performance measures and performance models for supply...

97Performance Measures and Performance Models for...

Performance Measures and PerformanceModels for Supply Chain

Decision Making

Numerous models, algorithms, and tools have been deployedin supply chain modeling and decision making. These arebased on stochastic models, optimization models, objectoriented models, and simulation. This paper provides anexpository introduction to decisions, performance measures,mathematical models, and software models in supply chainnetworks.

Y Narahari and Shantanu Biswas

Source: © Y Narahari and Shantanu Biswas. Reprinted with permission.

7

1 Introduction

Businesses today operate in a very tough environment that is constantly in flux[46, 57]. Customers have become increasingly demanding looking for better andinnovative goods and services that are specifically customized to meet their uniqueneeds. There is also an implicit requirement on the accuracy, timeliness,convenience, responsiveness, quality and reliability of the service offered to them.And all of this is desired at ever-lower prices. Simultaneously, the rapid pace ofinnovation has resulted in shorter product and technology cycles, leading touncertainties in supply and demand. An example of this trend is the PC industry,

98 SUPPLY CHAIN PERFORMANCE MEASUREMENT AND IMPROVEMENT

where new models are introduced every 6 to 9 months under intense competitivepressure on cost, functionality and service. All these challenges ultimately resultin managers asking for an outstanding design of the underlying supply chainwhich happens to be the backbone of any business company.

A supply chain is a network of supplier, manufacturing, assembly, distribution,and logistics facilities that perform the functions of procurement of materials,transformation of these materials into intermediate and finished products, andthe distribution of these finished products to customers. Supply chains are nowat the centrestage of business performance of manufacturing and service enterprises.The supply chain process is a complex, composite business process comprising ahierarchy of different levels of value-delivering business processes. Designing ahigh performance supply chain is a very challenging task due to the complexstructure of the supply chains and the ever changing business. Some of theimportant reasons for the complexity of the decision making process are:

• Large scale nature of the supply chain networks

• Hierarchical structure of decisions

• Randomness of various inputs and operations

• Dynamic nature of interactions among supply chain elements

Because of the inherent complexity of decision making in supply chains, thereis a growing need for modeling methodologies that can help identify and innovatestrategies for designing high performance supply chain networks. A large numberof manufacturing and service organizations are therefore seeking modeling systemsthat can help identify and implement strategies for designing and improvingtheir supply chain networks. This article provides a first level introduction todecision making in supply chains and succinctly describes the principal modelingapproaches.

1.1 Outline of the Paper

In Section 2, we discuss different aspects of supply chain decision making. Weclassify supply chain decisions according to the time horizon of decisions andalso according to functionality. We provide a representative listing of thesedecisions.


In Section 3, we describe different measures of supply chain performanceunder two main categories: financial and non-financial.

Section 4 is devoted to a discussion of performance modeling approaches tosupply chain networks. The models discussed are: optimization models, analyticalmodels, and simulation models.

In Section 5, we present an object oriented modeling approach for supplychain networks. We first describe various objects of an object library. The objectsbelong to two categories: structural objects and policy objects. We have usedUML (Unified Modeling Language [6]) for creating generic object models ofsupply chain elements. We provide an example of a LPG (Liquid Petroleum gas)supply chain to illustrate our object modeling approach.

Section 6 concludes the paper and provides pointers to important currentliterature.

2 Decisions in Supply Chains

Supply chain decisions can be classified based on temporal and functionalconsiderations.

2.1 A Temporal Classification

Supply chain decisions can be broadly classified into three categories: strategic,tactical, and operational according to the time horizon of the decisions.

• Strategic decisions target long-term objectives of a supply chain and guidethe supply chain policies from a design and planning perspective. Typicallythese decisions are not reviewed before a time horizon of a few to severalyears expires but the time horizon depends on a variety of factors.

• Tactical decisions are the decisions that are required to effectively managethe supply chains configured according to strategic level decisions. Thetime intervals of tactical decisions could range from weeks to months.

• Operational decisions are short-term decisions and are generally focusedon the real-time activities of a supply chain.


2.2 A Functional Classification

Functionally, there are four major decision areas in supply chain management:procurement, manufacturing, distribution, and logistics. In addition, there arealso certain global decisions whose scope extends over multiple functions. Thereare strategic, tactical, and operational questions in each of these areas. We providea list of important decisions here.

2.2.1 Procurement Decisions

• Supplier Selection: Where should we source raw materials, components,and sub-assemblies from? (strategic)

• Direct Delivery from Suppliers: Can intermediate warehouses be eliminatedso that materials can be delivered directly into manufacturing plants?(strategic)

• Vendor Managed Inventories: Should the inventories at the plants bevendor-managed? (strategic)

• Optimal Procurement Policy: What are the cost and service trade-offs inalternative procurement strategies? (strategic/tactical)

2.2.2 Manufacturing Decisions

• Plant Location: How many manufacturing plants should be set up andwhere should they be located? (strategic)

• Product Line Selection: What products should be produced at eachmanufacturing location? (strategic/tactical)

• Capacity Planning: How much capacity is needed in each plant? (strategic)

• Capacity Allocation: How do we allocate plant capacity to products?(tactical)

• Inventory Decisions: What raw materials/WIP/finished goods inventoryshould be stocked in each center? (strategic/tactical)

• Optimal Manufacturing Strategy: What are the cost and service trade-offsin alternative manufacturing strategies? For example, what would be thebest policy among make-to-stock, make-to-order, and assemble-to-order?(strategic/tactical)


• Input Control: How do we introduce work into the plant? (strategic/tactical/operational)

• Production Scheduling: How do we schedule the production to maximizethroughput and minimize cycle time? (tactical/operational)

• Constrained Supply: How do we optimize resource utilization when thesupplies are not enough to fulfill the requirements? (tactical)

2.2.3 Distribution Decisions

• Configuration of Distribution Facilities: What types of distribution centersare required? (Warehouses, cross-docks, drop-lots, etc.) (strategic/tactical)

• Location: Where should distribution facilities be located? (strategic)

• Customer Allocation: Which customers should a facility service? (strategic/tactical)

• Facility Configuration: What product should be handled by each facility?What products and in how many quantities should be stocked at eachfacility? What should be the replenishment strategy? (strategic/tactical)

• Optimal Distribution Strategy: What are the cost and service trade-offs ofalternative distribution strategies? (tactical)

2.2.4 Logistics Decisions

• Logistics Mode Selection: What transport modes and lanes should be usedto move products throughout the network? (strategic)

• Selection of Ports: Which ports should be used to bring product into andout of a country? (strategic)

• Direct Delivery: Which products should move directly from manufacturingcenters to customers? (strategic/tactical)

• Optimal Transportation Strategy: What are the cost and service trade-offsof alternative transportation strategies? (tactical)

2.2.5 Global Decisions

• Product and Process Selection: What product quantities, by facility, byprocess, should be produced and stored in each period to support customer


demands? What products to sell and to which customers to maximize profits?(tactical)

• Planning Under Uncertainty: What are the implications associated withseasonal or cyclical demand, capacity availability, cost fluctuations, or rawmaterial availability? (tactical/operational)

• Global Optimization of Operations: What are the cost and service trade-offs among procurement, manufacturing, distribution, and logisticsalternative strategies? (strategic/tactical)

• Real-time Monitoring and Control: How can the orders be dynamicallyrouted and scheduled through the supply chain in reaction to occurrenceof real-time events? (operational)

3 Supply Chain Performance Measures

Supply chain performance measures can be classified broadly into two categories[55]: qualitative measures (such as customer satisfaction and product quality)and quantitative measures (such as order-to-delivery lead time, supply chainresponse time, flexibility, resource utilization, delivery performance, etc.). In ourstudy we consider only the quantitative performance measures.

Quantitative metrics of supply chain performance can be classified into twobroad categories: Non-financial and financial.

3.1 Non-Financial Performance Measures

Important metrics include: cycle time, customer service level, inventory levels,resource utilization, performa-bility, flexibility, and quality. There is a detaileddiscussion of these in [55]. We will focus here on the first four measures.

3.1.1 Cycle Time

Cycle time or lead time is the end-to-end delay in a business process. For supplychains, the business processes of interest are the supply chain process and theorder-to-delivery process. Correspondingly, we need to consider two types oflead times: supply chain lead time and order-to-delivery lead time. The order-to-delivery lead time is the time elapsed between the placement of order by a customerand the delivery of products to the customer. If the items are in stock, then it


would be equal to the distribution lead time and order management time. If theitems are made to order, then this would be the sum of supplier lead time,manufacturing lead time, distribution lead time, and order management time.The supply chain process lead time is the time spent by the supply chain toconvert the raw materials into final products plus the time needed to reach theproducts to the customer. It thus includes supplier lead time, manufacturinglead time, distribution lead time, and the logistics lead time for transport of rawmaterials from suppliers to plants and for transport of semi-finished/finishedproducts in and out of intermediate storage points. Lead time in supply chains isdominated by the interface delays due to the interfaces between suppliers andmanufacturing plants; between plants and warehouses; between distributors andretailers, etc. Lead time compression is an extremely important topic because oftime based competition and the correlation of lead time with inventory levels,costs, and customer service levels.

3.1.2 Customer Service Level

Customer service level in a supply chain is a function of several differentperformance indices. The first one is the order full rate, which is the fraction ofcustomer demands that are met from stock. For this fraction of customer orders,there is no need to consider the supplier lead times and the manufacturing leadtimes. Another measure is the backorder level, which is the number of orderswaiting to be filled. To maximize customer service level, one needs to maximizeorder full rate, and minimize backorder levels. Another measure is the probabilityof on-time delivery, which is the fraction of customer orders that are fulfilled on-time, i.e., within the agreed-upon due date.

3.1.3 Inventory Levels

Since inventory carrying costs can contribute significantly to total costs, thereis a need to carry just about enough inventory to satisfy the customer demands.Inventories held in a supply chain belong to four categories: Raw materials,work-in-process (unfinished and semi-finished parts), finished goods inventory,and spare parts. Each type of inventory is held for different reasons and there is aneed to keep optimal levels of each type of inventory. Thus measuring the actualinventory levels will provide a useful picture of system efficiency.


3.1.4 Resource Utilization

A supply chain network uses resources of various kinds: manufacturing resources(machines, material handlers, tools, etc.); storage resources (warehouses, automatedstorage and retrieval systems); logistics resources (trucks, rail transport, air-cargocarriers, etc.); human resources (labor, scientific and technical personnel); andfinancial (working capital, stocks, etc.). The objective is to utilize these assets orresources efficiently so as to maximize customer service levels, minimize lead times,and optimize inventory levels.

3.2 Financial Measures

There are several fixed and operational costs associated with a supply chain.Ultimately, the aim is to maximize the revenue by keeping the supply chain costslow. Costs arise due to inventories, transportation, facilities, operations, technology,materials, and labor. The financial performance of a supply chain can be evaluatedby looking into the following items [2]:

• cost of raw material

• revenue from goods sold

• activity-based costs such as material handling, manufacturing, assembling, etc.

• inventory holding costs

• transportation costs

• cost of expired perishable goods

• penalties for incorrectly lled or late orders delivered to customers

• credits for incorrectly lled or late deliveries from suppliers

• cost of goods returned by customers

• credits for goods returned to suppliers

Typically, the financial performance indices can be put together using thefollowing major modules: activity based costing, inventory costing, transportationcosting, and inter-company financial transactions.


4 Performance Models of Supply Chains

A research summary of various quantitative models for supply chains is providedin [54]. These models can be broadly classified into optimization models,analytical performance models, and simulation models. The tools and techniquesused for optimization in industry for solving supply chain problems are discussedby Hicks [26, 27].

4.1 Optimization Models

A major portion of the supply chain literature consists of multi-echelon inventorycontrol models. A comprehensive review of these models can be found in Vollmanet al [59]. These methods generally deal with operational or tactical/operationallevels. Multi-echelon inventory models have been successfully implemented inindustry. Cohen et al [10] describe OPTIMIZER: IBM’s Multi-Echelon InventorySystem for Managing Service Logistics. They develop efficient algorithms anddata structures to achieve large scale systems integration. Ettl et al [15] considera supply network model to generate base stock levels at each store so as to minimizethe overall inventory capital and guarantee the customer service requirements.

The other major focus area of supply chain optimization models is to determinethe location of production, warehousing, and sourcing facilities, and the pathsthe products take through them. These methods provide models mostly forstrategic and strategic/tactical levels. One of the earliest works in this area is byGeoffrion and Graves [24]. They describe a mixed integer programming modelfor determining the location of distribution facilities. Cohen and Lee [11, 12]consider global manufacturing and distribution networks and formulate mixedinteger optimization programs. Lee and Billington [35] validate these models byapplying it to analyze the global manufacturing strategies of Hewlett-Packard.Arntzen et al [1] provide a comprehensive deterministic model for supply chainmanagement called Global Supply Chain Model (GSCM), to determine optimalmanufacturing and distribution strategies. A successful implementation of thismodel was done at the Digital Equipment Corporation.

4.2 Analytical Performance Models

Models of supply chains in a dynamic and stochastic environment consider thenetwork as a discrete event dynamic system. Such systems can be studied as


Markov chains, stochastic Petri nets and queueing network models [56, 44].Malone and Smith [38], in their study, have looked at organizational andcoordination structures, which constitute a key element of any business process.Raghavan and Viswanadham [43] discuss performance modeling and dynamicscheduling of make-to-order supply chains using fork-join queueing networks.Viswanadham and Raghavan compare make-to-stock and assemble-to-ordersystems using generalized stochastic Petri net models [58]. They also use integratedqueueing and Petri net models for solving the decoupling point location problem,i.e. the point (facility) in the supply chain from where all finished goods areassembled to confirmed customer orders [44].

4.3 Simulation and Information Models

Models discussed above are high abstraction models for business processes undersimplifying assumptions such as Markovian dynamics. To obtain very accurateand detailed models, one has to represent many realistic features, which is possiblein simulation models. Development of simulation models for understanding issuesof supply chain decision making has gained importance in recent years. Some ofthe studies are Malone [39], Connors et al.[13], and Feigin et al.[16]. Bhaskaranand Leung [4] describe re-engineering of supply chains using queueing networkmodels and simulation.

Feigin et al. have looked into enterprise modeling and simulation in an objectoriented environment [16]. Similar work has been done by Mujtaba et al. [41] andChu [9]. But typically developing and implementing object models for a givensupply chain takes a long time. A set of generic objects representing various entitiesof supply chain can greatly shorten this period. Swaminathan et al. [53] have takenthis approach.They have built a generic object-based agent framework with whichthey can build simulation models for a variety of supply chain networks.

The Integrated Supply Chain Management (ISCM) Project [31], [19] and[18] has led to the development of a unified testbed used by the agents built forsupply chain functions: Logistics, Transportation Management, Order Acquisition,Resource Management, Scheduling and Dispatching. These agents rely onontologies for activity, state, time, resources, cost, quality and organization as acommon vocabulary for communication and use the services of Information Agents


that automatically distribute information and manage information consistencyand evolution.

Huang, Cheng, and Yang [29] have proposed mobile objects as a naturalframework for structuring a supply chain information system. Much of the recentwork on web-enabled supply chain information systems [42] also uses the objectmodel as the conceptual framework.

4.4 Software Tools

A variety of commercial software packages for supply chains indicate the growingimportance of this area.

• IBM Supply Chain Simulator (IBM-SCS): The IBM supply chain simulatoris a software tool that can help a company or a group of companies makestrategic business decisions about the design and operation of its supplychain [7]. IBM-SCS deploys a mix of simulation and optimization functionsto model and analyze supply chain issues such as facilities location,replenishment policies, manufacturing policies, logistics strategies, stockinglevels, lead times, process costs, and customer service. SCS is built uponSIMPROCESS, a general purpose business process simulator.

• i2 Technologies (Rhythm) Rhythm or Trade Matrix family of supply chainmanagement products have emerged as leading supply chain products inrecent times [45]. The Rhythm suite of products includes: Supply ChainPlanner, Supply Chain strategist, Scheduler, Global Logistics Manager,and Demand Planner.

• SAP (APO): Advanced Planner and Optimizer of SAP corporation [49] isnow being extensively marketed and deployed. APO consists of the followingsoftware modules: Enterprise Planning, Demand Planning, ProductionScheduling, and Distribution Planning.

• Manugistics (Manugistics6): Manugistics6 [40] is a broad supply-chainplanning and management suite, consisting of the following: Configuration,Constraint-Based Master Planning, Demand Management, ManufacturingPlanning & Scheduling, Material Planning, Network Design andOptimization, Purchase Planning, Replenishment Planning, Supply Chain


Analytics, Transportation Management, Vendor Managed Inventory(VMI)/Continuous Replenishment Planning (CRP).

5 Software Models of Supply Chains

In this section, we first discuss the need for and the advantages of object orientedmodeling approach. Following this, we present our object library for supply chains.We provide a detailed description and classification of various objects in our library.Finally we bring out the utility of the object library using an illustrative example.

5.1 Rationale for Object Models

One reason for the widespread appeal of object oriented modeling [5, 47, 30] isthe natural mapping paradigm. This is made possible by the object construct,which allows a one-to-one mapping between objects in the system being modeled(e.g., distributor, supplier, plant, vehicle, etc.) and their abstractions in the objectmodel. Object oriented modeling also has a major effect on implementationthrough its facilitation of modular design and software reusability. When exploited,object oriented programming features such as encapsulation, inheritance, andpolymorphism facilitate code reuse and programming efficiency. The feature ofencapsulation enforces structured development. Object oriented modeling alsooffers potential advantages in the incremental development and verification oflarge-scale systems. Thus object oriented modeling applied to supply chainmodeling and analysis can be potentially very beneficial.

Supply chain decision making requires rapid and flexible modeling approachat various levels of detail. Object oriented modeling can be used to design andimplement reusable classes for building models of supply chains and create asupply chain object library. The concept of an object library facilitates rapidmodel development of any given supply chain and aids in application of themodeling architecture to specific scenarios at various levels of abstraction.

5.1.1 Structural Objects

5.2 Major Objects in the Supply Chain Library

We present an object library with reusability and extensibility features for modelingand analysis of supply chain networks. We classify the elements of our objectlibrary into two categories: structural objects and policy objects.


The structural objects are the physical entities of supply chain networks. Thephysical structure of the supply chain networks is modeled using these classes.Physically the supply chain network is composed of plants, warehouses,distributors, retailers, suppliers, customers, orders, vehicles, etc. The policy objectsembed business logic which is used for controlling the flow of products andinformation through the network. The policy elements provided in our libraryare inventory policy, order management policy, demand planning policy, supplyplanning policy, distribution policy.

The set of structural objects is used in conjunction with the policy objectsto build the object models of a supply chain. These models are used to providecustomized inputs for various decision problems to be studied. The classesrepresenting the structural objects are described below. We provide a taxonomyof the objects in Figure 1. The objects like suppliers, distributors, etc., are

Structural Objects

Customer Order Plant Warehouse Supplier Distributor

RetailerVehicle

Manufacturing PlantAssembly PlantLate-customization Center

External Customer OrderWarehouse OrderManufacturing OrderLate-customization OrderSupplier Order

InternalExternal

Figure 1: A Taxonomy of Structural Objects


considered to be external entities. Therefore their subclasses have not beenshown in the taxonomy of the supply chain. The supply chain consists of thefollowing entities:

Customer

A customer can be either an internal customer or an external customer. The internalcustomers are the various entities of the network like the plants and the distributors.The external customers are the consumers of the products (finished or semi-finished) of the supply chain. The customer class may also contain informationon the desired service level and priority of the customer.

Order

An order contains the name and the quantities of the desired products, the nameof the customer, and the name of the entity to which the order is placed. Anorder can belong to any of the following categories: external customer order,warehouse order, manufacturing order, late-customization order, and supplierorder. External customer orders are generated either from forecasts (demandplanning policies) or by the customer objects in a deterministic manner.

Plant

A plant manufactures or assembles finished or semi finished products from rawmaterials, and/or sub-assemblies. A plant may have its associated raw materialwarehouse, in-process inventory warehouse, and finished goods warehouse. Wehave the following derived classes from the base class plant: manufacturing plantand late-customization center. Manufacturing process requires specified cycle timeto convert input parts into output products. Several factors are responsible forthe performance of a plant, for example the capacity, availability of input parts,and the fluctuations in the cycle time.

Supplier

A supplier provides a plant with raw materials or sub-assemblies. A suppliercould be a manufacturing plant, or a late-customization center, or a full fledgedsupply chain. Since the supplier is outside the purview of the supply chain understudy, it is modeled as a class which can supply the required products under


specified constraints. We have considered only lead-time and capacity constraints,but other specific constraints can be easily added to the supplier class.

Retailer

An external customer generally buys the products from the retailer. A retailer hasan associated warehouse, where the inventories of the products are stored. A retailercan receive deliveries from distributor, or manufacturing plant, or late-customization center, or from some other retailer. The product is delivered tocustomer if it is available in the retailer’s warehouse. Otherwise the order is addedto a queue for the particular product, according to a pre-assigned priority. Theorder is delivered when the product is received (from distributor, or plant, orlate-customization center as the case may be).

Distributor

A distributor receives deliveries from manufacturing plant, or late-customizationcenter, or from other distributors. The distributor may have an associatedwarehouse. It supplies to the retailers, or sometimes to other distributors. It mayalso supply to the late-customization center with information on customer speci�edrequirements.

Vehicle

Transportation vehicles move products from one node of the network to another.Each vehicle has characteristics in terms of products it can carry, capacity (involume or weight), costs, and speed.

Warehouse

A warehouse is a storage facility which is characterized by the nature and capacityof the products it can store. A warehouse can be attached to the plant, thedistributor, and the retailer. A warehouse can be used for storage of raw-materialinventories, in-process inventories, and finished product inventories.

5.2.1 Policy Objects

The policy objects model strategies which describe the protocols used inprocurement, manufacturing, transportation, and distribution of material withinthe supply chain. Policy objects have an identity and behavior of their own and it


is therefore appropriate to model them as full-fledged objects. Moreover, separatingthe policy objects from the structural objects enables composing them in veryflexible and powerful ways to obtain comprehensive modeling power. Policy objectsenable a society of structural objects to have varying behaviors. For example, astructural object such as “Warehouse“ can be composed with a policy objectsuch as Inventory Policy to describe different types of warehouse managementand replenishment schemes. We have identified and defined the following policyobjects. A taxonomy of the policy objects is shown in Figure 2.

Policy Objects

Figure 2: A Taxonomy of Policy Objects

InventoryPolicy

ManufacturingPolicy

Multi-Echelon PolicyEOQ PolicyBase Stock PolicyMRP Policy

Make-to-Stock PolicyMake-to-Order PolicyAssemble-to-Order Policy

Order ProcessingOrder Scheduling

Order ManagementPolicy

DemandPlanning

Policy

ForecastingAuctions

Routing PolicyScheduling Policy

DistributionPolicy

SupplyPlanning

policy

Constraint PlanningResource Scheduling

Inventory Policy

Inventory policies guide the flow of materials in the supply chain networks.Different inventory policies include multi-echelon inventory policies, and EOQpolicies [28, 15, 59].


Manufacturing Policy

The manufacturing policy can be make-to-stock, or make-to-order, or assemble-to-order, or a combination of the these policies [13, 11, 34, 44].

• Make-to-stock Policy: The plant builds products according to advance plans,and pushes the finished products into the warehouses.

• Make-to-order Policy: The plant produces a product from its input partsonly when an order for that product is received.

• Assemble-to-order Policy: The manufacturing plant produces componentswhich can be assembled by the late customization center according tocustomer specification.

Order Management Policy

The order management policy models the order processing and scheduling atany node of the supply chain. The delay incurred in the process is also considered.For simplicity we do not allow any partial shipment of an order.

Demand Planning Policy

The demand planning policy generates forecasts of expected demands for futureperiods. Various forecasting policies like time series, and regression analysis (basedon factors like competition, economic condition, promotional efforts, etc.) areconsidered [17].

Supply Planning Policy

Supply planning is a critical process in determination of company’s service andinventory levels. This models the allocation of production and distributionresources to meet the actual and forecast demand under capacity and supplyconstraints [12, 17, 14].

Distribution Policy

The product distribution is the process of delivering a product from the suppliersite to the end customer. The scheduling policies include routing and schedulingof vehicles to optimize delivery schedules [20].


5.3 Design Patterns

Design patterns [21] have emerged as an important best practice in object orienteddesign. The use of design patterns enables robust and highly reusable objects tobe created. This is a critical requirement for object oriented modeling of supplychains. Several design patterns can be used as part of supply chain object library.These include abstract factory, strategy, mediator, composite, and chain ofresponsibility. We will discuss two of the design patterns [21] in some detail..

• The Strategy Pattern The Strategy pattern consists of a number of relatedalgorithms encapsulated in the form of classes. A class which requires aparticular service or function and which has several ways of carrying outthat function is a candidate for the Strategy pattern. Objects choose betweenthese algorithms based on computational efficiency or user choice. Therecan be any number of strategies and more can be added and any of themcan be changed at any time. There are a number of cases in programswhere we would like to do the same thing in several different ways. Forexample, we may have different distribution strategies (Figure 3). Strategypattern allows us to choose from several algorithms dynamically.

• The Mediator Pattern When a program is made up of a number of classes,the logic and computation is divided logically among these classes. However,

Figure 3: An Example of a Strategy Pattern

Interface Nodeaggregation Abstract Strategy

Strategy ()

Concrete Strategy A

Strategy ()

Distribution Policyimplements InterfaceNode

Concrete Strategy B

Strategy ()

Concrete Strategy C

Strategy ()


as more of these isolated classes are developed in a program, the problem ofcommunication between these classes becomes more complex. The moreeach class needs to know about the methods of another class, the moretangled the class structure can become. Further, it can become difficult tochange the program, since any change may affect code in several otherclasses. The Mediator pattern addresses this problem by promoting loosercoupling between these classes. Mediators accomplish this by being theonly class that has detailed knowledge of the methods of other classes.Classes inform the mediator when changes occur and the Mediator passesthem on to any other classes that need to be informed. The various supplychain objects in our system communicate with the help of mediator classobjects. For example we have created mediator objects to communicatebetween various suppliers, vehicles, and plants; plants, warehouses,distributors, retailers, and vehicles. The advantage is that the mediator isthe only class that needs to be changed if some classes change or newclasses are added. Thus reuse of the various objects become easier.

5.4 An Illustrative Example

To illustrate creation of object models, we discuss a real-world supply chainnetwork and present a first level object model that can be created. The system(see Figure 4) [48] can be described as follows. Shri Shakti (SS) LPG Limitedimports and markets LPG (Liquid Petroleum Gas) in South India. SS has itsimport facilities located at Kakinada and Mangalore ports. It has two storagefacilities, one for each port. The storage facilities are located in the vicinity ofthese ports. The LPG is transported from the port to the storage facilities throughpipelines. The quantity of LPG that SS can import at either of the ports ispractically unlimited. SS sells LPG in bulk to industrial customers directly fromthe two storage facilities. It markets LPG in packed form to domestic andcommercial establishments through its dealers. When a customer’s cylinder runsout of LPG, the dealer replaces the empty cylinder with a filled cylinder at thecustomer’s location (dealer and customers are typically at most five miles apart).The dealer recovers the cost of transporting cylinders from commission on a perrefill basis from SS.


Figure 4: Diagram Representing the Shri Shakti LPG Supply Chain

Each dealer town (town in which SS has one or more dealers) gets its suppliesfrom a designated bottling plant. Each bottling plant typically services a set ofdealer towns. Each dealer sends the empty cylinders in full truck loads to itsassociated bottling plant. Since the plant maintains an inventory of filled cylinders,the trucks are fully loaded with filled cylinders on their return trips to the dealer.Each bottling plant receives LPG in tankers from the cheaper of the two portsand storage facilities (those at Kakinada and Mangalore), cheaper in terms of perunit cost of procurement and transportation. The tankers are dedicated totransporting LPG, and hence, SS pays the transporters for both delivery andreturn trips to the storage facilities.

Pipeline


Figure 5: An Object Model of Shri Shakti LPG Supply Chain

An object model for this system can be built as follows. SS supply chainfollows make-to-stock manufacturing policy. This system consists of the followingentities: Port, Pipeline, LPG Storage Facility, Tanker, Industrial Customer, BottlingPlant, Truck, Dealer, Domestic Customer, and Commercial Customer. Themake-to-stock policy and the above objects are derived from our object library.The object model for this system is shown in Figure 5. This object model becomesthe basis for formulating and studying various strategic, tactical, and operationaldecision issues.

Note that this object model includes only structural objects. By includingpolicy objects appropriately, the model will be able to describe the supply chain


structure and behavior completely. For example, a policy object Inventory Policysuitably sub-classed into classes such as EOQ Policy , Multi-echelon Policy,etc., can be composed with a structural object such as Warehouse to describeinventory control dynamics in the supply chain. Similarly, a policy object suchas Distribution Policy can be composed with a structural object such asDistributor to represent a wide variety of distribution mechanisms.

6 Summary

This paper is an attempt to bring out the key role of mathematical and softwaremodels in supply chain modeling and decision making. First, we described thespectrum of strategic, tactical, and operational decisions in supply chain management.Then, we presented non-financial and financial metrics of describing supply chainperformance. We next described briefly the different kinds of mathematical modelsthat aid supply chain decision making. Finally, we brought out the role of objectoriented models in software modeling of supply chain networks.

We now provide a pointers to a few useful articles and books in the literaturefrom which more details can be obtained. There are excellent books on supplychain management; for example, see the books [25, 8, 51, 36, 55, 28]. Theedited volume by Tayur, Ganeshan, and Magazine [54] is a comprehensive sourceof research in quantitative models of supply chain networks until 1999. Theedited volume by Simchi-Levi, Wu, and Shen [37] is is an excellent source ofstate-of-the-art research in quantitative models of supply chain networks. Thepaper by Biswas and Narahari [3] describes a decision support systems that usesthe models and tools described in this paper. The papers by Garg, Narahari, andViswanadham [22, 23] describe the design of high performance supply chainsusing statistical techniques.

(Y. Narahari is currently a Professor at the Department of Computer Science andAutomation, Indian Institute of Science, Bangalore. He completed his Ph.D. at thesame department in 1988. He has earlier co-authored a widely acclaimed textbook onPerformance Modeling of Automated Manufacturing Systems (Prentice Hall, EnglewoodCliffs, NJ, 1992). He is currently on the editorial board of IEEE Transactions onSystems, Man & Cybernetics (Part A) and IEEE Transactions on Automation Science& Engineering. He can be reached at [email protected]


Shantanu Biswas is currently working as senior research engineer in EnterpriseResearch in Motorola Labs. His research interests include combinatorial Optimization,approxination algorithmes, game theory and mechanism design and ad-HOC networks.He can be reached at [email protected])

Endnotes

[1] BC Arntzen, G G Brown, TP Harrison, and L Trafton Global supply chain management atDigital Equipment Corporation. Interfaces, 25(1):69 93, January-February 1995.

[2] S Bagchi, SJ Buckley, M Ettl, and GY Lin. Experience using the IBM supply chain simulator.In Winter Simulation Conference, December 1998.

[3] S Biswas and Y Narahari Object oriented modeling for decision support in supply chain

networks. European Journal of Operational Research, Volume 153, 2004, pp. 704-726.

[4] K Bhaskaran and YT Leung. Manufacturing supply chain modelling and reengineering.

Sadhana - Academy Proceedings in Engineering Sciences, 22(2):165 187, April 1997.

[5] G Booch Object-Oriented Analysis and Design with Applications. Benjamin-Cummings,

Redwood, CA, USA, 1994.

[6] G Booch, J Rumbaugh, and I Jacobson. The Unified Modeling Language User Guide. Addison

Wesley Longman, Inc., 1999.

[7] SJ Buckley and JE Smith. Supply chain simulation. Technical report, Georgia Tech Logistics

Short Course, June 1997.

[8] Sunil Chopra and Peter Meindel. Supply Chain Management: Strategy, Planning, andOperation, Prentice Hall of India, 2002.

[9] PC Chu. Actors and scripts : object-oriented simulation of enterprise systems. Journal of SystemsSoftware, 37(3):187 199, June 1997.

[10] MA Cohen, PV Kamesam, P Kleindorfer, HL Lee, and A Tekerian. Optimizer: IBM’s multi-echelon inventory system for managing service logistics. Interfaces, 20(1):65 82, January-

February 1990.

[11] MA Cohen and HL Lee. Strategic analysis of integrated production distribution systems:Models and methods. Operations Research, 36(2):216 228, March-April 1988.

[12] MA Cohen and HL Lee. Resource deployment analysis of global manufacturing anddistribution networks. Journal of Manufacturing and Operations Management, pages 81 104,

1989.


[13] D Connors, C An, S Buckley, G Feigin, A Levas, N Nayak, R Petrakian, and R Srinivasan.Dynamic modeling of re-engineering supply chains. Technical report, IBM Research Division,

May 1996.

[14] B Dietrich, D Connors, T Ervolina, JP Fasano, R Srinivasan, R Wittrock, and R Jayaraman.Production and procurement planning under resource availability constraints and demand

variability. Technical report, IBM Research Division, 1995.

[15] M Ettl, GE Feigin, GY Lin, and DD Yao. A supply network model with base-stock control

and service requirements. Technical report, IBM Research Division, 1996.

[16] G Feigin, C An, D Connors, and I Crawford. Shape up, ship out. OR/MS Today, pages 24 30,

April 1996.

[17] ML Fisher, JH Hammond, and WR Obermeyer et al. Making supply meet demand. HarvardBusiness Review, 72(3):83 93, May-June 1994.

[18] MS Fox and M Gruninger. On ontologies and enterprise modelling. In InternationalConference on Enterprise Integration Modelling Technology 97. Springer-Verlag, 1997.

[19] MS Fox and M Gruninger. Enterprise modelling. AI Magazine, pages 109 121, 1998.

[20] JB Fuller, J O’Conor, and R Rawlinson. Tailored logistics: The next advantage. Harvard

Business Review, pages 87 98, May-June 1993.

[21] E Gamma, R Helm, R Johnson, and J Vlissides. Design Patterns: Elements of Reusable Object-

Oriented Software. Addison Wesley, October 1994.

[22] DGarg, Y Narahari, N Viswanadham. Design of Six Sigma Supply Chains. IEEE Transactionson Automation Science and Engineering Volume 1, Number 1, July 2004, pp. 38-57.

[23] DGarg, Y Narahari, N Viswanadham. Achieving sharp deliveries in supply chains throughvariance pool allocation. To appear in: European Journal of Operational Research, 2006.

[24] AM Geoffrion and GW Graves. Multi-commodity distribution system design by Bender’s

decomposition. Management Science, 20(5):822 824, January 1974.

[25] RB Handfield and E.L. Nichols, Jr. Introduction to Supply Chain Management. Prentice Hall,

1999.

[26] DA Hicks. A manager’s guide to supply chain and logistics problem solving tools and

techniques, Part I: Understanding the techniques. IIE Solutions, September 1997.

[27] DA Hicks. A manager’s guide to supply chain and logistics problem solving tools and

techniques, Part II: Tools, companies and industries. IIE Solutions, October 1997.

[28] WJ Hopp and ML Spearman. Factory Physics: Foundations of Manufacturing Management.

Irwin, 1996.


[29] E Huang, FT Cheng, and HC Yang. Development of a collaborative and event driven supplychain information system using mobile object technology. In 1999 International Conference

on Robotics and Automation, pages 1776 1781, 1999.

[30] I Jacobson. Object-oriented Software Engineering : A use-case approach. Addison-Wesley, 1994.

[31] Enterprise Integration Laboratory.The integrated supply chain management project. http://

www.eil.utoronto.ca/iscm-descr.html, 1996.

[32] Java Programming Language. http://java.sun.com. Sun Microsystems, 1999.

[33] HL. Lee and C Billington. Managing supply chain inventory: Pitfalls and opportunities.

Sloan Management Review, 33(3):65 73, 1992.

[34] HL Lee and C Billington. Material management in decentralized supply chains. OperationsResearch, 41(5):835 847, 1993.

[35] HL Lee and C Billington. Evolution of supply chain management models and practice atHewlett-Packard Company. Interfaces, 25(5):42 63, September-October 1995.

[36] David Simchi Levi, Philip Kaminsky, and Edith Simchi Levi. Designing and Managing the

Supply Chain: Concepts, Strategies, and Case Studies. Irwin McGrawHill, 2000.

[37] David Simchi Levi, David Wu, and W. Shen (Editors). Handbook of Supply Chain Analysis in

the eBusiness era, Kluwer Academic Publishers, 2005.

[38] TW Malone and SA Smith. Modeling the performance of organizational structures.

Operations Research, 36(3):421 436, May-June 1988.

[39] TW Malone. Modeling coordination in organization and markets. Management Science,

33(3):1317 1331, October 1987.

[40] Manugistics6. http://www.manugistics.com. Manugistics, 1999.

[41] MS Mujtaba. Enterprise modeling and simulation: Complex dynamic behavior of a simple

model of manufacturing. Hewlett-Packard Journal, pages 80 114, December 1994.

[42] ME Nissen. Supply chain process and agent design for ecommerce. http://web.nps.navy.mil/

menissen, 2000. [43] N. R. S. Raghavan and N. Viswanadham. Performance modeling anddynamic scheduling of make-to-order supply chains. In Third International Conference on

Agile and Computer Integrated Manufacturing, Rensselaer Polytechnic Institute, October 1998.

[44] NRS Raghavan. Performance Analysis and Scheduling of Manufacturing Supply Chain Networks.

PhD thesis, Dept.of Computer Science and Automation, Indian Institute of Science,

Bangalore, India, November 1998. [45] Rhythm and Trade Matrix Suite of Solutions. http://www.i2.com. I2 Technologies, 2002.


[46] Roshan Gaonkar, Dynamic Configuration and Synchronization of Collaborative E-SupplyChain Networks, Ph.D. thesis, Department of Mechanical Engineering, National University

of Singapore, Singapore, 2002.

[47] J Rumbaugh, M Blaha, W Premerlani, F Eddy, and W Lorensen. Object-Oriented Modeling

and Design. Engle-wood Cliffs, New Jersey, Prentice Hall, 1991.

[48] JK Sankaran and NRS Raghavan. Locating and sizing plants for bottling Propane in SouthIndia. Interfaces, 27(6):1 15, November-December 1997.

[49] SAP Corporation. http://www.sap.com. Systems, Applications, and Products, 2002.

[50] JF Shapiro. Bottom-up versus top-down approaches to supply chain modeling. In S. Tayur, R.

Ganeshan, and M. Magazine, editors, Quantitative Models for Supply Chain Management,pages 737 760. Kluwer Academic Publishers, 1999.

[51] Jeremy F. Shapiro. Modeling the Supply Chain. Duxbury Thomson Learning, 2001.

[52] Rational Software. Unified modeling language documentation, version 1.1. http://

www.rational.com/uml/ documen-tation.html, September 1997.

[53] J M Swaminathan, S F Smith, and N M Sadeh. A multi agent framework for modeling supplychain dynamics. Decision Sciences, 1997.

[54] S Tayur, R Ganeshan, and M Magazine, editors. Quantitative Models for Supply ChainManagement. Kluwer Academic Publishers, 1999.

[55] N Viswanadham. Analysis and Design of Manufacturing Enterprises. Kluwer AcademicPublishers, 1999.

[56] N Viswanadham and Y Narahari. Performance Modeling of Automated Manufacturing Systems.

Prentice Hall, Englewood Cliffs, NJ, 1992.

[57] N Viswanadham and RS Gaonkar, Strategic sourcing and collaborative planning in Internet-

enabled supply chain networks producing multi-generation products, Tech. Rep.,Department of Mechanical Engineering, National University of Singapore, Singapore, 2003.

[58] N Viswanadham and NRS Raghavan. Petri net models for performance analysis of supply

chain networks. In Operations Management Workshop, Indian Institute of Management,Bangalore, December 1998.

[59] TE Vollman, WL Berry, and DC Whybark. Manufacturing Planning and Control Systems. TheDow Jones-Irwin/APICS Series in Production Management, 1997.

performance measures and performance models for supply...

Documents