies which the companies adopt to make a successful SC are closely related to demand information management (Table 1). 3.1.1. Execution oriented community strategy Specialized digital equipment makers emerged as a result of increased part modularization in the 1990s. Each company independently forecasts its demand and makes its production/sales plan for only its own profitability. In this type of community strategy, companies are loosely coupled at execution level. The objective is to make the SC obtain smooth material flow.
3.1.3. Strategic coordinated community strategy Inside this community everyone along the SC closely coordinates at the planning level and the execution level. The flat screen TV industry at present is the typical example of this kind of community. A strategic coordinated community strategy helps to achieve efficiency across the entire SC and it is possible to achieve a global optimum.
3.2. Types of demand information We identify types of demand information according to their time stamp. There are three types of demand information classified according to where they are located along the time axis. These are realized demand information, planned demand information, and forecasted demand information. See Fig. 1. Realized demand information is acquired from a sales or shipment record. It is used in forecasting process as an input. When there is no certain tendency or high variability in market demand exists, it is difficult to make a precise forecast. There are two kinds of forecasted demand information: planned demand information and expected demand information. Planned demand information is the forecasted demand for the near future. It has high forecasting accuracy. It can be shared with upstream SC players. It is difficult, however, to share between players which are non-affiliated or have no partnership. Expected demand information is the forecasted demand for the more distant future. A greater degree of error is expected and more frequent change is possible than with planned demand information. Expected demand information can be shared easily than ‘planned demand information’, because it contains little interior information about the player. It can be shared among two or more players along the SC. The characteristics of demand information are summarized in Table 2 below.
4. Demand information management 3.1.2. Vertical inclined community strategy At the beginning of the 2000s, severe competition reduced the number of upstream part suppliers but made them more powerful. An upstream player can take advantage of the procurement plan of its downstream player. This type of community strategy make the whole vertical SC performs efficiently with lower inventory level.
In this section we define the types of demand information management based on which information is used in each system. We also identify two demand informationsharing methods—the planned demand transferring method (PDTM) and the forecasted demand distributing method (FDDM).
Table 1 Supply chain community strategies
Supply chain structure Main objective Degree of freedom Easiness of cooperation Controlling party Optimization type
Execution oriented community strategy
Vertical inclined community strategy
Strategic coordinated community strategy
Early stage horizontal integrated SC Own profit and efficiency High Low None Local optimum
Oligopolized horizontal integrated SC Smooth material flow Low Medium Oligopolized upstream player Near global optimum
Autonomous decentralized SC Efficiency of entire SC Medium High None Global optimum
ARTICLE IN PRESS S.-J. Ryu et al. / Int. J. Production Economics 120 (2009) 162–175
165
Expected Demand Market Demand
Planned Demand Realized Demand
Past
Near Future
Far Future
0
t
Fig. 1. Three zones of demand information.
Table 2 Characteristics of demand information
Target period Source information Accuracy Vulnerability Shared form Sharing method Sharing players
Realized demand information
Planned demand information
Expected demand information
Past Sales result Poor None Sales record N/A N/A
Near future Forecast Best Little Planned requirement Transferring Between two players
Far future Forecast Good High Forecasted demand Transferring or distributing Two players or more
Forecast
Forecast
Forecast
Planning Layer
Production Plan
Production Plan
Sales Plan
Execution Layer
Manufacture
Manufacture
Sale
Supplier
Maker
Market
Retailer
Fig. 2. Information flow of IDIM.
4.1. Isolated demand information management (IDIM) Each player in the SC creates its own forecasted demand information based on its sales record, and employs this in the planning and execution process. In this method there is no way of knowing the status of adjacent players. Furthermore, upstream players of SC have to make plans and execute production operations without knowledge of market demand (Fig. 2).
(Strength) If there is strong correlation between the sales record and future demand, it is possible to maintain high forecasting accuracy. (Weakness) Upstream players often over- or underproduce because there is no method of coordination between players. High variability in market demand exacerbates this. Each player has to maintain high inventory levels to handle the problem. The entire SC is quite vulnerable to market demand change.
ARTICLE IN PRESS 166
S.-J. Ryu et al. / Int. J. Production Economics 120 (2009) 162–175
4.2. Information-sharing methods Two information-sharing methods exist. One is PDTM and the other is FDDM. 4.2.1. Planned demand transferring method It is the retailer who creates forecast information about market demand in PDTM. After creating a procurement plan by subtracting on-hand inventory and scheduled receipt, it transfers the procurement plan to maker, the upstream player assembling final products and supplying retailer with them. The maker then creates its own production plan based on the information received from retailer. In the same way the maker decides the amount of purchasing from the supplier considering inventory conditions and then transmits the procurement plan to the supplier. In this method, the accuracy of shared information worsens, because time is needed for each player to create a plan. Nevertheless, it can be presumed that the level of consistency for planning results between adjacent players is quite high (Fig. 3). (Strength) Because upstream players make production plans based on their downstream player’s planned requirements, they can better manage downstream purchase orders. As a result, upstream players can keep their inventory level low by preventing unnecessary production. The service level remains high between two adjacent players. (Weakness) The retailer has to consider all upstream player’s planning and execution lead-time when making forecasting market demand. Therefore the accuracy of forecasted demand information worsens. If more tiers are introduced into the SC, accuracy falls sharply. 4.2.2. Forecasted demand distributing method In the FDDM, a third party organization (net demand requirement computation organization) is responsible for forecasting demand. This organization considers each player’s inventory status and lead-time, then distributes it to each player. As an example, the forecasted demand information that suppliers receive consists of the net expected requirement when all downstream inventory is subtracted from the forecasted market demand.
Forecasted Demand
Planning Layer
Execution Layer
In this method there is no time lost while downstream players make their own procurement plan. This means that the level of accuracy is better with the PDTM. However, it is also assumed that the level of consistency for planning results between adjacent players is worse than in the FDDM. The delivery lead-times to the market are, however, different with each player, and the accuracy of the forecasted demand information for each player has some level of discrepancy. For instance, the accuracy of supplier’s shared information is worse than that of his downstream players (Fig. 4). (Strength) Because it is possible to compile demand forecast information for the near future, the accuracy of shared information is to be higher than ‘PDTM’. Also, the service level to the market in ‘FDDM’ is assumed to be kept higher than in ‘PDTM’, and it is assumed that the average inventory of whole SC can be maintained at a low level. (Weakness) Because the delivery lead-time to the market is different with each player, the accuracy of the forecasted demand information for each player has some level of discrepancy. Therefore, it is assumed that the service level between adjacent players is worse than with the PDTM. We summarize the strengths and weaknesses of both demand information-sharing methods in Table 3 below. 5. Experiment design In this section, we will discuss experiment design of this research. 5.1. Simulation model 5.1.1. Market demand We define market demand using sine curve as Fig. 5. MarketDemand ¼ InitialDemand þ DemandIncreaseRate ðTimeÞ þ DemandVariationRange Time Sin 2p VariationCycle
Forecasted Demand
InitialDemand: market demand when Time ¼ 0.
Forecast
Production Plan
Sales Plan
Manufacture
Manufacture
Sale
Supplier
Maker
Retailer
Production Plan
Fig. 3. Information flow of PDTM.
Market
ARTICLE IN PRESS S.-J. Ryu et al. / Int. J. Production Economics 120 (2009) 162–175
Demand Requirement Calculation Organization
Forecasted Demand
Forecasted Demand Planning Layer
Production Plan
Execution Layer
Manufacture
Forecast
Forecasted Demand
Production Plan
Sales Plan
Manufacture
Sale
Maker
Retailer
Supplier
167
Market
Fig. 4. Information flow of FDDM.
Table 3 Strengths and weaknesses of DIS* methods Planned demand transferring method Strength
Low inventory level
Forecasted demand distributing method
Good service level to
between players
Good service level between two adjacent players of SC Weakness
Poor service level to market demand due to bad forecasting accuracy
market demand owing to high forecasting accuracy Well-distributed pipeline inventory
Low service level between two adjacent players
(DIS*: Demand information sharing).
DemandIncreaseRate: slope of demand change (indicate the pattern of change). DemandVariationRange: degree of demand variation. VariationCycle: time interval between demand variations. We assign the value of 100 to InitialDemand. And, DemandIncreaseRate, DemandVariationRange, VariationCycle will be used as variables of simulation to compare SC performance according to the different informationsharing types. 5.1.2. Operations of player 5.1.2.1. Demand information generation. We have the same logic of creating forecasted demand information between both of the information-sharing methods. We express it as following. ForecastedDemand ¼ ð1 ForecastingErrorÞ ActualDemand
ForecastingError: forecasting error between 0 and 1. ActualDemand: actual market demand to be realized. The ForecastingError increases when it comes to forecasting demand further in the future. Therefore, forecasted demand information is different according to the information-sharing method. Ltit(A,B): Information transferring lead-time from downstream player A to upstream player B LTp(A): planning lead-time of player A
Lte(A): execution lead-time of player A LTd(A,B): delivery lead-time from upstream player A to downstream player B 5.1.2.2. Production and procurement plan generation. We describe the details of plan generation schemes (see Figs. 6–8). 1. Supplier: The supplier creates its own production plan considering forecasted demand and inventory status. Production capacity plays a role as a constraint. Details of the plan generation logic are as follows: ProductionPlanðSÞ ! DemandForecatðSÞ ComponentInventoryðSÞ; ¼ min ProductionCapacityðSÞ 2. Maker: The maker generates his own production plan based on forecasted demand and inventory status. The capacity will be a constraint. The maker then calculates part requirements taking into account the production plan and part inventory when transferring procurement plan to the supplier, or upstream player. ProductionPlanðMÞ ! DemandForecatðMÞ ProdInventoryðMÞ; ¼ min ProductionCapacityðMÞ ProcurementPlanðMÞ ¼ ProductionPlanðMÞ ComponentInventoryðMÞ ProdInventory(M): product inventory of maker PartInventory(M): part inventory of maker 3. Retailer: The retailer draws up his own sales plan based on forecasted demand and inventory status. The sales capacity will be a constraint during the calculation. The retailer calculates product requirement considering the sales plan and product inventory when transferring procurement plan to the maker.
SalesPlanðRÞ ¼ min
DemandForecatðRÞ FinalGoodsInventory;
!
SalesCapacityðRÞ
ProcurementPlanðRÞ ¼ SalesPlanðRÞ ProductInventoryðRÞ
ARTICLE IN PRESS 168
S.-J. Ryu et al. / Int. J. Production Economics 120 (2009) 162–175
Market Demand
Intial Demand
Increase Rate
Variation Range
Cycle
0
t Fig. 5. Market demand model.
Production Capacity Forecasted Demand Production Plan
Plan Generation Component Inventory Fig. 6. Supplier’s plan generation.
Production Capacity Forecasted Demand
Plan Generation
Product Inventory
Production Plan Component Inventory
Plan Generation
Procurement Plan
Plan Generation
Procurement Plan
Fig. 7. Maker’s plan generation.
sales Capacity Sales Plan
Forecasted Demand Plan Generation Final-Goods Inventory
Product Inventory
Fig. 8. Retailer’s plan generation.
ARTICLE IN PRESS S.-J. Ryu et al. / Int. J. Production Economics 120 (2009) 162–175
Input Instruction
Shipment Instruction
Semi-Product Inventory Management
Manufacture
Shipment
Supplier
Purchase Instruction
Input Instruction
Part Inventory Management
Procure
Shipment Instruction
Product Inventory Management
Manufacture
Shipment
Maker
169
Purchase Instruction
Shipment Instruction
Fina-Goods Inventory Management
Procure
Shipment Retailer
Fig. 9. Player’s behavior model.
5.1.2.3. Instruction and execution. Fig. 9 shows how each player’s instruction and execution works within the proposed model. Instruction and execution is performed according to the generated by each player. 5.2. Analysis on SC performance We begin by specifying the analysis points for SC performance. 5.2.1. Control characteristics Here we lay out the analysis points relating to control characteristics. Analysis point 1: the strengths and weaknesses of information-sharing methods can be identified clearly when there is no change or variation in market demand. With neither change nor variation, it is the forecasting error of demand information which affects the performance of a SC.
PDTM: the level of consistency for planning results
between adjacent players is quite high. Therefore the service level between them is assumed to be high and the inventory level of whole SC can be kept low. However, players may have excess inventory as the accuracy of demand forecasts created by retailer is low. FDDM: it is possible to attain high accuracy in demand forecasts, so it is assumed that the service level to market will be high, and with less chance of having excess inventory. However, the inventory level of the whole SC is higher than in PDTM.
Analysis point 2: when there is an increasing trend in market demand, the discrepancy between forecasted and actual demand increases with both information-sharing methods. Thus, the service level worsens with both methods.
demand. This is because a high level of consistency in planning results between adjacent players is maintained. FDDM: the gap between forecasted and actual demand becomes larger for upstream players. Therefore, it is assumed that the inventory level, especially that of the supplier, increases.
Analysis point 3: when there is a decreasing trend in market demand, the discrepancy between forecasted and actual demand decreases with both information-sharing methods. Therefore it can be assumed that the service level to market demand and the service level between players improve with both methods. Furthermore, a low inventory level can be maintained throughout the whole SC. Analysis point 4: the gap between forecasted and actual demand grows when market demand is variable. Therefore, the service level to market demand is slightly worse than when there is no variability in market demand. Analysis point 5: the service level to market demand is kept low when there are constraints on each player’s execution capacity. This lowers the inventory level of both the SC as a whole, and that of each player within it. However, the service level between two adjacent players should improve when there is no constraint. Analysis point 6: when a player’s lead-time is shorten, the number of gaps between forecasted and actual market demand decreases. Therefore, the service level to market improves with both information-sharing methods. 5.2.2. Responsiveness to environmental change Analysis point 7: when market demand rises unexpectedly, the inventory level of each player and the whole SC increases. However, influence on the service level to market is assumed to be different with differing information-sharing methods.
PDTM: the inventory level of the whole SC and the service level between supplier and maker can be maintained when there is no change in market
PDTM: the discrepancy between the retailer’s forecasted demand and actual market demand grows.
ARTICLE IN PRESS 170
S.-J. Ryu et al. / Int. J. Production Economics 120 (2009) 162–175
Therefore the service level to market declines sharp, and it takes time to restore the previous service level. FDTM: at first, the service level to market worsens due to the sudden change of market demand. Prior service level, however, is soon restored.
5.3. Parameter setting We specify the parameters to verify the analysis points discussed in previous section. The forecasting error used in the simulation is depicted in Fig. 10. It is manipulated by four settings, which differ with their increased rate of forecasting errors such as high, medium, low, and no forecasting error. The other parameters are depicted in Table 4 below. Forecasting error is set at four levels; ‘high forecasting error’, ‘medium forecasting error’, ‘low forecasting error’ and ‘no forecasting error’. They are different with increased error across the time period. Demand variation is manipulated with four alternatives. They are ‘high demand variability’, ‘medium de-
mand variability’, ‘low demand variability’, and ‘no demand variability’. Execution lead-time has two values, which are ‘long lead-time’ and ‘short lead-time’. For instance, execution lead-time in ‘long lead-time’ is two times longer than planning lead-time. Finally, capacity tightness consists of either ‘high capacity tightness’ or ‘low capacity tightness’. For instance, execution (production or sales) capacity of ‘high capacity tightness’ is 1.1 times larger than average demand.
5.4. Evaluation index 5.4.1. Control characteristics We explain the evaluation index in this section, beginning with the evaluation index for control characteristics as depicted in Table 5. Throughput is the amount of shipment to an adjacent downstream player. The throughput of the whole SC is equivalent to that of the retailer.
Forecasting Error (%)
50 40 30 20 10
0
t-1
t
t+1
Period
Fig. 10. Forecasting error.
Table 4 Parameter setting Parameter
Manipulation
Detail
Forecasting error
HFE (high error) MFE (medium error) LFE (low error) NFE (no error)
IncreasedError ¼ 3*UnitIncreasedError IncreasedError ¼ 2*UnitIncreasedError IncreasedError ¼ 1*UnitIncreasedError IncreasedError ¼ 0*UnitIncreasedError
Demand variation
HDV (high variability) MDV (medium variability) LDV (low variability) NDV (no variability)
0.5*(average demand)pDVp0.5*(average demand) 0.3*(average demand)pDVp0.3*(average demand) 0.1*(average demand)pDVp0.1*(average demand) DV (average demand) ¼ 0
Execution lead-time
LLT (long lead-time) SLT (short lead-time)
LTe(A) ¼ 2*LTp(A) LTe(A) ¼ 1*LTp(A)
Capacity tightness
HCT (high tightness) LCT (low tightness)
Execution capacity ¼ 1.1*(average demand) Execution capacity ¼ 1.5*(average demand)
ARTICLE IN PRESS S.-J. Ryu et al. / Int. J. Production Economics 120 (2009) 162–175
171
Table 5 Evaluation index Throughput (TH)
Order fill rate (OFR)
Inventory level
Supplier Maker Retailer Whole SC
Throughput to maker Throughput to retailer Throughput to market Retailer’s throughput
OFR to maker OFR to retailer OFR to market Retailer’s OFR
Inventory (outflow) Inventory (inflow, outflow) Inventory (inflow, outflow) Pipeline inventory
SC Performance
SC player
Resdegree
Restime
0
Occurrence of Environmental Change
Restoration of Previous Status
t
Fig. 11. Evaluation index for responsiveness.
We also define order fill rate as service level. Order fill rate is computed by the equation as below. ThroughputUpstream 100 OrderFillRateð%Þ ¼ RequirementDownstream ThroughputUpstream: throughput of upstream player. RequirementDownstream: requirement from downstream player. Order fill rate refers to how well an upstream player serves the requirements of the downstream player. Similarly with throughput, the service level to market demand is equivalent to the retailer’s order fill rate. Finally, inventory level is an index of control characteristics. It covers parts inventory, product inventory, and final-goods inventory. We also take the WIP (work in process) inventory into consideration. The whole SC inventory relates to the pipeline inventory over the entire SC.
5.4.2. Responsiveness to environmental change To evaluate the responsiveness to sudden change, we define the index as depicted in Fig. 11. This shows the degree of damage from sudden change and the time taken to restore previous status after a sudden environmental change.
6. Analysis of results In the following sections we analyze the result of simulation based on simulator developed with STELLA.
6.1. Control characteristics Fig. 12 shows the simulation result of pipeline inventory level of SC at the four degrees of forecasting error. When there is no error in forecasting, PDTM maintains lower inventory level than FDDM. However, if there is error in forecasting, FDDM performs better than the PDTM. As the forecasting error increases, the difference in performance between two methods grows. FDDM has higher forecasting accuracy than PDTM. Fig. 13 shows details of inventory level over SC when there is high forecasting error. We identified that PDTM still performs well in terms of inventory level at the site of maker, although FDTM has lower inventory levels than PDTM at the other points along the SC. Fig. 14 shows the simulation result of throughput at degree of demand variability.
ARTICLE IN PRESS 172
S.-J. Ryu et al. / Int. J. Production Economics 120 (2009) 162–175
Inventory Level (unit)
300 250 200 150 100 PDTM
50
FDDM
0 No Error
Low Error
Medium Error
High Error
Fig. 12. Inventory level at degree of forecasting error.
Inventory Level (unit)
300 Inventory of supplier 250 Inventory of maker 200
Inventory of retailer
150
Inventory of SC
100 50 0 PDTM
FDDM
Fig. 13. Inventory level at high forecasting error.
Throughput (unit)
120 100 80 60 40
PDTM
FDDM
20 0 No Variability
Low Variability
Medium Variablity
High Variability
Fig. 14. Throughput at degree of demand variability.
Inventory Level (unit)
300 250 200 150 100 PDTM
50
FDDM
0 No Variability
Low Variability
Medium Variability
High Variability
Fig. 15. Inventory level at degree of demand variability.
In terms of throughput, the FDDM performs better than PDTM. As demand variability grows, the performance gap increases. Fig. 15 shows the inventory level at various degree of demand variability.
PDTM maintains lower inventory level than FDDM. However, when the variability becomes larger, FDDM performs better than PDTM. Fig. 16 shows the simulation results of service level at both ‘long’ and ‘short execution lead-time’.
ARTICLE IN PRESS S.-J. Ryu et al. / Int. J. Production Economics 120 (2009) 162–175
When the execution lead-time shortens, the service level to market demand rises sharply in PDTM. This is why execution lead-time has greater influence on service level to market in PDTM than in FDDM. Fig. 17 shows the simulation results of OFR (order fill rate) to market both ‘long’ and ‘short’ execution lead-time at various degrees of execution lead-time.
PDTM
Service Level (%)
The service level to market declines sharply when the execution lead-time lengthens. 6.2. Responsiveness Although there is a sudden demand increase, there are lesser effects on OFR to market in FDTM than in PDTM.
FDDM Short Lead-Time
Long Lead-Time
90
173
80 70 60 50 40 Market OFR
S-M OFR
Market OFR
S-M OFR
Fig. 16. Service level at degree of execution lead-time.
85
OFR to Market (%)
80 75 70 65 60 55
FDDM
PDTM
50 45 1
2
3 4 Execution Lead-Time
5
6
Fig. 17. Order fill rate at degree of execution lead-time.
OFR to Market (%)
100 75 50 7 weeks 25 PDTM
FDDM
13 weeks
0 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 weeks Fig. 18. Responsiveness to a sudden demand increase.
ARTICLE IN PRESS 174
S.-J. Ryu et al. / Int. J. Production Economics 120 (2009) 162–175
It also takes about half as long to obtain previous service level with FDTM (Fig. 18).
6.3. Feasibility study on electronics industry example What follows is a feasibility study on the application of FDTM to the electronics equipments industry in which FDTM is thought to perform better than PDTM.
Table 6 Planning parameter Layer
Planning cycle
Bucket
Span
Planning Execution
4 weeks 1 week
4 weeks 1 week
(S)20(R)36 weeks 4 weeks
The subjective SC is designed as a three-tier SC producing a single product. The parameters used in feasibility study are established below in Tables 6 and 7. The planning cycle and the planning bucket are both 4 weeks. And the planning span varies—20 weeks to 36 weeks by player. In execution layer, both the planning cycle and the planning bucket are 1 week, and the span of planning is 4 weeks. There is a strong tendency to predict aggressively for future market demand in the high-tech electronics industry. Companies are inclined to forecast optimistically for the sake of the market growth rate and increasing their own market share; and the accuracy of forecasting worsens as the target period becomes longer. Figs. 19 and 20 shows the feasibility study result. We verified that there is significant improvement of 23% in pipeline inventory level for FDTM, although there is
Table 7 Forecasting accuracy Period
1st mo.
2nd mo.
3rd mo.
4th mo.
5th mo.
6th mo.
7th mo.
8th mo.
9th mo.
Accuracy (%)
95
90
80
70
60
50
60
70
70
74
OFR to Market (%)
72 2%
70 68 66 64 62 60 PDTM
FDDM
Fig. 19. ORF to market for each method.
Pipeline Inventory of SC
700 600
23 %
500 400 300 200 100 0 PDTM Fig. 20. Pipeline inventory for each method.
FDDM
ARTICLE IN PRESS S.-J. Ryu et al. / Int. J. Production Economics 120 (2009) 162–175
little improvement in OFR (about 2%) between information-sharing methods. 7. Conclusions In this research, we introduced SC community strategies and the types of demand information management closely related to information-sharing methods. And we define two different information-sharing methods according to types of shared information and sharing procedures. One is the ‘planned demand transferring method (PDTM)’ and the other is the ‘forecasted demand distributing method (FDDM)’. We also analyzed control characteristics and responsiveness to compare the SC performance between the two information-sharing methods. We then evaluated the SC performance in terms of throughput, inventory level, and service level. We verify that FDDM shows better performance than PDTM in terms of throughput. Even when there is high forecasting error or high demand variability, FDDM maintain lower inventory levels than PDTM. However, if there is little demand variability, PDTM performs better than FDDM. When the execution lead-time shortens, the service level to market demand rises sharply in PDTM. The execution lead-time also has greater influence on service level to market in the ‘PDTM’ than in the ‘FDDM’. In the case of a sudden demand increase, there is less effect on OFR to market in FDDM than in PDTM. It also takes about half as long to restore previous service level in FDDM when compared to PDDM. Finally, we conducted a feasibility study to an electronics industry example. We verify that FDDM shows a significant improvement (23%) in pipeline inventory level for the entire SC, although there is little improvement in OFR (about 2%) between information-sharing methods. 8. Future work In this study, we have compared the SC performance of information-sharing methods with various environmental scenarios. We verified that FDDM is a valuable tool throughout the SC, including the improvement of inventory levels and service level to market demand. However, it should be noted that PDTM is also valid in terms of service level between adjacent players and inventory level in the case of little demand variability. Also, we found that if there is a sudden change in market demand, the consequences of the effect are quite different with information-sharing methods. From this result we can expect that the new information-sharing method, which captures the strengths of both methods, can perform better. As a last note, this study did not model the SC with cash flow, which means that this study lacks evaluation of costs such as information-sharing costs and SC configuration costs. These will be the subjects of our future research.
175
References Aviv, Y., 2001. The effect of collaborative forecasting on supply chain performance. Management Science 47 (10), 1326–1343. Bourland, K.E., Powell, S.G., Pykenm, D.F., 1996. Exploring timely demand information to reduce inventories. European Journal of Operational Research 92, 239–253. Byrne, P.J., Heavey, C., 2006. The impact of information sharing and forecasting in capacitated industrial supply chains: A case study. International Journal of Production Economics 103, 420–437. Cachon, G.P., Lariviere, M.A., 2001. Contracting to assure supply: How to share demand forecasts in a supply chain. Management Science 47 (5), 629–646. Chen, F., 1998. Echelon reorder points, installation reorder points, and the value of centralized demand information. Management Science 44 (12, part2), 221–234. Chen, F., 1999. Decentralized supply chains subject to information delay. Management Science 45 (8), 1076–1090. Chen, F., Yu, B., 2005. Quantifying the value of leadtime information in a single-location inventory system. Manufacturing & Service Operations Management 7 (2), 144–151. Chen, H., Chen, J., Chen, Y., 2006. A coordination mechanism for a supply chain with demand information updating. International Journal of Production Economics 103, 347–361. Dejonckheere, J., Disney, S.M., Lambrecht, M.R., Towill, D.R., 2004. The impact of information enrichment on the bullwhip effect in supply chains: A control engineering perspective. European Journal of Operational Research 153, 727–750. Gaur, V., Giloni, A., Seshadri, S., 2005. Information sharing in a supply chain under ARMA demand. Management Science 51 (6), 961–969. Gavirneni, S., Kapusucinski, R., Tayur, S., 1999. Value of information in capacitated supply chains. Management Science 45 (1), 16–24. Holweg, M., Disney, S., Holmstrom, J., Smaros, J., 2005. Supply chain collaboration: Making sense of the strategy continuum. European Management Journal 23 (2), 170–181. Jain, A., Moinzadeh, K., 2005. A supply chain model with reverse information exchange. Manufacturing & Service Operations Management 7 (4), 360–378. Kulp, S.C., Lee, H.L., Ofek, E., 2004. Manufacturer benefits from information integration with retail customers. Management Science 50 (4), 431–444. Lee, H.L., Padmanabhan, V., Whang, S., 1997. Information distortion in a supply chain: The bullwhip effect. Management Science 43 (4), 546–558. Lee, H.L., So, K.C., Tang, C.S., 2000. The value of information sharing in a two-level supply chain. Management Science 46 (5), 626–643. Li, L., 2002. Information sharing in a supply chain with horizontal competition. Management Science 48 (9), 1196–1212. Li, G., Lin, Y., Wang, S., Yan, H., 2006. Enhancing agility by timely sharing of supply information. Supply Chain Management: An International Journal 11 (5), 425–435. Tang, C.S., 2006. Perspectives in supply chain risk management. International Journal of Production Economics 103, 451–488. Watson, N., Zheng, Y.-S., 2005. Decentralized serial supply chains subject to order delays and information distortion: Exploiting real-time sales data. Manufacturing & Service Operations Management 7 (2), 152–168. Welker, G.A., van der Vaart, T., Pieter van Donk, D., 2008. The influence of business conditions on supply chain information-sharing mechanisms: A study among supply chain links of SMEs. International Journal of Production Economics 113, 706–720. Yao, D.-Q., Yue, X., Wang, X., Liu, J.J., 2005. The impact of information sharing on a returns policy with the addition of a direct channel. International Journal of Production Economics 97, 196–209. Yao, D.-Q., Kurata, H., Mukhopadhyay, S.K., 2008. Incentives to reliable order fulfillment for an Internet drop-shipping supply chain. International Journal of Production Economics 113, 324–334. Yue, X., Liu, J., 2006. Demand forecast sharing in a dual-channel supply channel. European Journal of Operational Research 174, 646–667. Zhao, X., Xie, J., Leung, J., 2002. The impact of forecasting model selection on the value of information sharing in a supply chain. European Journal of Operational Research 142, 321–344.