Cours du master de recherche Génie Industriel Supply Chain Logistics and Operations Management Xiaolan XIE, Professeur Centre Ingénierie et Santé CNRS URM 6158 LIMOS – Equipe ROGI Ecole Nationale Supérieure des Mines xie@emse.fr Supply Chain Logistics and Operations Management Xiaolan XIE, Professeur Centre Ingénierie & Santé CNRS UMR 6158 LIMOS-Equipe ROGI Ecole Nationale Supérieure des Mines xie@emse.fr polycopie sur http://www.emse.fr/~xie/master 1 2 Plan References Chapter 1. Introduction Chapter 2. Supply chain design Chapter 3. Managing economies of scale in a supply chain Chapter 4. Managing uncertainty in a supply chain Chapter 5. Value of information Chapter 6. Distribution strategies & strategic alliance Chapitre 7. Outils informatiques en SCM • D. Simchi-Levi, P. Kaminsky, E. Simchi-Levi, «Designing and managing the supply chain», Irwin MsGraw-Hill • S. Chopra, P. Meindl, « Supply chain management: strategy, planning and operation » • G. Cachon, Terwiesch, « Matching supply with demand » • H. Stadtler C. Kilger : "Supply chain management and advanced planning" 3 4 1 What is a Supply Chain? Chapter 1. Introduction to Supply chain management 1. 2. 3. 4. • A supply chain consists of all parties involved, directly or indirectly, in fulfilling customer requests What is SCM? Different views of supply chains Drivers and objectives of supply chains Supply chain decisions • The entire process from point of origin (raw materials) to point of consumption (final products bought by customers) • A network (interdependent system) of facilities including • materials supply from suppliers • transformation of materials to (inventories of) semi-finished and finished products • distribution of finished products to customers •Supply network or supply web. 5 6 What is a Supply Chain? Supply Chain Examples • SCM is a set of approaches utilized to efficiently integrate suppliers, manufacturers, warehouses, and stores, so that merchandise is produced and distributed at the right quantities, to the right locations, and at the right time, in orders to minimize systemwide costs while satisfying service level requirements. Example 1: Wal-Mart Procter & Gamble Da-Fa Clothing, Inc. (China) Wal-Mart or third-party distribution centers Wal-Mart Stores Customers Request: Buying detergent, clothes, TV, …... SONY Factory (Malaysia) 1: SCM takes into account every facility that has an impact on cost and plays a role in product making and distribution. Plastic Producer 2: The objective of SCM is to be efficient and cost-effective across the entire system. Chemical Producer Fabric Producer Electronics Components Producer Zipper Producer Plastic Producer Thread Producer SC of detergent : (i) custmer need for detergent, (ii) the Wal-Mart retail store he visits, (iii) FGI warehouse or DC supplying the store using trucks by a 3rd party; (iv) DC is stocked by the manufacturer P&G, (v) P&G plants receives raw materials fro suppliers, who may have been supplied by lower-tier suppliers (eg packaging from Tenneco while Tenneco receives raw materials from other suppliers. 3: Because SCM revolves around efficient integration of all suppliers-manufacturers-warehouses-stores, it encompasses the firm’s activities at many levels, from strategic to tactical and operational levels. 7 8 Pioneer of cross-docking system. 2 Example 3: Dell Example 2: Hewlett & Packard (HP) Suppliers Suppliers Suppliers IC Mfg US DCs Retailer PC Board Europe DCs Retailer Consumers Far East DCs Retailer Consumers Consumers Monitors by SONY (Mexico) FAT Subassembly Keyboards by Acer (Taiwan) CPU by Intel (USA) Dell Assembly Plant Customers order computers on-line Other components Suppliers SC : customer, Dell’s Web site, Dell assembly and all Dell’s suppliers and their suppliers FAT = Final assembly & test IC Mfg = Integrated circuit manufacturing PC Board = Printed circuit board Known for its direct sell and build-to-order system. Pioneer of Postponement or delayed differentiation 9 10 Supplier Supplier manufacturer Distribution system Customer A Flow View of SC A Typical Supply Chain Flows in both directions Dominant flow of products and services: Flow of products and semi-finished products, rework, recycling, etc Supplier Suppliers Manufacturing Plants Regional Warehouses (Distribution Centers) Field Warehouses (Distribution Centers) Dominant flow of demand and design information: Procurement order, demand data, inventory information, product information, prices, etc. Retail Stores Customers Financial flow: Payment, cash receivable, refund, consignment contracts, etc. 11 12 3 A Flow View of SC An Organization View of SC Example 1: Wal-Mart Procter & Gamble Da-Fa Clothing, Inc. (China) SONY Factory (Malaysia) Wal-Mart or third-party distribution centers Wal-Mart Stores Customers Request: Buying detergent, clothes, TV, …... Fabric Producer Electronics Components Producer Zipper Producer Plastic Producer • A supplier or a customer can be internal or belong to different companies • The supplier and customer may have different objectives and make decisions independently • The best performance can only be achieved when all members of the SC work for the same goal, in some way Plastic Producer • Two SC visions : Chemical Producer Thread Producer – Intra-organizational Supply Chain : cooperations • Wal-Mart provides the product, pricing and availability, info to customer. • Customer transfers funds to Wal-Mart. of all facilities of a large company • Wal-Mart conveys point-of-sales data and replenishment orders to warehouses or DC, who transfer the replenishment orders via trucks back to the store. – Inter-organizational Supply Chain : network of • Wal-Mart transfers funds to DC after the replenishment. 13 • DC provides pricing info and sends delivery schedules to Wal-Mart • Wal-Mart may send back packaging material to be recycled. enterprises that work together toward a commun goal 13 14 An Organization View of SC An Organization View of SC Focus company Intra SC Procter & Gamble Da-Fa Clothing, Inc. (China) Wal-Mart or third-party distribution centers Wal-Mart Stores supplier Customers Request: Buying detergent, clothes, TV, …... Cust. CF Assem DC supplier SONY Factory (Malaysia) Cust. CF Assem Cust. Fabric Producer Electronics Components Producer Plastic Producer Zipper Producer supplier CF Assem DC Plastic Producer Cust. Chemical Producer out-bound logistics In-bound logistics Thread Producer Sourcing 15 Making Delivery, sales, service 16 4 An Organization View of SC An Organization View of SC Single-location entreprise Multiple-location entreprise Factory 1 Factory 3 Factory 2 17 18 An Organization View of SC An Organization View of SC Supply chain Supply network Entreprise A –Factory 1 Stores Entreprise A –Factory 3 Entreprise A –Factory 2 Warehouse Tier 2 suppliers Tier 1 suppliers Procurement 19 20 5 Processes Involved in a Supply Chain Processes Involved in a Supply Chain A supply chain is a seqence of processes & flows that take places within and between different SC stages and combine to fill a customer need (To be analysed with SCOR). ___________________ 1. 2. 3. 4. Suppliers Manufacturers Cycle view: Processes in a SC are divided into a series of cycles, each performed at interface between two successive stages ___________________ 1. 2. 3. 4. Distributors ___________________ 1. Procurement 2. Manufacturing 3. Customer order filling 4. Delivery Retailers Consumers ___________________ 1. 2. 3. 4. Not every SC will have all four cycles clearly separrated. Customers Cust. Order cycle Retailers Replenishment cycle Distributors Manufacturing cycle Manufacturers Procurement cycle Suppliers Dell 22 Push/Pull view of SC processes Traditional Push/Pull view of intra-organisational supply chains Cycle view: Each cycle consists of six subprocesses. Each party works to improve the efficiency of each subprocess. Demand information changes between different cycles The scale of orders grows as we move farther from the customer Buyer stage places order Stages 21 Processes Involved in a Supply Chain Supplier stage markets product Cycles Delivery lead time Design Purchase manufacture Ship manufacture Assemble Engineering to order Make to order Ship PULL PUSH Buyer stage receives supply Assemble Delivery lead time Inventory Buyer returns reverse flows to supplier or 3rd party manufacture Delivery lead time Inventory Assemble Ship assemble to order Delivery lead time Supplier stage receives order Supplier stage supplies order manufacture 23 Assemble Inventory Ship Make to stock 24 6 Push/Pull view of SC processes Push/Pull view of SC processes Push/Pull view: Push / Pull boundary : point at which customer orders arrive Processes divided into 2 categories depending on whether they are executed in response to a customer order or in anticipation of customer orders. L.L. Bean : a mail-order company • Pull processes are initiated by a customer order (demand known with certainty) DELL : direct sell Customers PULL • Push processes are initiated in anticipation of customer order (demand unknown and must be forecast) Cust. Order & Replenishment cycle Customers PULL Cust. Order & manufacturing L.L. Bean cycle Replenishmnt &Manufacturing Manufacturer cycle PUSH Procurement cycle Manufacturing (DELL) Suppliers PUSH Procurement cycle Suppliers 25 Push/Pull view of SC processes Push / Pull boundary : point at which customer orders arrive • Raw material such as fabric purchased 6-9 months before demand arrival • Mfg begins 3-6 months before point of sale Only inventory of components Push/Pull view of SC processes Key point : Paint industry : production of the base, mixing colors, and packing. A cycle view clearly defines the processes involved and owners of each process. Useful when considering operation decisions because it specifies roles & responsibilities of each member & the desired outcome of each process Till 1980, all processes done in large factories and paint cans shipped to stores. Key point: Now, • base preparation and packing of cans in push phase and • color mixing at retail stores. Push / Pull view very useful when considering strategic decision as it is relating to supply chain design. Another example of gains from suitably adjusting the push/pull boundary: Result? 26 Make-To-Stock or Make-To-Order 28 7 Push-Pull Supply Chains Push-Pull Strategies The Supply Chain Time Line Customers Suppliers PUSH STRATEGY Low Uncertainty PULL STRATEGY • The push-pull system takes advantage of the rules of forecasting: – Forecasts are always wrong – The longer the forecast horizon the worst is the forecast – Aggregate forecasts are more accurate • The Risk Pooling Concept • Delayed differentiation is another example – Consider Benetton sweater production High Uncertainty Push-Pull Boundary 29 Objective of Supply Chain Management 30 Managing a Supply Chain is not Easy • SCM is concerned with the efficient management of a supply chain so as to maximize supply chain profitability across the entire supply chain 1. Uncertain _demand__ Supply chain profitability or surplus = total revenue - total cost 2. Conflicting _objectives_ across the supply chain • Manage globally, not locally (Companies do not compete, their supply chains do) • Consider both _cost_ and _customer service level__ (Optimal tradeoff between _cost_and _responsiveness_) • Match _supply_ and _demand_ dynamically (in real time) • infrastructure (design), operations, integration Manufacturers Distributors Large production batches Low inventory Few DCs Retailers Few stores Low inventory Little variety Close to DCs Consumers Convenience Short lead time Large variety of products Large shipments 31 32 8 Conflicting Objectives of SC functions Function Objectives Implication Marketing • High revenue • High • High product availability •Low Supply Chain Performance Drivers • Logistic drivers: Facilities, inventories, transportation Customer service • Cross-functional drivers: information, sourcing, pricing • While logistics remains a major part, SCM is increasingly becoming focused in cross-functional drivers. •Low production cost Production •High level production •Long production runs Finance • Low investment & cost •Fewer fixed costs •Many •Few • High Disruption to production Inventories • The goal is to strike the balance between responsiveness and efficiency •Low •Low inventories 33 Supply Chain Performance Driver: Inventory 34 Supply Chain Performance Driver: Transportation • Inventory is the major source of cost in a supply chain and changing inventory policies dramatically alters SC’s efficiency and responsiveness. • High inventory level will be more responsive to market demand, however, at a higher cost • A lower inventory level will be less responsive to customer demand, however, lower inventory carrying and obsolescence cost • Nordstrom targets upper-end customers with high responsive requirement. It stocks a large variety and quantity of inventory than other department stores. It charges a premium by providing a higher service level to those customer who can afford it 35 • Transportation can take the form of many combinations of modes and routes. • Air shipment versus sea or land shipment • Full truckload or less-than-full truckload • Transportation choices have large impact on SC responsiveness and efficiency • Responsiveness versus cost • Nabisco, Inc. pioneered a collaborative logistics effort with other firms to share trucks and warehouses to save logistics cost • Home Depot moves 85% of its merchandise directly from suppliers to stores in full truckload, save in transportation and warehousing costs • Laura Ashley favors responsiveness for its mail-order businees and works with FedEx to allow next day delivers. 36 9 Supply Chain Performance Driver: Facilities • Facilities are physical locations where product is stored, assembled or fabricated. • Two major types : production and storage sites. • Decisions: role, location, capacity, flexibility • Centralized facility provides economy of scale, is more efficient • Decentralized facilities can be located close to the market, is more responsive • Toyota has a policy of having manufacturing facilities in every major market it served to be more responsive to its customers (side benefits: protection from currency fluctuations, trade barriers) Supply Chain Performance Driver: Information • Potentially the biggest driver as it directly affects all other drivers. • Information serves as the connection between the different stages of a supply chain. • High quality and timely information provides the most important ingredient in making high quality decisions to achieve both efficiency and responsiveness • Andersen Windows built an information system that can give customers immediate price quotes and automatically send an order to the factory. This allowed the company to get customized product to market rapidly • Information about the demand pattern, shipping options can help improve SC 37 38 Supply Chain Performance Driver: Sourcing Supply Chain Performance Driver: pricing • Sourcing is the choice of who will perform a particular SC activitity (prod, storage, transp, info management). • At the strategic level, these decisions determine what functions a firm outsources and what a firm performs. • Sourcing affects both efficiency and responsiveness. • Pricing affects the behaviour of the buyer of the good or service, and hence the SC performance. • It affects the customer segments that choose to buy the product, as well as the customers’ expectations. • It affects the SC in terms of responsiveness required as well as demand profile. • Pricing is also a lever for matching supply and demand through short-term discount. • After Motorola outsourced much of its production to contract manufacturers in China, its efficiency improves but its responsiveness suffers because of the long distances. • To remedy, it started flying in some of its cell phones from China even though this increased the transportation cost. 39 • Costco, a membership-based wholesaler in the US, has a policy that prices are kept steady but low. Customers expect low prices but are comfortable with lower level of product availability. The Costo SC aims to be very efficient, at the expense of some responsiveness. • Amazon : free shipping (7-14 days), standard shipping at $4.98 (3-5 days), 2-days shipping at $11.47, 1-day shipping at $20.47.40 10 Supply Chain Key Performance Indices: KPI Supply Chain Key Performance Indices: KPI Customer oriented KPI Facility-related Metrics Inventory-related Metrics • Delivery performance (Time to full delivery, on time delivery) • Capacity • Average inventory • Order fulfilment performance (order fill rate, order leadtime) • Utilisation • Perfect order fulfilment • Products with more than N number of days of inventory • Supply chain responsiveness (supply chain response time) • Theorectical flow/cycle time of production • Production flexibility (vol. flexibility, product-mix flex.) • Actual average flow/cycle time • Average safety inventory • Flow time efficiency = ratio of above two • Seasonal inventory Supply chain oriented KPI • Total logistics (SC) management cost • Product variety • Value-added productivity per employee • Processing/setup/down/idle time • Warranty cost • Average replenishment batch size • Fill rate • Fraction of time out of stock • Average production batch size • Cash-to-cash cycle time • Production service level • Inventory days of supply • Volume contribution of top 20% SKUs and customers • Asset turns 41 Supply Chain Key Performance Indices: KPI Transportation-related Metrics • Average inbound transportation cost • Average incoming shipment size • Average inbound transportation cost per shipment • Average outbound transportation cost • Average outbound shipment size • Average outbound transportation cost per shipment 42 Supply Chain Key Performance Indices: KPI Sourcing-related Metrics Information-related Metrics Pricing-related Metrics • Forecasting horizon • Days payable outstanding • Profit margin • Frequency of update • Average purchase price • Days sakes outstanding • Forecast errors • Range of purchase price • Incremental fixed cost per order • Seasonal factors • Average purchase quantity • Incremental variable cost per unit • Variance from plan • Fraction on-time deliveries • Average sale price • Ratio of demand variability to order variability (to identify bullwhip effect) • Supply quality • Average order size • Supply lead time • Range of sale price • Range of periodic sales • Fraction transportation by mode 43 44 11 Quick summary of important factors of SC Supply chain analysis • The SC includes all activities and processes to supply a product or service to a final customer. • Based on Supply Chain Operations Reference (SCOR) model for modelling and analysis of a Supply Chain. • Developped by the Supply Chain Council • Any number of companies can be linked in the SC. – http://supply-chain.org/ • A customer can be a supplier to another customer so the total chain can have a number of supplier/customer relationships. • With 5 generic process types: plan, source, make, deliver, return • Objective : model the supply chain management processes • While the distribution system can be direct from supplier to customer, depending on the products and markets, it can contain a number of intermediaries (distributors) such as wholesalers, warehouses, and retailers. Plan • Product or services usually flow from supplier to customer and design and demand information usually flows from customer to supplier. Rarely is this not so. Deliver Suppliers’ Supplier Source Make Source Deliver Make Deliver Your Company Supplier Internal or External Source Make Deliver Customer Internal or External Source Customer’ s Customer 45 Why Study SCM? 46 Why Study SCM? • Nike: BW “The Swoosh Stumbles” (2000) Delivery efficiency Inventory reduction Time to full delivery Quality of demand forecast Overall productivity SC cost reduction Service level Capacity increase – trouble with i2 implementation leads to major inventory problems. 16% – 28% Improvement 25% – 60% Improvement 30% – 50% Improvement 25% – 80% Improvement 10% – 16% Improvement 25% – 50% Improvement 20% – 30% Improvement 10% – 20% Improvement • Sony Play Station 2 – 2000 launching: shipped 500k units as opposed to 1,000k desired. • Amazon.com 2001 Q4 profits attributed to “improved operating efficiencies” Source: 1997 PRTM ISC Benchmark Study 47 – Sorting orders Fewer items were put in the wrong locations in its distribution centers, and shipped 35% more units with the same number of workers as a year ago, which helped cut fulfillment costs by $22 million last quarter, even with sales increase of 15%. – Anticipating Using software to more accurately forecast purchasing patterns by region, slashed inventory levels by $31 million, or 18%, in the 4th quarter – Opening a marketplace Allowing other sellers to offer used books and other merchandise on Amazon.com helped boost sales in the holiday season by 23%, to 38 48 million. 12 Why Study SCM? (examples from text) Why Study SCM? • In 1998, American companies spent $898 billion (GDP 10.6%) in supply-related activities. – Transportation 58% – Inventory 38% – Management 4% • Wal-Mart vs. K-Mart: In 10 years, Wal-Mart transformed itself by changing its logistics system (own fleet of trucks); over 80% shipped to stores from its own 27 DCs, rest directly from suppliers; usually received within 48 hours. It has the highest sales per square foot, inventory turnover and operating profit of any discount retailer. (crossdocking, Everyday Low Price, innovative IS/IT) • Third party logistics services grew in 1998 by 15% to nearly $40 billion. • It is estimated that the grocery industry could save $30 billion (10% of operating cost) by using effective logistics strategies. K-Mart, on the other hand, ships less than 50% on own, – A typical box of cereal spends more than three months getting from factory to supermarket. and suffers from supply chain inefficiency (inability to respond quickly to customer demand) – A typical new car spends 15 days traveling from the factory to the dealership, although actual travel time is 5 days. 49 50 Why Study SCM? Logistics vs SCM • Effective supply chain management is a top priority in many companies, e.g., Both are concerned with efficient management of physical flows, and matching supply & demand Boeing: Streamlining its procurement process using Exostar (internet exchange) -- to eliminate paper trail, facilitate information exchange across the supply chain General Motors: Reengineering its distribution operations, partnering with Vector SCM (a thrid-party service provider) -- to reduce in-transit inventory & achieve fast delivery from factories to dealers Logistics Supply Chain Management Scope: Within a firm Scope: Entire supply chain Objective: minimizing logistics cost Objective: minimizing cost & maximizing customer services Tactical – ignored by top mgmt Strategic -- top mgmt attention ExxonMobil: Redesigned its logistics network after the merge of Exxon and Mobil -- to drive down cost and complexity P&G, Dell, Cisco, Amazon, Bristol-Myers Squibb, McDonald’s, etc. 51 52 13 Why is SCM hot now? Supply Chain Decisions TIME FRAME TYPE IT as a key enabler years Strategic 1) Communications: Internet ability to share data through SC 2) Computing: PC speed ability to solve/analyze complicated quant models 3 mo.- 1year 1 + 2 => DECISIONS daily Also, transportation flexibility (multi-modal) Tactical Operational TYPICAL DECISIONS •Supply chain strategies (Sell direct or through retailers? Outsource or in-house? Focus on cost or customer service?) •Supply chain network design (How many plants? Location and capacities of plants and warehouses?) •Product mix at each plant •Workforce & Production planning •Inventory policies (safety stock level) •Which locations supply which markets •Transportation strategies •Production scheduling •Distribution scheduling and routing •Place inventory replenishment orders •Lead time quotations 53 Some Strategic Supply Chain Questions 54 Some Strategic Supply Chain Questions Q1: Should my supply chain focus more on cost or customer service? Supply chain strategy should fit the company’s competitive strategy. Q2: Do I need retailers in my supply chain? The competitive strategy defines, relative to its competitors, the set of customer needs that its seeks to satisfy through its products and services. Q3: Do I need warehouses/DCs in my supply chain? Q4: Perform all the functions in-house or outsource some? 55 56 14 Q1: Should my supply chain focus more on cost or customer service? Some Strategic Supply Chain Questions Achieving strategic fit by Competitive strategy targets one or more market segments: 1. Understanding the customer and SC uncertainty Wal-Mart : Provide high availability of a variety of products of reasonable quality at low prices. 2. Understanding the SC capabilities McMaster-Carr (sells MRO products, Offers 400000+ products thru a catalog and a web site) : Provide the customer with convenience, availability and responsiveness. 3. Achieving strategic fit Dell (build to order): Customization and variety at a reasonable cost, with customer having to wait about one week. Gateway (selling eMachine PCs thru retailers): Low price, fast response time, help in product selection but with limited variety 57 Q1: Should my supply chain focus more on cost or customer service? Q1: Should my supply chain focus more on cost or customer service? Impact of customer needs on implied demand uncertainty 1. Understanding the customer and SC uncertainty Identify the needs of customer segments targeted by competitive strategy • Quantity of prod needed in each lot, • response time the customers are willing to tolerate • The variety of products needed • The service level required • The price of the product • The desired rate of innovation in the product All can be translated into a single metric: implied demand uncertainty. Implied demand uncertainty is the uncertainty of the demand the SC is targeting. Customer tolerated response time greatly impacts the implied demand uncertainty. 58 customer need Causes implied demand uncertainty to range of quantity required increases increases bcs it implies greater variance in demand lead time decreases increases bcs there is less time to react variety of products required increases increases bcs demand /prod becomes more disagregate and lack of scale Nb of distribution channels increases increases bcs demand disagregated over more channels rate of innovation increases increases bcs new prod tend to be more uncertain required service level increases increases bcs the firm has to handle unusual demand surges Products with high implied uncertainty are : • less mature, high product margin, • but more difficult to forecast, more difficult to match demand and supply, more season-end marked down 59 60 15 Q1: Should my supply chain focus more on cost or customer service? Q1: Should my supply chain focus more on cost or customer service? • Along with demand uncertainty, it is important to consider the uncertainty resulting from the capability of the supply chain. • Supply uncertainty is also strongly affected by the life-cycle position of the product. New produscs have higher supply uncertainty bcs designs and production processes are still evolving. Locating your demand and supply on the uncertainty spectrum Predictable supply & uncertain demand Predictable supply and demand Impact of supply source capability on Supply uncertainty Supply source capability increases Unpredictable and low yields increases Poor quality increases Limited supply capacity increases Inflexible supply capacity increases Evolving production processes increases or somewhat uncertain supply & demand An existing auto model Commonplace goods: gasoline, salt Causes implied supply uncertainty to Frequent breakdowns or Predictable demand & uncertain supply High uncertainty Entirely new products 61 Q1: Should my supply chain focus more on cost or customer service? 62 Q1: Should my supply chain focus more on cost or customer service? 2. Understanding the SC capability Consider and categorize the characteristics of the SC responsiveness SC responsiveness • Respond to wide ranges of quantities demanded • Meet short lead times • Handle a large variety of products • Build highly innovative products • Meet a high service level • Handle supply uncertainty High Cost-responsiveness efficient frontier low Responsiveness comes at a cost. High SC efficiency : the inverse of the cost of making and delivering a product to the customer. low cost 63 64 16 Q1: Should my supply chain focus more on cost or customer service? Q1: Should my supply chain focus more on cost or customer service? What is the Right Supply Chain for Your Product? Locating your SC on the responsiveness spectrum Responsive supply chain High somewhat Somewhat efficicient efficicient responsive Dell, Sport Obermeyer Highly responsive GM, HP Integrated steel mills scheduled weeks or months in advance Little flexibility Hanes apparel: A traditional make-to-stock with production leadtime = weeks Automotive production: Delivery variety of products in couple of weeks DELL: Cost-effective supply chain Customer made PC in a few days P&G Low (functional products) Demand uncertainty High (innovative products) 65 Q1: Should my supply chain focus more on cost or customer service? 66 Q1: Should my supply chain focus more on cost or customer service? Two “Extreme” Types of Supply Chains Understanding Your Product Cost effective supply chain Market-responsive SC Primary purpose Supply predictable demand efficiently at the lowest possible cost Respond quickly to unpredictable demand in order to minimize stockouts, and obsolete inventory Manufacturing focus (capacity utilization) Lowest cost via high utilisation Maintain capacity flexibility for unexpected demand Inventory strategy Minimum inventory to lower cost Maintain buffer inventory for unexpected demand Transportation strategy Mainly rely on low cost modes Use responsive modes Approach to choosing suppliers Selection based on cost & quality On speed, flexibility & quality Lead-time focus Reduce but not at the expense of cost Aggressively reduce even if the costs are significant Two “Extreme” Types of Products -(a) Functional products -- predictable demand & long life cycle -- Easy to meet customer demand (revenue predictable) -- Competition based on price offered to customers -- Strategy: Supply chain design should focus on minimizing total cost (b) Innovative products -- unpredictable demand & short life cycle -- Difficult to forecast demand -- Difficult meeting customer demand (oversupply or stockout) -- Competition based on customer service (product variety, degree of customization, lead time) -- Strategy: Supply chain design should focus on customer service 67 68 17 Q1: Should my supply chain focus more on cost or customer service? Q1: Should my supply chain focus more on cost or customer service? Product Life Cycle Market Volume Life Cycle Differences Ramp-up and growth Ramp-up and growth Maturity Maturity End-of-life Demand uncertainty End-of-life Competition Basis Role of inventory Cost of shortage Cost of overage Time Product Type 69 Q1: Should my supply chain focus more on cost or customer service? 70 Q2: Do I need retailers in my supply chain? Key: To achieve strategic fit, a firm must tailor its supply chain to best meet the needs of different customer segments Dell (Direct sell to customers) HP-Compaq (Sell through retailers) Customers Customers Dell Suppliers Pull Push Pull Retail stores Compaq Push Suppliers To retain strategic fit, SC strategy must be adjusted over the life cycle of a product and as te competitive landscape changes. Pros: 1. Customization (larger variety of computers) 2. Elimination of retailers (lower facility/inventory cost) 3. Price flexibility Cons: 1. Customers have to wait to get the computer 2. Higher transportation costs (2~3% relative to price of a PC) 3. Lack of interaction between customers and sales personnel 71 72 18 Q3: Do I need warehouses/DCs in my supply chain? Q4: Perform all the functions in-house or outsource some? Why Use Warehouses/DCs? Functions in a Supply Chain Suppliers Manufacturers Distributors Retailers Consumers Three major functions: Manufacturing Plants DCs/Warehouses 1. Manufacturing Suppliers -- parts, subassemblies Manufacturers -- final products Retailers 2. Distribution and Warehousing Suppliers to manufactures Manufacturers to distribution centers 3. Retailing Retailers to customers 73 Q4: Perform all the functions in-house or outsource some? Past vs Now 74 Q4: Perform all the functions in-house or outsource some? What Functions to Perform? Past: Perform all stages of manufacturing, distribution & warehousing Based on your core competencies Example 1: Wal-Mart Core competencies: Retailing & Logistics -- Most efficient retail stores in the world -- Manage its own DCs and Warehouses -- Products provided by hundreds of suppliers Now: Perform key stages of manufacturing Example 2: General Motor’s Saturn Division Core competencies: Manufacturing -- Focus on manufacturing -- Logistics and transportation taken care by Ryder Logistics The anwser depends to a large extent on the competitive strategy. 75 76 19 Review of supply chain concepts • What is meant by the push-pull boundary? • make to stock vs. build to order • primary source of difficulties in SCM • SC processes • SC focus • product life-cycle considerations • two “extreme” type of products 77 20 Major Network Design Decisions • Facility role number of warehouses, centralised / decentralised, flexibility Chapter 2. Supply chain design • Facility location location of a warehouse, a plant • Capacity allocation size of a warehouse, space for products • Market and Supply allocation Determining which products customers will receive from each warehouse 1 Discussion: Factors that Influence Facility Location? 2 Discussion: Factors that Influence Facility Location? • Stratgic factors depending on the competitive strategy • Technological factors : high fixed cost (semiconductor, bottling plants of Cocacola)? flexibility of production offshore facility: low-cost facility for export • Macroeconomic factors: tariffs, tax inventives source facility: lower-cost facility for global production • Exchange rate and demand risk server facility: regional production facility • Political factors contributor facility: regional production facility with development skill • Infrastructure factors outpost facility: regional facility built to gain local skills • Competitive factors : positive externalities between firms – locating to split the market lead facility: facility that leads in development & process technology • Customer response time and local presence • Logistics and facility costs 3 4 1 Objective of Network Design Discussion: Impact of Increasing # of Facilities? balance service level against • Impact of increasing number of retail stores? (market share, ...) • annual system-wide costs, including production, purchasing, inventory, facility costs, transportation costs • Impact of increasing number of warehouses? (service level, inventory costs, overhead & setup costs, outbound transportation costs, inbound transportation costs) Goal: find a minimal-annual-cost configuration of the distribution network that satisfies product demands at specified customer service levels. 5 A framework for network design competitive strategy internal constraints capital, growth strategy existing network Production tech., cost, scale/scope impact support required, flexibility phase I SC strategy competitive environment production methods skill needs, response time A framework for network design Phase I: define a supply chain strategy/design • Define the firm’s broad SC design: SC stages, in-house or outsourced SC functions • Starts with the definition of the competitive strategy with the set of customer needs, the SC capabilities. • Forecast the evolution of the global competition & whether competitors in each market local or global • Identify constraints on available capital, whether growth by acquisition/building new facilities/partnering. Global competition tariff & tax incentives phase II Regional facility configuration regional demand, size, growth homogeneity local specifications political, exchange rate demand rate phase III Desirable sites 6 available infrastructure competitive strategy Factors : costs, labor, materials, site specific phase IV Location choices logistics costs, transport inventory coordination internal constraints capital, growth strategy existing network 7 phase I SC strategy Global competition 8 2 A framework for network design A framework for network design Phase II: define the regional facility configuration Phase III: Select a set of desirable potential sites • Select a set of desirable potential sites within each region • Based on an analysis of the infrastructure availability to support the production technologies • Hard infrastructure requirements: availability of suppliers, transportation services, communication, utilities, warehousing • Soft infrastructure requirements: availability of skilled workforce, workforce turnover, community receptivity • Identify regions where facilities will be located, their roles, their approximate capacity • Starts with the forecast of the demand by country, measure the size of the demand, homogeneity of demand across different countries • Identify whether economies of scale or scope can play a significant role in reducing cost, given available production technologies. • Identify demand risk, exchange risk, political risk , regional tarrifs, local production quota, tax incentives, export/import restrictions of different regional markets • Identify competitors in each region and make a case for whether to locate close to or far from competitors tariff & tax incentives Production tech., cost, scale/scope impact support required, flexibility competitive environment phase II Regional facility configuration regional demand, size, growth homogeneity local specifications political, exchange rate demand rate production methods skill needs, response time 10 Data for Network Design Phase IV: Location choices • Select a precise location and capacity allocation for each facility • Restricted to sites selected in Phase III • Designed to maximize total profits phase IV Location choices available infrastructure 9 A framework for network design Factors : costs, labor, materials, site specific phase III Desirable sites 1. All products including vol. & transp. mode 2. Location of customers, existing warehouses, DCs, plants, and suppliers 3. Demand for each product by customer location 4. Facility, labor, material costs by site 5. Transportation rates by mode 6. Inventory costs by site and as a function of quantity 7. Sale price of product in different regions 8. Taxes and tarrifs 9. Customer service goals 10. Shipment sizes by product and frequencies of customer delivery logistics costs, transport inventory coordination 11 12 3 Customers Impact of Aggregating Customers • Customers located in close proximity are aggregated using a grid network or clustering techniques. All customers within a single cell or a single cluster are replaced by a single customer located at the centroid of the cell or cluster (referred to as a customer zone). • The customer zone balances 1. Loss of accuracy due to over aggregation 2. Needless complexity • What affects the efficiency of the aggregation? Baltimore metro Western MD 1. The number of aggregated points, that is the number of different zones Eastern Shore DC metro 2. The distribution of customers in each zone. Southern MD Before clustering After clustering 13 Why Aggregate? 14 Recommended Approach for Aggregation • Use at least 150-200 aggregated points. • Make sure each zone has an equal amount of total demand. • Place the aggregated point at the center of the zone. In this case, the error is typically no more than 1%. One solution : aggregate according to Zip-code 15 16 4 Product Grouping Transportation rates • Companies may have hundreds to thousands of individual items in their production line • Rates are almost linear with distance but not with volume. • Collecting all data and analyzing it is impractical for so many products • Internal or external transportation fleet. • In practice, items are aggregated into a reasonable number of product groups, based on • For internal fleet, cost per mile per SKU (Stock Keeping Unit) can be easily estimated from the annual costs per truck, annual mileage per truck, annual amount delivered and truck’s effective capacity. 1. Distribution pattern 2. Product type 3. Shipment size 4. Transport class of merchandise • It is common to use no more than 20 product groups. 17 18 Mileage estimation Transportation rates • For external fleet, two modes : TL (truckload) and LTL (less than truckload). • Step 1: estimation of distance (in miles) Dab 69 • For TL, cost/mile is available from a zone-tozone cost table. lona lonb 2 lata latb (straightline, good for short distance) 2 2 2 lat latb lon lonb Dab 2(69) sin 1 sin a cos(lat a ) cos(latb ) sin a (for long distance) 2 2 • Step 2: Taking into account a circuity factor , i.e. Dab with = 1.3 for urban area and = 1.14 otherwise. • For LTL, three types of freight rates : class (classification tariff), exception; commodity • Example, Chicago (-87.65, 41.85) and Boston (-71.06, 42.36). Distance can also be checked with internet mapping systems. 19 20 5 Warehouse cost Warehouse cost • Storage costs: inventory holding costs which are proportional to average inventory level. • Handling cost : labor and utility costs which are proportional to annual flow • Fixed costs: depending on the warehouse capacity in a non linear way • Estimation of inventory level by inventory turnover ratio (taux de rotation des stocks) $1500000 inventory turnover ratio = $1200000 annual sales average inventory level $800000 20000 40000 60000 80000 Warehouse capacity (sq. ft) 21 Warehouse capacities 22 Potential warehouse locations • Step 1: according to the EOQ model, Potential warehouse locations must satisfy a variety of conditions: Warehouse capacity = 2 average inventory level • Geographical and infrastructure conditions • Step 2: taking into account empty space needed for access and handling • Natural resources and labor availability Warehouse capacity = 2 average inventory level) • Local industry and tax regulation Where > 1 is typically chosen to be 3. • Public interest On-hand Inventory Demand Rate Average Cycle Inventory Time 23 24 6 Service level requirements Network Design Solution Approaches • Maximum distance between each customer and the warehouse serving it • Heuristics • Percentage of population whose distance to their assigned warehouse is within a given distance. -- rules of thumb -- sub-optimal & usually no guarantee of performance • Exact Algorithms -- Guarantee optimal solutions for the problem -- Optimization techniques: Linear Programming (LP), Integer Programming (IP) 25 26 Optimization Approach Two Network Design Problems • Problem 1: Given facility locations (plants, warehouses), find the best distribution strategy from plants to warehouses to markets. • Problem 2: Given a set of candidate locations, find the best locations for warehouses and best distribution strategy from plants to warehouses to markets. Step 2 Step 1 Analyze intangible aspects List of potential sites Optimization: •Selection of one or more sites •Allocation of demand to sites Our focus here Possible Approaches: Heuristics and Exact Algorithms 27 Fall 2007 SCM 28 7 Phase II: network optimization model • Phase II: network optimization model To decide regions in which facilities are to be located based on regional demand, tarriffs, economies of scale, aggregate factor costs The capacity plant location model Example: To optimize the supply chain of the SunOil to meet its demand from five regions. Cost and demand data for SunOil are as follows: Demand region production and transportation cost per 1000000 units Fixed low N. S. supply region America America Europe Asia Africa cost capacity N. America 81 82 101 130 115 6000 10 S. America 117 77 108 98 100 4500 10 Europe 102 105 95 119 111 6500 10 Asia 115 125 90 59 74 4100 10 Africa 142 100 103 105 71 4000 10 Demand 12 8 14 16 7 Fixed • N = nb of potential plant/location/capacity • M = nb of markets or demand points • Dj = annual demand from market j • Ki = potential capacity of plant i • Fi = annualized fixed cost of opening plant i • Cij = cost of producing & shipping one unit from plant i to market j (production, inventory, transportation, tarriffs) high Decision variables cost capacity 9000 20 6750 20 9750 20 6150 20 6000 20 29 Phase III: Gravity center model • Yi = 1 if plant i is open, 0 otherwise • Xij = quantity shipped from plant i to market j To be continued. 30 Phase III: Gravity center model Identify the geographic location of potential sites Example: Locate a Distribution Center (X, Y) Locate a new assembly plant supplied by three plants and serving 5 markets. to serve N markets located at (Xn, Yn) with demand Dn and unit shipping cost Fn per unit of product and per kilometer. Algorithm: Step 1. Select any location (X, Y) of the DC Sources/markets supply sources buffalo memphis St Louis Markets atlanta boston jacksonville philadelphia new york Step 2. For each market n, evaluate its distance from DC X X n Y Yn Step 3. Obtain a new location (X’, Y’) dn 2 2 Dn Fn X n dn , Y ' X ' DF dn n n Dn FnYn dn DF dn n n Step 4. Replace (X, Y) by (X’, Y’) and repeat 2-3 till convergence. 31 transp cost quantity in $/ton.mile (Fn) Tons (Dn) Xn Yn 0,9 0,95 0,85 500 300 700 700 250 225 1200 600 825 1,5 1,5 1,5 1,5 1,5 225 150 250 175 300 600 1050 800 925 1000 500 1200 300 975 1080 32 8 Phase IV: Network optimization models Allocating demand to production facilities • Similar to regional configuration-Phase II but with facility locations given Locating plants: the capacity plant location model • Similar to regional configuration model but with location-specific costs and duties Locating plants: the capacity plant location model with single sourcing • Similar to the previous one but with each customer point sourced by one plant Locating plants and warehouse simultaneously 9 Case study I: Distribution network redesign for automotive industry Two scenarios: centralised / distributed distribution Facility location models Boat Plant 1 Boat Train HB1 HB2 Train Customer 1 DC 1 Truck Train Train Train Train Plant 2 TP1 Train Customer 2 Truck Train Train Train Truck Train Truck DC 3 Xiaolan Xie Customer ... Customer 16 Truck DC 4 Customer 1 Boat Plant 1 Customer ... DC 2 Plant 3 Boat Train HB1 HB2 Train Truck DC 1 Customer 2 Truck Train Train Plant 2 Truck TP1 Train Customer ... Truck Truck Customer ... Plant 3 Customer 16 1 Case study I: Distribution network redesign for automotive industry Requirements: Scenario analysis Complex operation processes (operating rules) Performance evaluation considering stochastic aspects Trade-off between costs and customer service level Leadtime balance between MTS and MTO products 2 Case study II: supplier selection in textile industry Enterprise @Europe AS-IS Boat +truck Delivery Supplier 1 Far East Distributor Customer Boat +truck Plane + truck Boat + planet + truck Supplier 1 Far East Enterprise Boat +truck TO-BE Plane + truck Delivery Supplier 2 Far East Boat + planet + truck Distributor Customer Truck Simulation is the only possible tool for faithful evaluation of the performances Truck Supplier 3 Europe Supplier 4 Europe 3 4 1 Case study II: supplier selection in textile industry Requirements: Strategic + operational decisions Supply chain network design Order assignment (split) ratio Replenishment level Huge nb of alternatives Dynamic in nature Demand seasonality Unstable transportation time Multiple criteria Total costs, Backlog ratio, service levels Existing models Simulation + optimisation 5 Models for location decision The p-Median Problem Set covering location model 6 The p-Median Problem warehouses SS-Capacitated Facility Location Problem Capacitated Facility Location Problem Multi-Commodity Capacitated Facility Location Problem Multi-Commodity Tri-Echelon Model Multi-Commodity Network Design retailers Problem: Locate p warehouses from m potential locations to serve n retailers. Assumptions: 1. No fixed cost for opening a warehouse 2. No capacity constraint on the warehouses 7 8 2 The p-Median Problem Min n m i 1 j 1 C ij X m s .t . X X ij Y m Y j 1 j : total transport ij cost Solution techniques: Lagrangian relaxation heuristics (Beasley93, Bramel& Simchi-Levi99) Heuristic rules: ADD (Kuehn&Hamburger, 63), DROP (Feldman et al., 66), SHIFT, ALA, VSM j Performances: P Y j 0 , 1 : warehouse X ation 1 ij j 1 ij The p-Median Problem Duality gap < 2,5 % on large pb with m = n = 200, p = 10 j opening 0 , 1 : warehouse j serving Poor LP-bounds Commercial package isn't efficient decision retailer i Possible extensions: Handling fee of products, Distance limit of links, Consideration of existing facilities. 9 10 SS-Capacitated Facilty Location SS-Capacitated Facilty Location SS-Location/Allocation Min transport n m i1 j 1 Single-Source Capacitated Facility Location Problem: C m Select warehouses from m potential locations to serve n retailers. warehouses total s .t . X ij j1 retailers diX i1 Y Assumptions: X j ij X ij cost j 1 f jY ij q jY costs : j j 0 , 1 : warehouse 0 , 1 : warehouse fixed m 1 n ij ation j opening j serving decision retailer i 1. Fixed cost for opening a warehouse 2. Maximal capacity of each warehouse 11 12 3 SS-Capacitated Facilty Location Capacitated Facilty Location Location/Allocation Problem Solution techniques: Lagrangian relaxation heuristics (Barcelo&Casanovas 84, Klincewicz& Luus 86 , Sridharan 93, Beasley 93, Bramel & Simchi-Levi, 99) Mutli-Source Capacitated Facility Location Problem: Branch & Bound (Neebe&Rao 83), LP+heuristics (Daskin&Jones 93) Select warehouses from m potential locations to serve n retailers. Performances: Related to the transportation problem Duality gap < 1,2 % on large pb with m = n = 100 warehouses retailers Poor LP-bounds Commercial package isn't efficient Assumptions: Possible extensions: Handling fee of products, Distance limit of links, Consideration of existing facilities. 1. Fixed cost for opening a warehouse 13 Capacitated Facilty Location Min total n transport ation m cost fixed costs C ij X ij 14 Capacitated Facilty Location Solution techniques: : m i 1 j 1 2. Maximal capacity of each warehouse Branch&Bound based on omega/delta rules (Akinc&Khumawala 77) f jY j j 1 Heuristic rules: ADD, DROP, SHIFT, ALA, VSM (Jacobsen 83) m s .t . X ij d i Lagrangian relaxation heuristics (Beasly 93) j 1 Extension to Multi-commodity CFLP problem: n X ij q j Y j Benders decomposition (Geoffrion&Graves 74) i 1 Y j 0 , 1 : warehouse opening X ij 0 : demande allocation decision Cross decomposition (Lee, 93) Lagrangian relaxation (Mazzola&Neebe 98) 15 16 4 Multi-Commodity Tri-Echelon Model Multi-Commodity Tri-Echelon Model (Pirkul & Jayaraman 96) Plants warehouses MIN Plant-warehouse shipping costs + Warehouse-Retailer transportation costs + Fixed plant openning costs retailers + Fixed warehouse openning costs s.t. - Each retailer supplies from 1 warehouse for each prod. - Warehouse capacity of any site is not exceeded - Plant supply capacity for each prod. is not exceeded Mutli-commodity Capacitated Facility Location Problem: - Material balanving at each warehouse Select warehouses and plants from a list of potential locations to serve n retailers with p products. - Nb of openned warehouses - Nb of openned plants 17 18 Multi-Commodity Tri-Echelon Mode Multi-Commodity Network Design Solution techniques: Commodity 1: Origin Lagrangian relaxation heuristics 20 Performances: Commodity 2: Origin 10 15 20 16 Duality gap < 2,7% for large problems (100 retailers, 20 warehouses, 10 plants, 3 prod.) Destination 2 Destination 1 Very poor LP-bounds Multi-Commodity Network Design Problem: Possible extensions: • warehouse handling fees, distance limit of any link, selection different warehouse types (small, medium, large) Select the set of capacitated arcs to open to minimize the fixed costs and flow transportation costs • Include many location models as special cases • Possibility to take into acount multi-modal logistics 19 20 5 Multi-Commodity Network Design Multi-Commodity Network Design Solution techniques: MIN C ijk X ijk k (i , j ) st Dual-ascent method for uncapacitated case (Magnanti et al 89, 94) f ij Y ij Lagrangian relaxation (Holmberg&Yuan 00) (i, j ) d k , k k X ij X ji d k , 0, j j X ijk U ij Y ij : if i o ( k ) Benders decomposition (Geoffrion & Graves 74, Magnanti&Wong84) if i d ( k ), : Flow conservati on at node i otherwise Branch & Bound (Crainic et al 93, Melkote & Daskin 01) Performances: Arc capacity Flow of commodity k Good LP-bounds for a disagregating formulation of the uncapacitated case Arc openning decision of arc (i, j) Very poor LP-bounds for capacitated case k X ijk 0 : Y ij 0 , 1 : 21 22 Fixed Charge Facility Location Uncapacitated Fixed Charge Facility Location Problems Select warehouses from n potential locations to serve n retailers in order to minimise opening costs and transportation costs similar to p-median SS-Capacitated Facilty Location 23 24 6 Fixed Charge Facility Location €100 15 €200 10 15 €150 18 €130 12 B €225 5 25 E €210 16 15 G 24 12 30 H s .t . €175 24 €165 13 12 X F €230 22 22 25 19 K J €215 20 Y 22 transport n n i1 j 1 n s .t . j1 ation h i d ij X ij cost fixed X ij j 1 serving j j 0 , 1 : warehouse 0 , 1 : warehouse ij f jY 1 i opening decision j serving retailer = unit transportation cost i 26 costs : Locate:At site that minimises sum of fixed and routing costs f jY j opening j 1 n 0 , 1 : warehouse ij : Heuristic algorithm: ADD X ij 1 X ij Y j Y j 0 , 1 : warehouse j Y X costs dij = distance from node i to warehouse j 21 Fixed Charge Facility Location total ij ij ij fixed cost n hi = demand of node i 25 Min X hid ation fj = fixed cost of opening a warehouse at site j L 19 j1 X 19 i1 j 1 I €125 19 n n 12 20 transport n 16 18 €190 11 total D C 24 Min 18 22 A Fixed Charge Facility Location Assign:Demand nodes to Find:Facility site that reduces nearest facilities total cost the most decision retailer yes j Locate:At cost reducing site Case where the warehouses are given, i.e. Y is given: Cost reducing site found? serve each node i from the closest warehouse, i.e. the warehouse j with minimal dij among all warehouse j such at Yj = 1. No STOP 27 28 7 Heuristic algorithm: ADD Locate the first warehouse A A B C D E F G H I J K L total fixl fix+0,35total 0 150 444 990 120 1440 198 528 624 880 1102 1340 7816 100 2836 B C D E 225 555 825 360 0 220 400 380 264 0 216 192 720 324 0 612 190 80 170 0 1248 720 288 864 363 451 649 275 768 448 736 192 546 260 312 312 1210 1276 1364 1034 1159 741 817 703 1220 780 680 860 7913 5855 6457 5784 200 130 150 225 2970 2179 2410 2249 F 900 520 360 216 180 0 627 672 156 1100 589 440 5760 175 2191 Heuristic algorithm: ADD dij*hi hi * min{dij, diI} H J K L I 270 495 720 600 870 1005 330 480 420 550 610 610 492 336 240 696 468 468 1062 828 432 1116 774 612 125 60 120 235 185 215 1368 1008 288 1200 744 528 0 165 495 242 440 671 240 0 480 592 400 736 585 390 0 494 247 247 484 814 836 0 418 880 760 475 361 361 0 399 1220 920 380 800 420 0 6936 5971 4772 6886 5576 6371 190 210 165 230 125 215 2618 2300 1835 2640 2077 2445 Locate the second warehouse G A B C D E F H J K L G I A 0 225 555 720 360 720 270 495 720 600 720 720 B 150 0 220 400 380 420 330 420 420 420 420 420 C 240 240 0 216 192 240 240 240 240 240 240 240 D 432 432 324 0 432 216 432 432 432 432 432 432 E 120 120 80 120 0 120 120 60 120 120 120 120 F 288 288 288 288 288 0 288 288 288 288 288 288 G 198 363 451 495 275 495 0 165 495 242 440 495 H 480 480 448 480 192 480 240 0 480 480 400 480 I 0 0 0 0 0 0 0 0 0 0 0 0 J 836 836 836 836 836 836 484 814 836 0 418 836 K 361 361 361 361 361 361 361 361 361 361 0 361 L 380 380 380 380 380 380 380 380 380 380 380 0 total 3485 3725 3943 4296 3696 4268 3145 3655 4772 3563 3858 4392 fix 265 365 295 315 390 340 355 375 165 395 290 380 fix+0,35total 1485 1669 1675 1819 1684 1834 1456 1654 1835 1642 1640 1917 fi + fI 29 Heuristic algorithm: ADD hi * min{dij, diI , diG} 30 Heuristic algorithm: ADD Locate the third warehouse hi * min{dij, diI , diG , diA} B C D E F H J K L A G I A 0 225 270 270 270 270 270 270 720 270 270 270 B 0 220 330 330 330 330 330 420 330 330 330 150 C 0 216 192 240 240 240 240 240 240 240 240 240 D 0 432 216 432 432 432 432 432 432 432 432 324 E 80 120 0 120 120 60 120 120 120 120 120 120 F 0 288 288 288 288 288 288 288 288 288 288 288 G 0 0 0 0 0 0 0 495 0 0 0 0 H 0 480 240 240 240 240 240 240 240 192 240 240 I 0 0 0 0 0 0 0 0 0 0 0 0 J 0 418 484 484 484 484 484 484 484 484 484 836 K 0 361 361 361 361 361 361 361 361 361 361 361 L 0 380 380 380 380 380 380 380 380 380 380 380 total 2695 2770 2647 2689 2929 2641 3145 2845 4772 2661 2718 2765 fix 455 555 485 505 580 530 355 565 165 585 480 570 fix+0,35total 1398 1525 1411 1446 1605 1454 1456 1561 1835 1516 1431 1538 Locate the fourth warehouse B C D E F H J L A G I K A 0 0 0 0 0 270 0 720 0 0 0 0 B 0 150 150 150 150 330 150 420 150 150 150 150 C 0 216 192 240 240 240 240 240 240 240 240 240 D 0 432 216 432 432 432 432 432 432 432 432 324 E 80 120 0 120 120 60 120 120 120 120 120 120 F 0 288 288 288 288 288 288 288 288 288 288 288 G 0 0 0 0 0 0 0 495 0 0 0 0 H 0 480 240 240 240 240 240 240 240 192 240 240 I 0 0 0 0 0 0 0 0 0 0 0 0 J 0 418 484 484 484 484 484 484 484 484 484 836 K 361 361 361 361 361 361 361 361 361 361 0 361 L 0 380 380 380 380 380 380 380 380 380 380 380 total 2695 2545 2307 2239 2479 2191 3145 2395 4772 2211 2268 2315 fix 455 655 585 605 680 630 355 665 165 685 580 670 fix+0,35total 1398 1546 1392 1389 1548 1397 1456 1503 1835 1459 1374 1480 31 32 8 Heuristic algorithm: ADD Heuristic algorithm: ADD hi * min{dij, diI , diG , diA , diK} Locate the fifth warehouse Locate the sixth warehouse (STOP) hi * min{dij, diI , diG , diA , diK , diD} B C E F H J L A D G I K A 0 0 0 0 270 0 720 0 0 0 0 0 B 0 150 150 150 150 330 150 420 150 150 150 150 C 0 216 192 216 240 216 240 216 240 216 240 216 D 0 0 0 0 432 0 432 0 432 0 432 0 E 80 120 0 120 120 60 120 120 120 120 120 120 F 0 288 288 288 288 288 288 288 288 288 288 288 G 0 0 0 0 0 0 495 0 0 0 0 0 H 0 480 240 240 240 240 240 240 240 192 240 240 0 0 0 0 I 0 0 0 0 0 0 0 0 J 0 418 418 484 418 418 418 418 418 484 418 836 K 0 0 0 0 361 0 361 0 0 361 0 0 L 0 380 380 380 380 380 380 380 380 380 380 380 total 2695 1662 1556 1812 1620 1524 3145 1512 4772 1394 2268 1432 fix 455 930 860 730 955 905 355 940 165 960 580 945 fix+0,35total 1398 1512 1405 1364 1522 1438 1456 1469 1835 1448 1374 1446 B C E F H J L A D G I K A 0 0 0 0 270 0 720 0 0 0 0 0 B 0 150 150 150 150 330 150 420 150 150 150 150 C 0 216 192 240 240 240 240 240 240 240 240 240 D 432 432 324 0 432 216 432 432 432 432 432 432 E 80 120 0 120 120 60 120 120 120 120 120 120 F 0 288 288 288 288 288 288 288 288 288 288 288 G 0 0 0 0 0 0 495 0 0 0 0 0 H 0 480 240 240 240 240 240 240 240 192 240 240 I 0 0 0 0 0 0 0 0 0 0 0 0 J 0 418 418 484 418 418 418 418 418 484 418 836 K 0 0 0 0 361 0 361 0 0 361 0 0 L 380 380 380 380 380 380 0 380 380 380 380 380 total 2695 2118 1880 1812 2052 1764 3145 1968 4772 1850 2268 1888 fix 455 780 710 730 805 755 355 790 165 810 580 795 fix+0,35total 1398 1521 1368 1364 1523 1372 1456 1479 1835 1458 1374 1456 33 Heuristic algorithm: DROP 34 A Lagrangian Relaxation Approach Initial problem Locate:At all candidate facility sites L* Find:Facility site whose removal Assign:Demand nodes to M in s .t . X j1 X yes Remove:Facility from cost reducing site n n i1 j1 n reduces total cost the most nearest facilities Y Cost reducing X ij j ij X ij n j 1 f jY j 1 j 0 , 1 : w a r e h o u s e o p e n i n g 0 , 1 : w a r e h o u s e s e r v i n g ij ij Y hid d e c isio n re ta ile r i site found? No STOP 35 36 9 A Lagrangian Relaxation Approach A Lagrangian Relaxation Approach Relaxed problem L M in n n i 1 j 1 h i d ij X ij n j1 Solving the relaxed problem f jY j n i 1 i 1 n X j 1 ij L s .t . X ij X ij Yj Yj 0 , 1 : w a r e h o u s e X s e rv in g re ta ile r i X Constraint relaxed: n j1 ij s .t . Min h i d ij i X ij f j j1 i 1 X ij Y j , Y j 0 , 1 , X ij 0 , 1 i 1 i d ij i d ij i X X ij ij f j Y j f j Y j n i i 1 n i 1 i Yj ij 0 , 1 : w a r e h o u s e o p e n i n g 0 , 1 : w a r e h o u s e s e r v i n g d e c is io n r e t a i le r i 38 A Lagrangian Relaxation Approach Solving the relaxed problem L h n i 37 A Lagrangian Relaxation Approach n n i 1 X ij 1 L() L* n n j1 s .t . Yj h M in n j1 0 , 1 : w a r e h o u s e o p e n i n g d e c i s i o n M in Dual Problem Y j i i 1 L * n L MAX s .t . L Min n n i 1 j 1 For each j, X ij Y j Y j 0 , 1 : warehouse Vj = fj + i min(0, hidij – i) X ij 0 , 1 : warehouse Yj = 1, if Vj < 0 and Yj = 0, if not. h i d ij X ij opening serving n j 1 f jY j n i 1 i 1 X ij j 1 n decision retailer j Xij = 1, if hidij – i < 0 and Xij = 0, if not. L() = j min(0, Vj ) + i i 39 40 10 A Lagrangian Relaxation Approach Dual gradient L i 1 Gradient method for solving dual problem Step 1: Initialisation: i = 0. n j 1 X ij Step 2: Solving the relaxed problem L() as L Min n n i 1 j 1 A Lagrangian Relaxation Approach h i d ij X ij n j 1 f jY j X ij Y j , Y j 0 , 1 , X ij 0 , 1 n i 1 i 1 X ij j 1 n Step 3: Derive a feasible solution from solution of L() and save the bestso-far solution Step 4: Derive gradients L()/ i. Step 5: Update : i i + s . L()/ i with s C L * L L i n i 1 2 Step 6: If not converge, i.e. iL()/ i , go to Step 2. 41 A Lagrangian Relaxation Approach Derive a feasible solution from solution of L() 42 A Lagrangian Relaxation Approach Derive a feasible solution from solution of L() Let Xij, Yj be the solution of the relaxed problem L(). Let Xij, Yj be the solution of the relaxed problem L(). If no warehouse is opened, i.e. Y = 0, open any warehouse and assign all demans to it If no warehouse is opened, i.e. Y = 0, open any warehouse and assign all demans to it If some warehouses are opened, i.e. Yj = 1 for some j, then assign each demand site to its nearest warehouse such that Yj = 1 . If some warehouses are opened, i.e. Yj = 1 for some j, then assign each demand site to its nearest warehouse such that Yj = 1 . Can also be solved by column generation. Can also be solved by column generation. 43 44 11 Hodder & Dincer (1986) The state of the art of mathematical models for Global Supply chain network configuration Cohen and Lee (1989) A deterministic non-linear MIP model, based on EOQ techniques, to develop a global resource deployment policy. Maximises the total after-tax costs for the manufacturing facilities and distribution centres. Subject to "managerial constraints" (resource and production constraints) and "logical consistency constraints" (feasibility, availability, demand limits). Determines product flow among vendors, plants, DCs, markets and transportation channels. International plant location with financial capacities. Include exchange rate fluctuations, market prices, international interest rates, and fixed costs. Maximise Expected after-tax profit - Risk aversion factor Variance of profit. Constraints: plant capacity, an upper bound on the market demands, financial constraints, and bounds on decision variables. A large-scale non-linear MIP solved by heuristics. Cohen and Moon (1990) Extend Cohen and Lee (1989) by developing a constraint optimisation model, called PILOT. Investigate the effects of parameters on supply chain cost. Consider the location problem of manufacturing facilities and distribution centres. Conclude that there are a number of factors that may dominate supply chain costs under a variety of situations, and that transportation costs play a significant role in the overall costs of supply chains operations. 12 Vidal and Goetschalckx (00) Kouvelis and Rosenblatt (1997) First attempt to include supplier reliability in a strategic production-distribution model. Consider a zero-echelon system. Include deterministic exchange rates, material flow linkage constraints, and a set of linearized suppliers' reliability constraints. Probability of being on time of all suppliers shipping to each plant is at least a specified target value. A mixed integer programming model to prescribe an optimal design for an international logistic network. Focus on policies that individual countries adopt to attract international trade, including taxation, subsidised financing and local content rules. Demonstrate the sensitivity of the global network design to even small changes in the policies of just one country. GSCM model of Arntzen et al (1995) Develop a MIP model, called global supply chain model (GSCM) with numerous experiences at DEC and later become JD Edwards. Can accommodate multiple products, facilities, stages, time periods, and transportation modes. Minimises a composite function of: activity days AND total cost. Input: bills of materials, demand volumes, costs and taxes, and activity day requirement Output: Nb and location of DCs, customer-DC assignment, Nb of stages, the product-plant assignment. GSCM model of Arntzen et al (1995) Use Global Bill Of Materials (GBOM) to represent possible material flows across the supply chain. May include Sourcing options depending on prod., location, and stage. Disk PC Box UK, Canada, Taiwan Media Motherboard CPU Chip Customer Region 1 Monitor Taiwan, Spain, Mexico Head Disk Array Memory Printer Japan, Italy, USA Software Germany, USA, HK System UK USA Taiwan Canada Parent Customer Region 2 Customer Region n Children 13 GSCM model of Arntzen et al (1995) Minimise [facility production & MH costs + transportation costs + Duty + Taxe + facility fixed charges + product line fixed costs + fixed costs associated with mfg methods - duty drawback – duty avoidance] GSCM model of Arntzen et al (1995) Production/inventory/shipping constraints + (1-)[Processing activity days for all product-facilities + Transit activity days for all prod. at all links for all modes] GSCM model of Arntzen et al (1995) Offset trade and local content constraints Local value added in nation n >= nNational sale values. Local value added in nation n >= nWorldwide sale values (USA, EU). Demand for each product-period-customer region. Balance of inventory, production and shipping for each product-period-facility. Total weight of products through a facility is limited Production at each facility-method is limited Lower bound <= Production, inventory and shipping at each facility-period-product <= Upper bound GSCM model of Arntzen et al (1995) Duty drawback and duty relief constraints Duty credit to total export of each product out of a nation-group. Duty credit to total export of each product into a nation-group 14 GSCM model of Arntzen et al (1995) GSCM model of Arntzen et al (1995) Solution method based on: Special structure of the problem. Elastic penalties to distinguish hard and soft constraints. Row factorisation for computation of cascaded material balance constraints (Brown & Olson 94) Constraint-branching enumeration. Integrality gap < 0.0005 percent (?). Unsolved issues of GSCM models System configuration constraints Limit on Nb of facilities making a product. Limit on Nb of facilities for each type. Limit on Nb of facilities using a method. Given opened facilities, product-facility decisions, and facility-method decisions. Thèse de l’Université de Metz Une approche d’optimisation basée sur la simulation The following factors are particularly important in designing global supply chains: Impact of uncertainties, Good estimation of network operation performance measures (KPI), Realist model for transportation facilities, Country related costs. pour la conception des chaînes logistiques : Applications dans les industries automobile et textile présentée par Hongwei DING sous la direction de Mrs. L. BENYOUCEF et X. XIE MACSI 15 Contexte et motivations (1) Contexte et motivations (2) De plus en plus, Sous l’effet de la globalisation Marchés instables (offre-demande, …) Systèmes manufacturiers, informatiques, … complexes Cycle de vie des produits/technologies très réduits … Conception et pilotage des chaînes logistiques Besoins industriels Chaînes logistiques / Supply chains Dans un futur proche, la concurrence ne sera pas entre entreprises mais entre chaînes logistiques. Outils d’aide à la décision Prise en compte des interactions entre les différents niveaux décisionnels Prise en compte des aspects aléatoires et dynamiques Prise en compte du niveau de service client Manques des méthodes existantes Modèles trop simplifiés et données trop agrégées Analyses peu réalistes (prise en compte insuffisante des impacts des aléas et des stratégies opérationnelles) Monocritère orientées coût [Aikens 85], [Verter et Dincer 92], [Geoffrion et Powers 95], [Slats et al. 95], [Vidal et Geotschalckx 97], [Beamon 98], [Schmidt et Wilhelm 00], [Snyder 04], etc. - Christopher 1992 61 62 Contexte et motivations (3) Contexte et motivations (4) Projet Européen GROWTH - ONE Optimization methodology for Networked Enterprises Période : 02/2001 – 02/2004 Partenaires industriels : Partenaires académiques : Nos contributions dans le projet ONE Développement d’une approche d’optimisation basée sur la simulation Définition des règles de pilotage pour la simulation des chaînes logistiques Implémentation et intégration de l’approche dans l’outil ONE Application de l’approche à deux cas d’étude Reconfiguration d’un réseau de production et de distribution Objectif: Développer des approches pour la conception et le pilotage des chaînes logistiques, s’appuyant sur des modèles réalistes, avec la prise en compte : des coûts, des délais, des taux de service des impacts sociaux et environnementaux Choix de fournisseurs 63 64 16 Plan de la présentation Problème de conception et pilotage d’une chaîne Conception Conception et pilotage des chaînes logistiques L’approche d’optimisation basée sur la simulation Application dans l’industrie automobile Application dans l’industrie textile Conclusions et perspectives Choix de fournisseurs Localisation des sites Choix des technologies utilisées … Pilotage Planification de l’approvisionnement Planification de la production Gestion de stock … Décisions stratégiques Décisions tactiques / opérationnelles Indicateurs de performances (KPIs) Financiers : Coût d’investissement/désinvestissement, coût de production, coût de transport, coût de stockage, etc. Logistiques : Taux de demande satisfaite par le stock, pourcentage des commandes livrés dans les délais souhaités, etc. 65 66 Etat de l’art (1) Etat de l’art (2) Choix de fournisseurs Conception des réseaux de production et de distribution Critères de choix Prix, délai de livraison, qualité, capacité de production, etc. AHP (Analytical Hierarchy Process) Programmation mathématique [Gaballa 74], [Weber et Current 93], [Weber et Ellram 93], etc. Analyse par simulation Modèles généraux de simulation des chaînes logistiques [Jain et al. 01], [Rossetti et Chen 03], [Herrmann et al. 03], [Biswas et Narahari 04] [Bagchi et al. 98], [Schriber et Brunner 03], [Kilgore 03], etc. Manques Localisation des sites et gestion de stock [Erlebacher et Meller 00], [Nozick et Turnquist 01], [Daskin et al. 02], [Shen et al. 03], etc [Narasimhan 83], [Dyer et Forman 92], [Korpela et Tuominen 96], etc. Localisation des sites [Geoffrion et Graves 74], [Cohen et Lee 85, 89], [Arntzen et al. 95], [Verter et Dincer 95], [Bel et al. 96], [Canel et Khumawala 97], [Jayaraman et Pirkul 01], etc. Méthodes existantes Modèles de programmation mathématique [Dickson 66], [Ellram 90], [Weber et al. 91], [Barbarosoglu et Yazgac 97], etc. [AHP] Les attributs importants associés à chacun des fournisseurs sont connus avec certitude. => Absence des incertitudes et de la dynamique de la chaîne Seuls les aspects relatifs aux fournisseurs sont considérés => Non prise en compte des aspects liés au transport, à la gestion de stock, etc. Analyses spécifiques [Towill et al. 92], [Petrovic et al. 99], [Bhaskaran 98], etc. Manques des modèles d’optimisation 67 [Vidal97] Peu de modèles considèrent les aspects stochastiques, e.g. les délais, etc. Non prise en compte des aléas. Aspects tactiques et opérationnels souvent ignorés lors de la conception. Monocritère orientés coût. 68 17 Objectif de la thèse Plan de la présentation Développement d’une approche de conception fondée sur un modèle réaliste avec la prise en compte des : Interaction entre les décisions à différents niveaux Incertitudes tout au long de la chaîne Décisions stratégiques Décisions tactiques Décisions opérationnelles Demande aléatoire Délai de transport aléatoire Fournisseur non-fiable, … Multicritères Dans le but de concevoir une chaîne qui est opérationnellement efficace Coûts d’investissement et opérationnels Conception et pilotage des chaînes logistiques L’approche d’optimisation basée sur la simulation Application dans l’industrie automobile Application dans l’industrie textile Conclusions et perspectives Coûts d’investissement/désinvestissement, coût d’approvisionnement, coût de production, coût de transport, coût de stockage, etc. Niveau de service client … Pourcentage des produits livrés dans les délais souhaités, etc. 69 70 Approche proposée (SIM-OPT) Module de simulation Algorithmes Génétiques Ce module a pour objectif d’évaluer les performances Module d’optimisation configuration + règles de pilotage Indicateurs de performance Une configuration de la chaîne étudiée Module de simulation 71 Fournisseur, usine, centre de distribution, client Connexion de transport Liaison d’information, entreprise Le système de pilotage associé Gestion de stock, planification de la production, gestion des ordres de production, etc. Affectation des ordres d’approvisionnement, répartition des produits transportés, etc. 72 18 Caractéristiques des entités Fournisseur Usine Centre de distribution Prix, délai d’approvisionnement, quantité minimale acceptée par ordre, etc. Capacité de production, délai de production, coût de production, etc. Affectation des ordres de production Capacité de stockage, coût de stockage unitaire, etc. Exemple de modélisation Client Gestion de stock Demande moyenne, fréquence des demandes, type de comportement, etc. Connexion de transport Liaison d’information Entreprise Capacité de transport, délai de transport, coût de transport unitaire, etc. Expéditeur des ordres, récepteur des ordres, etc. Ordonnancement des ordres Pour rôle la gestion des flux d’information => information centralisée et partagée Planification de la production Chargement des moyens de transport 73 Règles de pilotage d’une chaîne 74 Caractéristiques des règles Règles locales DIFFICULTE !!! Ordonnancement des ordres de fabrication et livraison Politique de gestion de stock Règles de chargement des moyens de transport Règles de départ des moyens de transport … Règles globales Planification du service client ‘qui sert qui’ Affectation des approvisionnement internes Affectation des approvisionnement externes Choix des connexions de transport utilisées … Création automatique des modèles de simulation pour différentes configurations. Module ’optimisation dd’optimisation Moduled’ Départ régulier Prêt à partir Départ suivant un planning configuration + règles de pilotage Indicateurs de performance Module Modulede desimulation simulation SOLUTIONS Règles génériques et flexibles capables de s’adapter à différentes configurations. 75 76 19 Exemple d’une règle de pilotage (1) Exemple d’une règle de pilotage (2) Le plus proche Le plus rapide Optimisation statique … Configuration 1: Configuration 2: DC1 + DC3 DC1 + DC2 + DC3 Règle choisie: Le plus proche C1 Décisions : Fermeture/ouverture des centres de distribution (DC) Hypothèse : Chaque client est servi par exactement un DC Question : Pour chaque client, quel DC assurera son service ? Règle choisie: Le plus rapide C2 DC1 C3 DC2 C4 DC3 Cn 77 Implémentation Module d’optimisation Exigences d’optimisation de l’approche SIM-OPT Un environnement pour la simulation par événements discrets Ordonnanceur des événements, moteur de simulation, etc. Un cadre de simulation des chaînes logistiques 78 Entités principales, règles de pilotage, etc. CFacility CTransportationLink 2 1..* C++ CTransportationNetwork CSupplier Optimisation combinatoire avec variables qualitatives et quantitatives Bruits importants des résultats de la simulation Capable d’apprendre des expériences de simulation Multicritères CInformationLink #id : int #name : string #location : string #existing : bool #closable : bool #active : bool #associate_cost : float #predecessor_list : list<int> #successor_list : list<int> #incoming_link_list : list<int> #outgoing_link_list : list<int> 2 Pourquoi les algorithmes génétiques ? 1..* Algorithme de recherche [Goldberg 89] De nature stochastique et itérative CEnterprise CManufacturer CDistributer CCustomer 79 Fitness = Qualité Besoin uniquement de la fitness pour guider la recherche des solutions optimales Cherche les solutions d’une population à une autre 80 20 Algorithmes génétiques multicritères (MOGA) Codage adopté Décisions stratégiques Non-dominated Sorting GA-II (NSGA-II) Un des meilleurs MOGAs jusqu’à présent Optimalité au sens Pareto Sélection élitiste Méthode de classement efficace [Deb et al. 02] Décisions tactiques/opérationnelles Configuration de la chaîne Choix de règles ‘1’ : ouverture du site Paramètres associés ‘0’ : fermeture du site ‘1’ : choix de la règle numéro 1 ‘2’ : choix de la règle numéro 2‘R’ : point de commande ‘Q’ : quantité commandée 81 Pseudo algorithme (1) 82 Pseudo algorithme (2) Étape1: Création d’une population initiale 1 chromosome = 1 solution candidate (configuration + règles) Étape3: Classement et sélection des chromosomes pour l’opération de croisement Étape2: Évaluations de toutes les solutions candidates Étape4: Opérations de croisement et mutation pour la reproduction d’une nouvelle population f1 = ∑(différents coûts); f2 = Niveau de service; f3 = …; etc. 0 1 1 0 Étape5: Lancement de la procédure de vérification et de réparation des chromosomes Fournisseur 1 Fournisseur 2 0 0 1 1 0 1 0 0 0 0 1 1 Fournisseur 3 Centre de distribution Fournisseur 2 KPI Fournisseur 3 1 0 0 0 Mutation Croisement Client Faisabilité ? Fournisseur 4 Centre de distribution Client Coût d’approvisionnement Coût de transport Coût de stockage Taux de demandes satisfaites … 83 0 0 1 1 0 0 1 0 0 1 1 0 0 1 1 1 0 0 0 0 Réparation 1 0 0 0 84 21 Pseudo algorithme (3) Plan de la présentation Étape1: Création d’une population initiale 1 chromosome = 1 solution candidate (configuration + règles) Étape2: Évaluations de toutes les solutions candidates Conception et pilotage des chaînes logistiques L’approche d’optimisation basée sur la simulation Application dans l’industrie automobile Application dans l’industrie textile Conclusions et perspectives f1 = ∑(différents coûts); f2 = Niveau de service; f3 = …; etc. Étape3: Classement et sélection des chromosomes pour l’opération de croisement Étape4: Opérations de croisement et mutation pour la reproduction d’une nouvelle population Étape5: Lancement de la procédure de vérification et de réparation des chromosomes Aller à Étape2 85 86 Cas d’étude Stratégies de production et de distribution Italie Reconfiguration d’un réseau de production et de distribution Ouverture/fermeture de sites Gestion de stock Usine1 train HB1 HB2 Client1 train camion camion camion train train train train train Client2 RDC1 train CDC Usine2 Affectation des ordres de production Allemagne bateau train camion camion camion camion Usine3 Véhicules standards 87 MTS MTO Make-to-Stock (MTS) Volume important Délai de commande court Coût de stockage important Client... RDC2 Client... RDC3 camion RDC4 Client16 Véhicules haut de gamme Make-to-Order (MTO) Délai de commande long Coût de stockage faible 88 22 Modélisation de la chaîne Analyse par scénarios Seules les décisions de fermeture des RDCs (2,3,4) Toutes les règles de pilotages sont fixées à l’avance Performances évaluées par simulation Production Gestion de stock multi-niveaux Affectation des clients aux DCs Configuration décentralisée actuelle Délais de transport aléatoires Configuration centralisée suggérée Demandes aléatoires Indicateurs de performances Coûts: désinvestissement, production, transport, stockage Temps de réponse moyen aux demandes clients Pourcentage des véhicules livrés dans les délais promis La configuration centralisée est préférée Espace de solutions limité ! Réduction du coût de stockage (économie d’échelle) Réduction du temps de réponse aux demandes clients 89 Optimisation basée sur la simulation Affectation par ratios 90 Expériences numériques et analyses (1) (R, Q) ou (s, S) Paramètres GAs Le plus proche Simulation Décisions à prendre Fermeture/ouverture des usines et des RDCs (2, 3, 4) Politiques de gestion de stock dans CDC et RDC1 Affectation des ordres de production aux usines ouvertes 91 NSGA-II adapté Nombre de générations : 2000 Taille de la population : 100 individus Sélection: Tournoi binaire Croisement: Un point et deux points, Pcross = 0.1 Mutation: Uniforme, Pmut = 0.9 Horizon de simulation : 3 ans Période de réchauffement : 3 mois 5 simulations pour chaque solution candidate Paramètres déterminés selon des tests effectués Indicateurs CPU Pentium IV 1.5 GHz, 256 Mb de mémoire Temps de calcul : 17.2 heures 92 23 Expériences numériques et analyses (2) Min. du coût total moyen Désinvestissement, production, transport, stockage Demandes générées Min. du temps de réponse moyen aux demandes clients Temps écoulé entre l’instant de la réception d’une commande et la livraison de cette commande. 5700 5600 Coût total moyen (€) g1 g2 5662 8.7 Politique de stockage Configuration CDC 5500 5400 10 12 14 Temps de réponse moyen (jour) 16 Poids d'affectation R Q s/R S/Q P1 P2 P3 1648 1200 1936 3099 4 6 4 5638 8.8 P1+P2+CDC+RDC1 (R,Q ) (s,S ) 8.9 P1+P2+CDC+RDC1 (R,Q ) (R,Q ) 1648 1200 1997 1998 4 6 7 9 P1+P2+CDC+RDC1 (R,Q ) (s,S ) 1648 1200 1997 3154 4 6 8 5349 9.1 (s,S ) 1934 309 1964 3134 8 8 4 9.3 9.4 P1+CDC+RDC1 P1+CDC+RDC1 P1+CDC+RDC1 (R,Q ) (R,Q ) (R,Q ) (s,S ) (s,S ) 1755 1973 1973 1201 329 322 1996 1953 1980 3166 3190 3195 4 3 4 6 6 7 4 7 9.5 P1+CDC+RDC1 (R,Q ) (s,S ) 1922 322 1962 3187 2 7 4 9.6 P1+CDC+RDC1 (R,Q ) (s,S ) 1816 320 1856 3026 2 7 4 10.2 Conception et pilotage des chaînes logistiques L’approche d’optimisation basée sur la simulation Application dans l’industrie automobile Application dans l’industrie textile Conclusions et perspectives 4 5295 5227 5218 5100 RDC1 (s,S ) RDC1 5573 5303 5200 P1+P2+CDC+RDC1 (R,Q ) CDC 5615 5340 Frontière Pareto 5300 8 Plan de la présentation P1+CDC+RDC1 (R,Q ) (R,Q ) 1642 322 1987 1218 5 6 3 5216 10.4 P1+CDC+RDC1 (R,Q ) (R,Q ) 1487 352 1990 1617 2 8 8 5211 10.5 P1+CDC+RDC1 (R,Q ) (R,Q ) 1685 351 1679 1672 7 2 6 5202 10.6 P1+CDC+RDC1 (R,Q ) (R,Q ) 1394 314 1965 1680 2 8 7 5193 11.1 P1+CDC+RDC1 (R,Q ) (R,Q ) 1188 388 1996 1669 1 8 3 5191 11.2 P1+CDC+RDC1 (R,Q ) (R,Q ) 1395 302 1965 1680 1 8 7 5181 11.4 P1+CDC+RDC1 (R,Q ) (R,Q ) 1395 341 1755 1652 1 6 6 5176 11.9 P1+CDC+RDC1 (R,Q ) (R,Q ) 1177 348 1965 1230 7 5 5170 12 P1+CDC+RDC1 (R,Q ) (R,Q ) 970 1305 1648 1454 8 5 5167 12.9 P1+CDC+RDC1 (R,Q ) (R,Q ) 664 335 1990 1675 5 8 4 5162 13 P1+CDC+RDC1 (R,Q ) (s,S ) 545 1305 1755 2963 8 7 2 5160 13.8 P1+CDC+RDC1 (R,Q ) (R,Q ) 638 339 1990 1669 5 2 4 5142 14.1 P1+CDC+RDC1 (R,Q ) (R,Q ) 613 328 1933 1652 3 8 7 3 3 93 Situation actuelle 94 Présentation du problème Choix de fournisseurs Actuel Moins cher Moyen Fournisseur Transport Plus rapide Délai de disponibilité pour le transport assez long Prix non compétitif par rapport à de nouveaux fournisseurs Délai de transport très long et aléatoire Client Demande saisonnière 95 Décisions à prendre Choix parmi les 4 fournisseurs candidats Paramètres de la politique de gestion de stock du DC1 Affectation des ordres d’achat (DC1 -> Fournisseurs sélectionnés) Répartition des produits sur les différentes connexions de transport 96 24 Modélisation Expériences numériques et analyses (1) Gestion de stock (R,Q) Paramètres GAs Demandes aléatoires Affectation par ratios NSGA-II adapté Nombre de générations : 2000 Taille de la population : 100 individus Sélection: Tournoi binaire Croisement: Un point et deux points, Pcross = 0.1 Mutation: Uniforme, Pmut = 0.9 Délais de transport aléatoires Répartition par ratios Règles de pilotage utilisées Simulation Gestion de stock par point de commande et quantité économique (R, Q) Règle d’affectation des ordres d’achat basée sur les ratios Règle de répartition des produits transportés basée sur les ratios Horizon de simulation : 3 ans Période de réchauffement : 3 mois 5 simulations pour chaque solution candidate Indicateurs CPU Pentium IV 1.5 GHz, 256 Mb de mémoire Temps de calcul : 15.6 heures 97 98 Expériences numériques et analyses (2) Min. du coût total moyen Engagement, approvisionnement, transport, stockage Max. du pourcentage des demandes client satisfaites par le DC1 sans délai Optimisation locales Seule optimisation des paramètres quantitatifs en fixant le choix de fournisseurs à l’avance. > 97% 95% 90% 85% < 97% 80% 23 25 27 29 31 33 35 Coût unitaire (€/paire) 37 39 100% f1(€) f2 (%) Fournisseurs choisis Poids d’affectation R Q 40,26 100 S2 + S3 L2(16, 23, 16) + L3(31) 6351 1652 39,51 99,99 S2 + S3 L2(23, 10, 23) + L3(31) 6465 1590 … … … … … … 28,71 97,19 S2 + S3 L2(16, 23, 21) + L3(31) 5363 1681 28,60 97,17 S2 + S3 L2(16, 19, 24) + L3(31) 5309 1625 28,56 96,90 S2 L2(24, 21, 22) 7196 1187 28,39 96,76 S2 L2(19, 30, 24) 6780 1216 … … … … … … 23,67 82,08 S2 L2(29, 24, 24) 6033 1005 23.66 81.86 S2 L2(25, 27, 17) 5760 1001 Pourcentage des demandes client satisfaites sans attente Pourcentage des demandes clients satisfaites sans attente 100% Fournisseur S2 Fournisseurs S2 + S3 > 97% 99% GOpt 98% LOpt_S2 LOpt_S2+S3 97% 28 41 30 32 34 36 38 Coût unitaire (€/paire) 99 Pourcentage des demandes client satisfaites sans attente Vérifications (1) 40 42 44 97% 96% 95% 94% 93% 92% 91% 90% 89% 88% 87% 86% 85% 84% 83% 82% 81% 80% < 97% GOpt LOpt_S2 LOpt_S2+S3 23 24 25 26 27 28 Coût unitaire (€/paire) 29 30 100 25 Vérifications (2) Plan de la présentation Recherche exhaustive Un problème simplifié avec seulement deux fournisseurs Frontière Pareto réelle obtenue par la recherche exhaustive Solutions obtenues en appliquant l’approche proposée Pourcentage des demandes client satisfaites sans attente 100% 95% 90% Conception et pilotage des chaînes logistiques L’approche d’optimisation basée sur la simulation Application dans l’industrie automobile Application dans l’industrie textile Conclusions et perspectives Frontière Pareto réelle 85% Resultats obtenus par optimisation 80% 17 19 21 23 25 27 29 31 33 35 37 39 Coût unitaire (€/paire) 101 102 Conclusions Innovations et originalités de l’approche Perspectives Évaluation fiable des performances de la chaîne sous des conditions réalistes Incertitudes bien couvertes Optimisation combinatoire et multicritères pilotée par GA Validation de l’approche sur deux cas d’étude Développement d’un module de simulation plus générique et complet Développement d’un logiciel d’optimisation/simulation et intégration dans l’outil ONE 103 Conception et implémentation de mécanismes de pilotage pour une simulation plus réaliste de la chaîne Développement d’autres techniques d’optimisation Comparaison avec des méthodes analytiques existantes sur des exemples numériques 104 26 Chapter 3. Managing economies of scale in a supply chain: cycle inventory Learning objectives: 1. Balance the appropriate costs to choose the optimal amount of cycle inventory in a supply chain 2. Understand the impact of quantity discount on lot size and cycle inventory 3. Devise appropriate discounting schemes for a supply chain 4. Understand the impact of trade promotions on lot size and cycle inventory 5. Identify managerial levers that reduce lot size and cycle inventory in a supply chain without increasing cost Role of cycle inventory 1 The role of cycle inventory in a supply chain Why do companies hold inventory? Why might they avoid doing so? • WHY? • A lot or batch size is the quantity that a stage of a SC either produces or purchases at a time. • The lot size is usually larger than the quantities demanded by the customer. • Cycle inventory is the average inventory in a SC due to this difference. – To take advantage of economic purchase order size : economy of scale (cycle inventory) – To meet anticipated customer demand – To account for differences in production timing (smoothing) – To protect against uncertainty (demand surge, price increase, lead time slippage) – To maintain independence of operations (buffering) • WHY NOT? – Requires additional space – Opportunity cost of capital – Spoilage / obsolescence 2 Key point : Cycle inventory exists in a SC bcs different stages exploit the economies of scale to lower total cost. The costs considered include: material cost, fixed ordering cost, and holding cost. 3 4 1 The role of cycle inventory in a supply chain Two Decisions in Inventory Management On-hand Inventory • Example: Consider a computer store selling an average of D = 4 printers a day but ordering Q = 80 printers from the manufacturer each time. • Cycle inventory = lot size/2 = Q/2 = 40 • Average flow time = cycle inventory/demand rate = 40/4 = 10 days (inventory holding time) • Inventory turnover (taux de rotation), inventory coverage (taux de couverture) Q Q/2 • When is it time to reorder? • If it is time to reorder, how much? Demand Rate, D Average Cycle Inventory, Q/2 Time 5 6 Economies of scale to exploit fixed costs: Economic Order Quantity Model Time Between Orders On-hand Inventory (Cycle Time) T = Q/D Economies of scale to exploit fixed costs: Economic Order Quantity Model Q Demand Rate, D Average Cycle Inventory, Q/2 Q/2 Reorder Point, R Place Order Receive order Time Lead Time, L 7 8 2 Economic Order Quantity Cost Model: Constant Demand, No Shortages Basic EOQ Assumptions Constant Demand Rate Constant Lead Time Orders received in full after lead-time. Constant Unit Price (no discounts) = = = = = = total annual inventory cost annual demand (units / year) order quantity (units) cost of placing an order or setup cost ($) cost per unit annual interest rate Total Annual Inventory = Cost TC = Annual Ordering Cost Annual + Holding Cost (D / Q) K + (Q / 2) Ic 9 10 (Constant Demand, No Shortages) Many orders, low inventory level Total Cost Carrying Cost On-hand Inventory Trade-off in EOQ Model: Inventory Level vs. Number of Orders Cost Relationships for Basic EOQ Q Time Q Few orders, high inventory level Ordering Cost Q* Order Quantity (how much) EOQ balances carrying costs and ordering costs in this model. On-hand Inventory • • • • TC D Q K c I Time 11 12 3 EOQ Results (How Much to Order) Determining When to Reorder (Constant Demand, No Shortages) Economic Order Quantity = Q* = • Quantity to order (how much…) determined by EOQ • Reorder point (when…)determined by finding the inventory level that is adequate to protect the company from running out during delivery lead time • With constant demand and constant lead time, (EOQ assumptions), the reorder point is exactly the amount that will be sold during the lead time. 2DK Ic Number of Orders per year = D / Q* Length of order cycle T = Q* / D Example: Total cost = TC = (D / Q*) K + (Q* / 2) Ic 13 14 Exercise EOQ Example Question: What if the company can only order in multiples of 12? (That is, order either 0 or 12 or 24 or 36, etc…)? D = 1,000 units per year BE CAREFUL! S = $20 per order IC = $8.33 per unit per month HOW MUCH TO ORDER? WHEN TO ORDER? Number of orders per year = Length of order cycle = T = Total cost = 15 16 4 Robustness of EOQ model Example: EOQ Robustness • Suppose that in the last problem, you have mis-specified the order costs by 100% and the holding costs by 50%. That is, Very Flat Curve - Good!! – S used in the computation = $40/order (actual cost = $20 / order) – IC used in computation = $150 / unit / year (actual = $ 100) – Then, using these wrong costs, you would have gotten Total Cost TC Q*-Q Q* Q*+Q Q' Order Quantity Would have to mis-specify Q* by quite a bit before total annual inventory costs would change significantly. 2(1,000)40 23.1 150 Your actual TC (computed substituting Q’ into TC, using correct costs of S = $20, and h = $100: TC 1,000 23 20 100 $2,019 23 2 Only 1% above minimum TC! 17 18 Key points KP1 : Total ordering and holding costs are relatively stable around the economic order quantity. A firm is often better served by ordering a convenient lot size close to the EOQ rather than the precise EOQ (robustness). KP2 : If the demand increases by a factor of k, the optimal lot size increases by a factor of k0,5 . The number of orders placed per year increases by a factor of k0,5. Flow time due to cycle inventory decreases by a factor of k0,5. KP3 : To reduce the optimal lot size by a factor of k, the fixed cost K must be reduced by a factor of k2. KP4 : Aggregating replenishment across products, retailers, or suppliers in a single order allow for a reduction of lot size of individual products bcs the fixed costs are now spread across differents aggregated entities. 19 Lot sizing with multiple products or customers 20 5 Assumptions : • In general, the ordering, transportation, and receiving cost of an order grows with the variety of products or pickup points. • A portion of the fixed cost of an order can be related to transportation (this depends only on the load but not on the product variety) • A portion of the fixed cost is related to loarding and receiving (this cost increases with variety on the truck) 21 Three approaches : • similar to EOQ model except the followings. • Di : annual demand for product i • S: order cost incurred each time an order is placed, independent of the variety of products included • si: additional order cost incurred if product i is included in the order. 22 Example: • Best Buy sells 3 models of computers, the Litepro, the Medpro, the Heavypro. • The annual demands are DL = 12000, DM = 1200, DH = 120. • Each model costs Best Buy 500$. • A fixed transportation cost of 4000$ is incurred each time an order is delivered. For each model ordered and delivered on the same truck, an additional fixed cost of 1000$ is incurred for receiving and storage. • Best Buy has an annual holding cost of 20%. 1. Each product manager orders his model independently (highest cost) 2. The product managers jointly order every product in each lot (easy to administer and implement, but not selective enough and expensive joint ordering if product specific order cost high) 3. Product managers order jointly but not every order contains every product, i.e. each lot contains a selected subset of products. 23 24 6 Approach 1 : Independent ordering • QL = 1095, QM = 346, QH = 110. • Oder frequencies : 11/year, 3,5/year, 1.1/year. • Total inventory cost = 155140 $ • Other measures of interest : cycle inventory, annual holding cost/prod, annual ordering cost, flow time. Approach 2 : Lots ordered and delivered for all • Combined fixed order cost/order : K = S + si • The optimal order frequency is (to explain, express total cost in T): k n* D I c i 1 i i i 2K • Example : n* = 9.75, annual inventory cost = 136528$, i.e. a reduction of 13% over approach 1. 25 Approach 3 : Lots ordered and delivered jointly for a selected subset of products Step 1. Identify most frequently ordered product assuming each being ordered independently. n max ni i ni Di I i ci 2 S si 26 Step 2. Identify the frequency with which other products are included. • Calculate the order frequency as a multiple of n • As the most frequently ordered product is in each order, the inclusion of a product i incurs an additional product specific fixed order cost of si. • Product i is included once every mi orders mi n ni The most frequently order products i* is included each time an order is placed 27 ni Di I i ci 2si 28 7 Step 3. Recalculate the order frequency of the most frequently order product n. n Step 4. For each product, evaluate the order frequency ni = n/mi and the total cost of such an ordering policy. D I c m 2S s m i i i i i Example : n = 11.47, mL = 1, mM = 2, mH = 5, annual total inventory cost = 130767$, a reduction of 4% over approach 2. i Why? (order cycle T for n, order cycle miT for i) 29 30 Key point: • A key to reducing cycle inventory is the reduction of lot size. • A key to reducing lot size without increasing costs is to reduce the fixed cost associated with each lot. • This may be achieved by reducing the fixed cost itself or by aggregating lots across products, customers, suppliers. • When aggregating, tailored aggregation is best, especially if product specific costs are large. Economies of scale to exploit quantity discounts 31 32 8 Introduction Two basic questions • Pricing schedule often displays economies of scale, with prices decreasing as lot size increases. • A discount is lot size based if the pricing schedule offers discounts based on the quantity ordered in a single lot. • A discount is volume based if the discount is based on the total quantity purchased over a given period. • Two commonly used lot size based discount schemes : all unit quantity discounts, marginal unit quantity discount or multiblock tarriffs • Given a pricing schedule with quantity discount, what is the optimal purchasing decision for a buyer? How does this affect the SC in terms of lot size, cycle inventories, flow times? • Under what conditions should a supplier offer quantity discounts? What are appropriate pricing schedules that a supplier should offer? 33 34 EOQ with all quantity discount Example • Pricing schedule : The unit purchase cost is Ci if the order quantity is at least qi with q0 = 0 < q1 < q2 < … < qr = ∞ and c0 > c1 > c2 > … • Drug Online (DO) is an online retailer of prescription drugs. Demand for vitamins is 10000 bottles per month. DO incurs a fixed order cost of 100$ each time an occurs is placed with the manufacturer. DO has an annual holding cost of 20%. • The pricing schedule of the manufacturer is the all unit discount schedule: • The retailer’s objective is to maximise its profit, i.e. minimise the sum of material, order, and holding costs. Order quantity 35 Unit Price ($) 0‐5000 3 5000‐10000 2,96 10000 or more 2,92 36 9 Solution Example (draw TC(Q)) Step 1. Determine the EOQ Qi for each price Ci Step 1: Q0 = 6324, Q1 = 6367, Q2 = 6410 Step 2: Order quantity 0‐5000 Q0 >= 5000, TC0 ignore 5000‐10000 5000 < Q1 < 10000, TC1 = 358969 $ 10000 or more Q2 < 10000, TC2 = 354520$ Step 3: Optimal order size = q2 = 10000, TC = TC2. Qi 2 DK Ici Step 2. Determine the total annual cost TCi for each price range Case 1: Qi >= qi+1, ignored as it is considered for Qi+1 Case 2: Qi < qi, D q TCi K i Ici Dci 2 qi Remarks : • Presence of quantity discount leads to Larger order size of 10000 units than the normal EOQ = 6324 • If S = 4$, order size under all unit discount schedule is still 10000 units and is 8 times the normal EOQ = 1265. Case 3: qi <= Qi < qi+1, D Q TCi K i Ici Dci 2 Qi Step 3. Determine the optimal order quantity. Unit Price ($) 3 2,96 2,92 37 38 EOQ with marginal quantity discount Solution • Pricing schedule : The pricing schedule contains specified break points q0 = 0 < q1 < q2 < … < qr = ∞. The marginal cost of a unit decreases at the break points to ci if the order quantity is at least qi with c0 > c1 > c2 > … Step 1. Determine the EOQ Qi for each price range Ci (why?) • The purchasing cost Vi of qi units is determined as follows: V0 = 0, Vi+1 = Vi + ci (q i+1 – qi), for i = 0, 1, … • Purchasing cost of an order of Q such that qi <= Q < qi+1 is: C(q) = Vi + ci(Q-qi) Qi Ici Step 2. Determine the total annual cost TCi for each price range Case 1: Qi < qi, Qi* = qi Case 2: Qi > qi+1, Qi* = qi+1 Case 3: qi <= Qi < qi+1, Qi* = Qi TCi Q 39 2 D K Vi qi ci Vi Q qi ci 1 V Q qi ci K I i Q Q D 2 Q Q D V Q qi ci 1 K I Vi Q qi ci i Q D 2 Q D Step 3. Determine the optimal order quantity. 40 10 Example (draw TC(Q)) V0 = 0, V1 = 15000, V2 = 29800 Step 1: Q0 = 6324, Q1 = 11028, Q2 = 16961 Step 2: Order quantity 0‐5000 Q0 >= 5000, TC0 = 363900$ 5000‐10000 Q1 > 10000, TC1 = 361780 $ 10000 or more 10000 < Q2, TC2 = 360365$ Step 3: Optimal order size = Q2 = 16961, TC = TC3. Key point Unit Price ($) 3 2,96 2,92 Remarks : • Much larger order size of 16961 units than the normal EOQ = 6324 • If S = 4$, order size 15755 is 12,5 times the normal EOQ = 1265. • There can be significant increase of order size and cycle inventory in the absence of fixed order costs as long as quantity discounts are offered. • Quantity discounts lead to a significant buildup of cycle inventory in a supply chain. • In many SC, quantity discounts contribute more to cycle inventory than fixed ordering cost. • Value of quantity discount in a supply chain? 41 42 Coordination to increase total SC profits Quantity discount for commodity products • For commodity products, a competitive market exists, the market sets the price, the firm’s objective is the lower costs. Why quantity discount? • For the retailer DO (Drug Online), its lot sizing decision is based on costs it faces. 43 44 11 Coordination to increase total SC profits Coordination to increase total SC profits Quantity discount for commodity products Quantity discount for commodity products DO : D = 10000 bottles vitamins/month, Kr = 100$, I = 20%, cr DO : TC_inv = D/Q*100 +0,2*3*Q/2 Manufacturer : = 3$, EOQ = 6324, TC_inv = 3795 $. Manufacturer : processing, packing & shipping DO orders • • • • • • • • • • • A line packing bottles at a steady rate matching the demand. Fixed setup cost Km = 250$ / order Production cost cm = 2$/bottle Holding cost = 20% Annual setup cost = 120000/6324*250 = 4744$ Annual holding cost = 6324/2*0,2*2 = 1265$ Total manufacturer setup & holding cost = 6009$ Total SC cost = 6009 + 3795 = 9804$ Fixed setup cost Km = 250$ / order Production cost cm = 2$/bottle Holding cost = 20% Total setup & holding cost = D/Q*250+0,2*2*Q/2 Total SC cost = D/Q*350 + (0,2*3+0,2*2)*Q/2 SC lot size Q = [2*D*350/ (0,2*3+0,2*2)]0,5=9165 Opt SC cost = 9165 $, gain = 9804 – 9165 = 638$ 45 46 Coordination to increase total SC profits Coordination to increase total SC profits Quantity discount for commodity products Quantity discount for commodity products Pricing scheme for achieving opt SC profit: • C = 3$/bottle if Q < 9165, C = 2.9978$ otherwise. DO : • has an incentive to order Q = 9165, • material cost reduction just enough to offset the increase of ordering & holding cost Total SC cost = opt SC cost = 9165 $ In practice, the manufacturer may have to share the increase of SC profit of 638$. Key point • For commodity products for which price is set by the market, manufacturer with large fixed costs per lot can use lot-size quantity discounts to maximise total SC profits. • Lot size-based discounts, however, increase cycle inventory in the SC. • The benefit of quantity discount decreases as the setup cost of the manufacturer decreases. (Importance of coordination between marketing & production) 47 48 12 Coordination to increase total SC profits Coordination to increase total SC profits Quantity discount for products for which the firm has market power Quantity discount for products for which the firm has market power When decisions are coordinated: • SC profil : Prof_SC = (p – Cs)(360000 – 60000p) • Consider the scenario in which the manufacturer has invented a new vitamin pill, vitaherb, for which few competitors exist. • The price p at which DO sells vitaherb influence demand. • Assume that: D = 360000 – 60000p. • Production cost Cs = 2$/bottle • The manufacturer decides the price Cr to charge DO Results: • p = 3 + Cs/2 = 4 • D = 120000, • Prof_SC = 240000$ 49 50 Coordination to increase total SC profits Coordination to increase total SC profits Quantity discount for products for which the firm has market power When decisions are made independently: • Manufacturer : MAXCr Prof_m = (Cr – Cs)(360000 – 60000p) • DO : MAXp Prof_r = (p – Cr)(360000 – 60000p) Results: • p = 3 + Cr/2, Cr = 3 + Cs/2 = 4, p = 5 • D = 60000, • Prof_m = 120000$, Prof_r = 60000$, SC profil = 180000 • Loss of 60000$ due to independent price setting, phenomenon known as double marginalization Quantity discount for products for which the firm has market power Key point • The supply chain profit is lower if each stage of the supply chain makes its pricing decisions independently, with the objective of maximizing its own profit. • A coordinated solution results in higher profit. 51 52 13 Coordination to increase total SC profits Coordination to increase total SC profits Quantity discount for products for which the firm has market power Quantity discount for products for which the firm has market power Pricing schemes to achieve the coordinated solution Pricing schemes to achieve the coordinated solution Two-part tariff : • The manufacturer charges its entire profit as an up-front franchise fee and then sells to the retailer at cost. • It is then optimal for the retailer to price as though the two stages are coordinated. Example : • Opt Prof_SC = 240000 $, Prof_DO = 60000$ (when no coordination) • Pricing scheme: charge the DO of the franchise fee of 180000$ and material cost of Cr = 2$/bottle. • DO maximises its profit if it sets p = 4$. Volume-based quantity discount: • The two-tariff is a volume-based quantity discount as the average material cost of DO declines as the purchase increases. • Design discount scheme to encourage DO the purchase the opt quantity 120000. • Pricing scheme : Cr = 4$ if the purchase < 120000, and Cr = 3.5$ otherwise. • DO optimal solution: p = 4, Prof_DO = 60000$, D = 120000, Prof_SC = 240000$. 53 54 Coordination to increase total SC profits Quantity discount for products for which the firm has market power Key point • For products for which the firm has market power, two-part tariffs or volume-based discounts can be used to achieve SC coordination and maximize SC profits. • Lot size-based discounts are not optimal even in the presence of inventory costs. In such as setting, either two-part tariff or a volumebased discount, with the supplier passing on some of its fixed cost to the retailer, is needed for the SC to be coordinated. • Lot size based discount tends to raise the cycle inventory. In contrast, volume based discounts are compatible with small lots. Use lot size based discount only when the supplier has high fixed cost. • Volume-based discounts suffer from orders peak toward the end of financial horizon. Volume discount based on a rolling horizon could 55 help. Short-term discounting: trade promotions 56 14 Introduction Introduction • Manufacturers use trade promotions to offer a discounted price and a time period over which the discount is effective. Key goals (from the manufacturer perspective) • Induce retailers to use price discount, displays or advertising to spur sales • Shift inventory from the manufacturer to the retailer and the customer • Defend a brand against competition • Ex: 10% off for any purchase from 12/15 to 01/25. • The goal is to influence retailers to act in a way that helps the manufacturer achieve its objectives. Need to understand the impact of trade promotion on the behaviour of a retailer and SC performances. 57 Introduction 58 Forward buy Retailer’s options facing a trade promotion 1. Pass through some or all of the promotion to customers to spur sales (increase the sales of the whole SC) 2. Pass through very little of the promotion to customer but purchase in greater quantity during promotion period to exploit the temporary reduction in price (forward buy and no increase of sales) 59 • d$: discount per product offered • Q* : EOQ at normal price • Qd : lot size ordered at discounted price Assumptions: • Discount is offered once • Retailer takes no action to influence demand • Qd is an integer multiple of Q*. 60 15 Forward buy Qd dD c d I cQ * cd Forward buy = Qd – Q* Qd • Why? Profit maximization (gain in fix cost, gain in purchase, loss of inventory cost). Q* Time 61 62 Forward buy Forward buy • Let T = Q/D be the period covered by short-term promotion • Cost during T without promotion Q Q * K Qc 0.5 Ic Q * T Example: • DO is a retailer selling vitaherb. Demand is 120000 bottles/year. The manufacturer currently charges 3$/bottle and DO has an annual holding cost of 20%. Fixed order cost K = 1000 $. What is the current lot size Q* of DO, cycle time, average flow time? • The manufacturer has offered a discount of 0.15$ for all bottles purchased by the retailer over the coming month. How many bottles should DO order given the promotion? • Cost during T with promotion K Q c d 0.5 I c d QT • Cost Gain during T Q Q Q * 1 K Qd 0.5 I c d Q IcQ * T • The optimal Q is obtained from Q Q 0 KD Q * 0.5 Ic Q * (optimal EOQ) 63 Answer: • Q* = 6324, Qd = 38236, Forward buy = 31912 Remark: • 5% discount causes the lot size to increase by 500+%. 64 16 Forward buy Impact on the demand Key point : • Trade promotions lead to a significant increase in lot size and cycle inventory because of forward buying by retailer. • This generally results in reduced SC profits unless the trade promotion reduces demand fluctuations. Example: • Assume DO selling at price p faces a demand of D = 300000 – 60000p. The normal price charged by the manufacturer is Cr = 3$/bottle. Ignoring the inventory related costs, evaluate the optimal response of DO to a discount of 0.15$ per bottle. Answer: • Without discount and Cr = 3$, p = 4$, D = 60000 • With discount of 0.15$ and Cr = 2.85$, p = 3.925$, D = 64500. • 7.5% increase in demand, DO pass only half of the trade promotion discount to Customers. 65 66 Impact on the demand Key point • Faced with a short term discount, it is optimal for retailers to pass through only a fraction of the discount to the customer, keeping the rest for themselves. • Simultaneously, it is optimal for retailers to increase the purchase lot size and forward bur for future periods. • Thus, trade promotions often lead to an increase of cycle inventory in a SC without a significant increase in customer demand. • Trade promotion should be designed so that retailers limit their forward buying and pass along more of the discount to end customers. Managing multiechelon cycle inventory 67 68 17 A mutliechelon distribution supply chain One manufacturer supplying one retailer (Instantaneuous production, lotsize Q) No synchronization : production right after delivery, average INV = 3Q/2 mfg inventory retailer inventory stage 1 stage 2 stage 3 shipping production stage 4 A multiechelon supply chain has multiple stages and possibly many players et each stage. Goal: decrease the total costs by coordinating orders across the SC Synchronization : production after before delivery, average INV = Q/2 69 70 distributor replenshment order arrives Simple multiechelon with one player at each stage Distributer replenishes every two weeks Integer replenishment policy: • lot size at each stage = integer multiple of the lot size of its immediate customer • Coordination of ordering across stages allows for a portion of the delivery to a stage to be cross-docked on to the next stage • Extent of cross-docking depends on the ratio of fixed ordering cost S and holding cost H at each stage. The closer the ratio, the higher the optimal percentage of cross-docked product. retailer shipment is cross-docked Retailer replenishes every week retailer shipment is from inventory Retailer replenishes every two weeks retailer shipment is cross-docked Shown to be quite close to optimal. retailer shipment is cross-docked Retailer replenishes every four weeks 71 18 One distributor supplies multiple retailers Integer replenishment policies Integer replenishment policy: • Distinguish retailers with high demand from those with low demand • Group retailers such that all retailers in one group order together • fr = n*fd or fd = n*fr, for each retailer r where n is an integer and fr and fd are retailer and distributor order frequencies • Each player orders periodically with reorder interval equal to an integer multiple of some base period • Divide all parties within a stage into groups such that all parties of a group order from the same supplier and have the same reorder interval • Set reorder intervals across stages such that the receipt of a replenishment order at any stage is synchronized with the shipment of a replenishment order to at least one of its customers. The synchronized portion can be cross-docked. • For customers with a longer reorder interval than the supplier, make the customer reorder interval an integer multiple of the suppliers' interval and synchronize their replenishment to facilitate cross-docking • For customers with a shorter reorder interval, make the supplier's reorder interval an integer multiple of the customer's interval and synchronize the replenishment • The relative frequency of reordering depends on the setup cost, holding cost and demand at different parties. Shown to be near optimal by Roundy. 73 Key points 74 Integer replenishment policies • Integer replenishment policies can be synchronized in multiechelon supply chains to keep cycle inventory and order costs low. • Under such policies, the reorder interval at any stage is an integer multiple of a base reorder interval. • Synchronized integer replenishment policies facilitate a high level of cross-docking. • Whereas the integer policies synchronize replenishment and decrease cycle inventories, they increase safety inventories because of the lack of flexibility with the timing of a reorder • These policies make the most sense for supply chains in which cycle inventories are large and demand is relatively predictable. 75 • Divide all parties within a stage into groups such that all parties of a group order from the same supplier and have the same reorder interval • Set reorder intervals across stages such that the receipt of a replenishment order at any stage is synchronized with the shipment of a replenishment order to at least one of its customers. The synchronized portion can be cross-docked. • For customers with a longer reorder interval than the supplier, make the customer reorder interval an integer multiple of the suppliers' interval and synchronize their replenishment to facilitate cross-docking • For customers with a shorter reorder interval, make the supplier's reorder interval an integer multiple of the customer's interval and synchronize the replenishment • The relative frequency of reordering depends on the setup cost, holding cost and demand at different parties. 76 19 Echelon inventory • Ordering policies based on echelon inventory (s, S), (r, Q) • Problems: where to locate the inventory, how to allocate the inventory warehouse echelon inventory supplier warehouse warehouse echelon lead time 77 20 Chapter 4. Managing uncertainty in a supply chain: safety inventory Learning objectives: • Understand the role of safety stock in a SC • Identify factors that influence the required level of safety stock • Utilise managerial levers to lower safety stock and improve product availability Role of safety inventory in a supply chain 1 Two Decisions in Inventory Management 2 EOQ Model when there is no uncertainty Time Between Orders On-hand Inventory (Cycle Time) T = Q/D • When is it time to reorder? • If it is time to reorder, how much? Q Demand Rate, D Average Cycle Inventory, Q/2 Q/2 Reorder Point, R Place Order Receive order Time Lead Time, L How many to order : Q = EOQ When to order: reorder point R = L.D 3 4 1 Effects of Demand / Lead Time Variability on Reorder Point (When) Role of satefy inventory On-hand Inventory • Inventory carried to satisfy demand that exceeds the forecasted demand. • Needed because of uncertain demand and uncertain product supply. Variable demand Expected demand at average demand rate d Q QUESTION: How much inventory is needed during lead time L? s Q/2 Safety Stock level Cycle Inventory Place order safety Inventory Receive order L KEY POINT: s is larger when there is uncertainty about demand or L Time 5 6 Safety Stock Safety Stock • Stock carried to provide a level of protection against costly stockouts due to uncertainty of demand during lead time • Safety Stock Criterion: – stockouts occur when demand during lead time (DL) … – service level (1 - )100%. • Stock outs occur when – Demand over the lead time is larger than expected Inventory Level s= ROP • DL is a random variable. What kind of probability distribution? Expected demand Time 7 8 2 Computing s … Assumption: Demand over lead-time is normally distributed 1- Probability {Demand over lead-time < s} = Determining appropriate level of safety inventory and choosing an inventory control policy Service Level Probability distribution of demand over L 1- s 9 10 Computation of Variance for Demand over Lead Time: Variability Comes From Two Sources Computing s: Normal Distribution Probability distribution of demand over L: 1: Suppose only demand di in day i is variable; lead time is constant at AVGL DL d1 d 2 ... d AVGL Mean = ; Std Dev = 1- AVGL times Var {DL } Var {d1 d 2 ... d AVGL } Var {d1 } Var {d 2 } ... Var {d AVGL } AVGL STD 2 s di’s are independent di’s are identically distributed 2: Now, suppose only lead time is variable; daily demand is constant at AVG 1- .90 .95 .98 .99 .999 z 1.28 1.65 2.05 2.33 3.09 z 1- s s z 3: Adding the two terms, we get to our result From normal table or, in Excel, use: =normsinv (0.90) 0 DL L AVG Var {DL } Var {L AVG } AVG 2 Var {L} AVG 2 STDL2 Var {DL } AVGL STD 2 AVG 2 STDL2 z 11 12 3 More specifically…. Mean demand over LT Safety factor (std normal table) Standard deviation of demand over LT Example: Safety stock SS s AVG AVGL z STDL2 AVG 2 STD 2 AVGL • Consider inventory management for a certain SKU at Home Depot. Supply lead time is variable (since it depends on order consolidation with other stores) and has a mean of 5 days and std deviation of 2 days. Daily demand for the item is variable with a mean of 30 units and c.v. of 20%. Find the reorder point for 95% service level. AVG 30; Note: •If lead time is constant, STDL 0 • If demand is constant, STD 0 STD 0.2(30) 6 AVGL 5; STDL 2 95% service level z = 1.64 s AVG AVGL z STDL2 AVG 2 STD 2 AVGL Note: This is a very good approximation even when demand is not normally distributed. 30 5 1.64 22 302 62 5 150 100.8 251 13 Inventory The (s,S) Policy: Fixed Ordering Costs The (s,S) Policy: Fixed Ordering Costs • Order when: inventory position (IP) drops below s • Order how much: bring IP to S S sR • Compute s exactly as in the base-stock model: Average demand during lead time s AVG AVGL z STDL2 AVG 2 STD 2 AVGL Safety Stock L Order placed 14 Order arrives • Compute Q using the EOQ formula, using mean demand D = AVG (be careful about units…): Time Q s should be set to cover the lead time demand and together with a safety stock that insures the stock out probability is within the specific limit (When to reorder). S depends on the fixed order cost – EOQ (How much) 2 K AVG h • Set S = s + Q 15 16 4 Example: (s,S) Model Summary of Inventory Models Use EOQ •How much: EOQ formula •When: inventory level drops to d*L • Consider previous Home Depot example, however, there are fixed ordering costs, which are estimated at $50. Assume that holding costs are 15% of the product cost ($80) per year. Also, assume that the store is open 360 days a year. s 251 yes (from previous calculations) h (.15)80 / 360 0.0333; K 50; 2 AVG K 2(30)50 Q 300 h 0.0333 Is demand rate and lead time constant? AVG 30 S s Q 251 300 551 Are there fixed ordering costs? yes no no Use (s, S) policy •How much: Q = S – s (Q is from EOQ formula) •When: IP drops below s (basestock policy formula) Use base stock (s) policy •When: IP drops below s •How much: necessary to bring IP back to s 17 18 Periodic review Periodic review Review interval: T, i.e. order every T time units. Review interval: T, i.e. order every T time units. Use modified EOQ •How much: EOQ covering N periods •When: inventory level drops to d*min(L, T) R ( L T )d z T L yes Is demand rate and lead time constant? Reorder point R = Stock level to cover demand DT+L Are there fixed ordering costs? no yes Use (s, S) policy •How much: Q = S – s (Q is from EOQ formula) •When: IP drops below s (basestock policy formula) Safety stock SS z T L T no Use base stock (s) policy •When: IP drops below s •How much: necessary to bring IP back to s L Point de commande R ou s = stock nécessaire sans commander à l'instant 0 pour couvrir la demande DT+L jusqu'à l'arrivée de la prochaine commande à la date T+L 19 S - s = quantité économique 20 5 Risk Pooling • (safety) stock based on standard deviation – square root law: stock for combined demands usually less than the combined stocks (depends on what?) Risk pooling or impact of aggregation • Example: independent demand X2 Y X2 Y2 X Y X2 Y2 X Y 21 22 HP Example: Benefits of a Universal Product Risk Pooling • (safety) stock based on standard deviation Because of a different power supplies, HP had two laser printers, one for Europe and one for N. America. A universal product (with a universal power supply) has been proposed, but costs $30 extra. Is it worthwhile? N. America Europe N(200,60) N(150,50) – square root law: stock for combined demands usually less than the combined stocks (depends on what?) • Centralizing inventory control reduces safety stock, hence average inventory level for the same service level. (This phenomenon is called risk pooling) • works best for Consider z = 2 (98% of service level) – negatively correlated demand. Why? – high coefficient of variation, which increases required safety stock. What is the difference is safety stocks required? assume independent demand seen by HP (NA and Europe) • other kinds of risk pooling: across markets, products, time 23 24 6 Example Risk Pooling Example • Consider two systems: Warehouse 1 Market 1 Warehouse 2 Market 2 Decentralized System: Two warehouses, each serving one customer Supplier Market 1 Supplier Centralized System: One warehouse, serving both customers Warehouse Market 2 Questions: Q1: For the same service level, which system will require more inventory? Q2: For the same total inventory level, which system will have better service? AVG STD SS s Q S Average Inventory Warehouse 1 39.3 13.2 25.08 65 132 197 91 Warehouse 2 38.6 12.0 22.8 62 131 193 88 Centralized Warehouse 77.9 20.7 39.35 118 186 304 132 Safety Stock SS = z ·STD · L Reorder Point s = AVG·L + SS Order Quantity Q = sqrt(2K·AVG/h) Order-up-to-level S = s + Q Average Inventory SS + Q/2 Decentralized system: total SS = 47.88 total avg. invent. = 179 25 Centralized vs. Decentralized Calculating demand variability of centralized system Warehouse 1 Warehouse 2 Market 1 Market 2 Market 1 Warehouse Market 2 d1: (1, 12) d2: (2, 22) : correlation coefficient of d1, d2 2 = 1 22 + 212, where -1 1 d1+d2: (, 2) = 1 + 2 = ?? Conclusions: 1. Stdev of aggregated demand is less than the sum of stdev of individual demands 2. If demands are independent or negatively correlated, the std of aggregated demand is much less 26 2+ Inbound transportation cost (from factories to warehouses) 1. If d1, d2 positively correlated, > 0 2. If d1, d2 are independent, = 0 3. If d1, d2 negatively correlated, < 0 1+2 12 22 N.C. Decentralized Facility/Labor cost 1+ 2 -1 Centralized Outbound transportation cost (from warehouses to retailers) Inventory cost Responsiveness to customers (lead time) Safety stock, overhead, economy of scale, lead time, service, transportation costs. 0 Ind. P.C. 1 27 Hybrid policy possible depending on the products (low demand product centralised and high demand products at local warehouse) 28 7 Design for Supply Chain • Also called: design for logistics -- DFL Take supply chain costs into account when designing product and manufacturing processes Design for Logistics • Approaches 1. Economic Packaging and transportation (obvious, design products that can be packed compactly and efficiently, IKEA) Modularity & Component Commonality Postponement 2. Modularity & Component Commonality 3. Postponement Benefits: Reduced inventory and transportation costs in the supply chain 29 Approach 2: Modularity/Component Commonality • Modularity allows components of the product be decoupled and manufactured in parallel. 30 Approach 2: Modularity/Component Commonality • Some modules are common across different products (example?) Serial processing PC board Europe Printer Customer (Europe) 1 Asia printer housing (motor, printhead, housing) 2 Without component commonality Parallel processing Europe Asia 2 With component commonality • Benefits: Customer (Europe) PC board 1 -- Modularity allows parallel processing reduced lead times -- Risk pooling reduced inventory cost -- Fewer components reduced inventory handling and procurement costs Printer Housing 31 32 8 Modular vs. Integral Design Why is Modular Design Preferred, from a Supply Chain Standpoint? Modular design One-to-one mapping between functional elements and components Interfaces between components not coupled • Example: Consider Chrysler. It needs to renew its Durango and Cherokee lines. Currently, each car has very little component commonality with the other, since both use integral designs. Chrysler is considering a modular platform design, in which 60% of the components, in terms of dollar value (chassis, transmission, underbody, etc.) are common to the new Cherokee and Durango. Suppose the monthly demand for the Cherokee, in 000s, N(50,202), whereas the demand for the Durango is N(40,202). Compute the monthly holding cost savings regarding inventory safety stock, if the modular design is used. Assume that each car costs Chrysler US$15,000 to manufacture, and that lead-time across components is constant at one month (for simplicity). Consider annual holding cost of a component to be 12% of the component value. Assume a 95% service level. Integral design Complex mapping from functional elements to components Interfaces between components are coupled integral modular 33 Illustration of Chrysler Product Strategies 34 Solution to the Chrysler Example AVGC 50, STDC 20, AVGD 40, STDD 20, L 1, z 1.64 Integral Design: Current: Integral Designs Safety stock C = z STDC2 L 1.64 202 (1) 33 Safety stock D z STDD2 L 1.64 202 (1) 33 Total safety stock: 33 + 33 = 66 Modular Design: AVG 50 40 90, STD 2 202 28.3 Safety stock = z STD 2 L 1.64 28.32 (1) 46.4 Proposed: Modular Design Monthly inventory holding cost savings = (66-46.4)*15,000*0.60*(0.12/12) =1,764 (in 000s), or US$ 21 million per year! 35 36 9 Approach 3: Postponement Approach 3: Postponement • Design product and manufacturing processes so that decisions about specific products can be delayed as late as possible • Concepts for delayed differentiation: -- For example, first manufacture a generic product, then differentiate it to make specific products -- Aka. Delayed product differentiation -- DPD Without postponement Raw materials components – – – – Resequencing product manufacturing steps Commonality Modularity Standardisation With postponement Raw materials components Generic product Different products Different products • Benefits: – Reduced demand uncertainty (why?) Higher service level or/and reduced inventory cost 37 38 Benetton Manufacturing Process Postponement Example: Benetton • A world leader in knitwear Old Sequence Knitting Wool Plant in Castrette, near Treviso. Knitting division. Computerized knitting loom capable of automatically producing the most complex product designs New Sequence Dyeing Purchase Yarn Purchase Yarn Dye Yarn Knit Garment Parts Finish Yarn Join Parts Knit Garment Parts Dye Garment Join Parts Finish Garment This process is postponed Dyeing vats for the finished knitted product. 39 40 10 Benetton Example: Evaluating the Value of Dyeing Postponement Process Redesign for Supply Chain: Postponement at Benetton Dye yarn only after the season’s fashion preferences become more established (knit lead-time much longer than dyeing lead-time). Consider the previous example (say, a sweater). Suppose demand for each of the 4 different sweater colors has a mean of 10000. Dyeing time is 1 month; knitting time is 6 months. Due to the short season and long production lead times, there is only one production run before the season. The standard deviation of demand (forecast error), however, depends on how long the forecast is done before the season starts. If seven months before the season, the standard deviation is 6000. If done one month before the season (color preferences are well known by then), the standard deviation is 1000. Compute approximate inventory holding cost savings in WIP as a result of dyeing postponement. Assume that each sweater costs $30, annual holding cost is 12% of the product value, and 95% service level. Example: single product; four colors knit Dyeing operations postponed dye dye knit Outcome: Reduces demand uncertainty & inventory 41 Benetton Example Solution Fall 2005 42 HP Printers Cases Quantities for yarn are based on the scheduled production for the final product, which is shown below: Case 1 1) What were the problems (crises) facing HP? What were the causes? 2) Resolving the crises – What alternatives did HP consider for solving the problems? For each color, STD = 6,000, z =1.65 (consider dyeing only) SS = z*STD = 9840, total SS = 4(9,900) = 39,600 Case 2 Aggregate demand (pooling), STD = 4(1000)2 = 2,000, z =1.65 SS = z*STD = 3300 savings in monthly holding cost: 36300*$30*(0.12/12) = $11K, for 100 SKUs, annual savings exceed $10M ! 43 44 11 Inventory-Service Crisis HP DeskJet Printer Supply Chain What is the crisis? US DC Customer Europe DC Customer Far East DC Customer Vancouver Plant Suppliers Excess inventory in some products Shortages in other products What caused it? Poor demand forecasting Large product variety due to many markets Long shipping lead time from plant to DC Inventory levels set incorrectly HP 45 Setting the Right Level of Inventory 46 Drivers of Safety Stock How are current inventory levels set at each DC? SS z•STD• Rule of thumb How should they be set? -- Use (s, S) policy -- Reorder point: s = AVG · L + z · STD · Safety stock Drivers of safety stock: L -- Use EOQ model to determine optimal order quantity – Q* L 1. Service level Z 2. Demand uncertainty STD 3. Order lead time L 2 · K·AVG = h – Order-up-to level: S = Q* + s 47 48 12 Resolving the Crisis -Options Discussed in the Case Options Postponement: DC Localization Strategy Pros/Cons • Plant in Europe • Shorten lead time; May not have enough volume; Lose economy of scale in manufacturing • Air shipments • Shorten lead time; May be too expensive • Better forecasts • Would be great! But unclear how to do that … • Hold more inventory (higher z) • Better service; but already a problem • Factory-Localization Strategy: (customization performed at the factory) HP DC customers (manufacturer) • DC-Localization Strategy: (customization performed at the DCs) HP DC customers (manufacturer) Postponement: Delaying the point of differentiation Postponement? 49 50 Risk pooling strategies Existing strategies for coping with uncertainties Risk pooling strategies Location pooling Product pooling Lead Time Pooling Capacity pooling • Collecting data to ensure best demand forecast • Make-to-order production • Reactive capacity Risk pooling strategies : to redesign the SC, the production process or the product to either reduce the uncertainty the firm faces or to hedge uncertainty so that the firm is in a better position to mitigate the consequence. 51 52 13 Location pooling Risk pooling strategies Q: How many different locations should the firm store inventory? Q: To keep one stockpile inventory per sales representative or to serve demand from multiple territories from a single location? Q: 1 DC or N DC? Risk-pooling strategies to reduce and hedge uncertainty • Location pooling • Product pooling • Lead Time Pooling • Capacity pooling Concepts of location pooling : • pooled territory & pooled inventory • individual inventory & individual territory Important issues : Supply chain responsiveness: response time to orders 53 54 Product pooling Location pooling • Concepts: universal design to serve demand with fewer products CV: Coefficient of Variation expected inventory in days of demand CV • Pooling of 2 demands with mean and variance 2. E pooled demand 2 pool 2 1 coefficient of correlation average inventory CV pool 1 1 coefficient of correlation 2 Drawbacks of a universal design # of territories pooled • • • # of DC 55 May not provide the needed functionality to consumers with special needs May be more expensive or cheaper to produce than focused products May eliminate some brand / price segmentation opportunities product line rationalisation 56 14 Lead Time Pooling Lead Time Pooling Concept 2: Delayed differentiation / Postponement. • To remedy the drawbacks of previous strategies: – location pooling creates distance – product pooling degrades product functionality Good when: • customers demand many version and variety is important Concept 1: Consolidated distribution • Less uncertainty wrt total demand than wrt individual versions Keep inventory close to customer while avoiding location imbalance • variety is created late in the production process 8 week LT store 1 1 wk LT store 1 • variety can be added quickly and cheaply • components needed to create variety are inexpensive relative to the generic component. Without postponement Supplier Supplier store 100 8 wk LT With postponement retail DC store 100 Raw materials components Generic product Different products 57 Capacity pooling with flexible manufacturing Raw materials components Different products 58 Capacity pooling with flexible manufacturing 10 links : no flexibility plants vehicles 1 1 • Mainly used in automotive assembly lines for different car models • Traditionally, assembly lines are dedicated to one model • The future trend tends to flexible assembly lines capable of assemblying different models (modèles mixtes) plants 1 vehicles 1 2 2 2 2 3 3 3 3 10 10 10 10 20 links 59 11 links Total flexibility plants 1 vehicles 1 plants 1 vehicles 1 2 2 2 2 3 3 3 3 10 10 10 10 60 15 Capacity pooling with flexible manufacturing Capacity pooling with flexible manufacturing Impact of adding flexibility Chaining : groups of plants & groups of vehicles expected sale 20 plants total flexibility 1 11 12 links no flexibility 80 100% capacity utilization A configuration of 20 links has approximately equal capability to respond to demand uncertainty than totally flexible configuration 61 vehicles 1 2 2 3 3 4 4 9 9 10 10 62 16 01/10/2013 Base Stock Policy under Periodic Review Supply Chain Logistics and Operations Management • Assumptions Chapter 5 Value of Information qt Dt – Review period is 1 – Lead time is L – No need to consider fixed cost • Bullwhip Effect • Causes of Bullwhip • Means to Counter Bullwhip • Policy – At review point, if inventory position IP is lower than the base stock level, place an order to bring the inventory position back to the base stock level – Base stock level = Lμ + zβ σ L1/2 1 Moving Average Forecast Sequence of Events at the Beginning of a Period: • Demand forecast updated • Base stock level computed St = Lμt^ + zβ σt^ L1/2 • Order is placed if needed qt = St – St-1 + Dt-1 • Order, if any for the period, is delivered • Demand is realized • Inventory level is recorded • Inventory position is recorded • In period t, we update the demand estimate using the realized demands in the last p periods, Dt-1, …, Dt-p • The base stock level for each period is computed with the updated mean and variance • The updated estimates of mean and variance for period t ˆ t D t p Dt 1 p ( D t p ˆ t ) 2 ( Dt 1 ˆ t ) 2 , ˆ t p 1 2 1/ 2 3 4 1 01/10/2013 t 1 2 3 4 5 6 7 8 9 3-Period Moving Average Example (1) 3-Period Moving Average Example (1) L =3, β=0.90, zβ=1.285 L =3, β=0.90, zβ=1.285 D 120 101 115 102 113 112 120 92 110 μ^ 110 108 112 σ^ 10 11 11 112 106 110 109 115 108 10 8 7 6 4 14 IP Q μQ^ σQ^ t 1 2 3 4 5 6 7 8 9 D 120 101 115 102 113 112 120 92 110 μ^ 110 108 112 σ^ 10 11 11 112 106 110 109 115 108 10 8 7 6 4 14 IP Q 352 348 360 358 335 346 341 355 356 80 116 113 112 79 123 107 134 93 μQ^ σQ^ 5 t 1 2 3 4 5 6 7 8 9 6 3-Period Moving Average Example (1) 3-Period Moving Average Example (1) L =3, β=0.90, zβ=1.285 L =6, β=0.90, zβ=1.285 D 120 101 115 102 113 112 120 92 110 μ^ 110 108 112 σ^ 10 11 11 112 106 110 109 115 108 10 8 7 6 4 14 IP Q 352 348 360 358 335 346 341 355 356 80 116 113 112 79 123 107 134 93 μQ^ 103 114 102 105 103 121 σQ^ t 20 2 19 23 22 14 1 2 3 4 5 6 7 8 9 7 D 120 101 115 102 113 112 120 92 110 μ^ 110 108 112 σ^ 10 11 11 112 106 110 109 115 108 10 8 7 6 4 14 IP Q μQ^ σQ^ 691 683 707 703 661 682 673 704 693 80 111 125 111 60 134 103 151 82 105 116 99 102 99 129 23 8 35 38 38 24 8 2 01/10/2013 The Phenomenon Observed Bullwhip Effect if there is no collaboration at all Order Size The bullwhip effect is a phenomenon observed in supply chains wherein the demand variability increases as one moves upstream from retailers to distributors to manufacturers Customer Demand Distributor Orders Retailer Orders Production Plan Retailers Warehouses/ Distributors Time production distributor retailer customer Manufacturers 9 10 Bullwhip Effect Example 2: Campbell Soup Bullwhip Effect Example 1: P&G Diapers 11 12 3 01/10/2013 Conclusions Is the Bullwhip Effect Good or Bad? Why? • Good or bad? -- Bad. It distorts the order information & amplifies order variability. • Order variability is amplified up the supply chain: bullwhip effect • Upstream echelons face higher variability • Multiple causes and can be complex • Impact of Bullwhip Effect: -- Inventory: More safety stock needed -- Transportation: Lower utilization of transportation -- Warehousing: More warehouse capacity needed Higher costs -- Manufacturing: Lower capacity utilization -- Customer Service: Lower service level, more likely to cause stockouts and lost sales 13 14 Cause 1: Demand Forecast Updating (Demand Signal Processing) Causes of Bullwhip Effect • Root Causes: 1. Demand forecast updating Order Qt goes to upstream 2. Order batching upstream 3. Price fluctuation 4. Rationing and shortage gaming Orders from downstream in the past p time periods Dt-p, Dt-p+1, …, Dt-1 Lead time L downstream • Moving average to forecast demand at period t based on Dt-p, Dt-p+1, …, Dt-1 • Use base-stock policy to determine order Qt Var(Q) 1+ Var(D) 15 2L2 2L + 2 p p Demand variability gets amplified from downstream to upstream! 16 4 01/10/2013 Cause 1: Demand Forecast Updating (Demand Signal Processing) Q3 Q2 Stage 3 L3 Stage 2 L2 No collaboration stage i+1 at period t based on Qit-p, …, Qit-1 • Use base-stock policy for each stage to determine order Qit Var D Dt • Demand forecasting – No visibility of end demand – Forecast base on orders not the end demand – Long lead time increases forecast inaccuracy Stage 1 L1 Information sharing • Moving average to forecast demand of Var QI Q1 Cause 1: Demand Forecast Updating (Demand Signal Processing) I 2 L 2 L2 1 i 2i p p i 1 • Moving average to forecast demand of stage i+1 at period t based on Dt-p, …, Dt-1 • Use local or echelon base-stock policy for each stage to determine order Qit Var QI Var D 1 2 L1 ... LI p 2 L1 ... LI 2 p2 Information sharing reduces but cannot cancel Bullwhip effect 17 18 Cause 2: Order Batching Example of Order Batching Order Q* > Dt to upstream supplier Company waits several periods before placing an order for Q* units Order from downstream per time period Dt Demand/order Price incentives from upstream supplier (even if K = 0): •Quantity discounts •Promotions Orders to manufacturers (once every 4 weeks) Fixed ordering costs K Q* 2DK h Orders from retailers (once a week) 4 19 8 12 16 weeks 20 5 01/10/2013 Example of Order Batching Cause 3: Price Fluctuation • Reasons for batching – – – – • Estimates indicate that 80 percent of the transactions between distributors and manufacturers in the grocery industry are made in a forward buy arrangement. (Kurt Salmon Associates) High ordering cost Full truckload economies Quantity discounts Push ordering, salespersons need to fill sales quotas • A forward buy is one in which items are bought in advance of requirements, usually because of a manufacturer’s attractive price offer. • With price fluctuations, customers buy in quantities that do not reflect their immediate needs: -- They buy in larger quantities and stock up when price is low -- They postpone purchases when price is regular or high 21 22 Examples of Initiatives to Counteract the Bullwhip Cause 4: Rationing and Shortage Gaming • When product demand exceeds supply, a manufacturer often rations its product to customers. Example: Car Manufacturer Available = 200 Dealer 1 Order = 100 Received = 67 Dealer 2 Order = 200 Received = 133 Only 2/3 of the order can be fulfilled • Knowing the manufacturer policy, customers exaggerate their real needs when they order (game the system). Example: Car Manufacturer Available = 500 Dealer 1 Need = 120 Order = 180 Received = 180 Dealer 2 Need = 180 Order = 270 Received = 270 Order more than needed so that if only 2/3 of the order is filled you still get what you actually need • As a result, customers’ orders give the supplier little information on a product’s real demand, a particularly vexing problem for new products 23 Cause of Bullwhip Initiative(s) 1. Demand Signal Processing •Use of point-of-sale (POS) data •Electronic data interchange (EDI) •Vendor-managed inventory (Barilla Case) •Lead-time reduction 2. Order Batching •Use of EDI (to reduce ordering costs) •Logistics outsourcing 3. Price Fluctuations •Every day low price (EDLP) 4. Shortage Gaming •Sharing sales and inventory data •Allocation based on past sales 24 6 01/10/2013 Information for Effective Forecasting Information for Coordination • Questions • Pricing, promotion, new products – Who will optimize? – How will savings be split? – Different parties have this information – Retailers may set pricing or promotion without telling distributor – Distributor/Manufacturer might have new product or availability information • Information needed – – – – – • Collaborative Forecasting addresses these issues. Production status and costs Transportation availability and costs Inventory information Capacity information Demand information 25 Lead-Time Reduction Information to Address Conflicts • Lot Size – Inventory – Advanced manufacturing systems – POS data for advance warnings • Inventory-Transportation – Lead time reduction for batching – Information systems for combining shipments – Cross docking – Advanced DSS • Why? – – – – 26 Customer orders are filled quickly Bullwhip effect is reduced Forecasts are more accurate Inventory levels are reduced • How? – EDI – POS data leading to anticipating incoming orders. 27 • Lead Time-Transportation – Lower transportation costs – Improved forecasting – Lower order lead times • Product Variety-Inventory – Delayed differentiation 28 7 01/10/2013 Background Quick Response at Benetton Strategy Quick Response at Benetton • Benetton, the Italian sportswear manufacturer, was founded in 1964. In 1975 Benetton had 200 stores across Italy. • Ten years later, the company expanded to the U.S., Japan and Eastern Europe. Sales in 1991 reached 2 trillion (lira). • Many attribute Benetton’s success to the successful use of communication and information technologies. • Benetton uses an effective strategy, referred to as Quick Response, in which manufacturing, warehousing, sales and retailers are linked together. In this strategy a Benetton retailer reorders a product through a direct link with Benetton’s mainframe computer in Italy. • Using this strategy, Benetton is capable of shipping a new order in only four weeks, several week earlier than most of its competitors. 29 Coping with Bullwhip Effect at Benetton 30 Managerial Insights • Bullwhip effect exists, in part, due to the retailer’s need to estimate the mean and variance of demand. • The increase in variability is an increasing function of the lead time. • The more complicated the demand models and the forecasting techniques, the greater the increase. • Centralized demand information can significantly reduce the bullwhip effect, but will not eliminate it • Information has value and cannot be shared freely • Integrated Information Systems – Global EDI network that links agents with production and inventory information – EDI order transmission to HQ – EDI linkage with air carriers – Data linked to manufacturing • Coordinated Planning – Frequent review allows fast reaction – Integrated distribution strategy Benetton Benetton 31 32 8 Distribution Strategies Chapter 6. Distribution strategies & strategic alliance • • Direct shipping Distribution strategies - direct shipping - shipiping via warehouses - shipping via cross-docks – Examples: JCPenney • Shipping via warehouses – Examples: • Strategic alliance - Third Party Logistics (3PL) - Retailer-Supplier Partnerships (RSP) • Shipping via cross docks – Cross docks serve as _inventory coordination_ points – Products spend _very little time_ at cross docks – Examples: Wal-Mart • allowing transshipments (often at the retailer level) 1 Strategy 1: Direct Shipping manufacturer retailer manufacturer retailer manufacturer retailer manufacturer 2 Strategy 2: Shipping via Warehouse retailer manufacturers retailers warehouse Type 1. Single origin single destination Type 2. Single origin multiple destinations Type 3. Multiple origins single destination Type 4. Multiple origins multiple destinations Type 1. Without milk runs manufacturers retailers warehouse Type 2. With milk runs Direct shipping with milk runs Common when the retail store requires fully loaded trucks Mandated by powerful retailers or in situation when lead time is critical 3 4 1 Strategy 3: Shipping via Cross Docks Role of Warehouses • Warehouses: • Warehouses play important roles in the supply chain Receiving, Sorting, Storing, Order Picking, Shipping – Position _products_ __close_ to customer – _coordination_ function • Cross Docks = Warehouses without inventory • _inbound_ shipments from multiple suppliers • _outbound___ shipments to multiple customers Receiving, Sorting, Shipping (from incoming trucks to outgoing trucks in < 12h) Shipping Receiving – Even if firms sell products directly to customers (no retailers), they may still use warehouses Example: Sorting Inbound shipments Requires coordination & IT support & fast and responsive transport & Forecast critical, needs of info. Sharing & Good for large distribution systems 5 Comparison of the Three Strategies Facility Direct Shipping Shipping via warehouses Shipping via cross docks no facility needed wareho use cross docks Inventory Transportation costs increased transport cost reduced inbound costs reduced inbound costs high holding costs lower holding costs No holding costs Outbound shipments 6 Pro & cons of different transportation networks Lead Time from mfg to retailer reduced lead times long reduced SS proportional to what factors? 7 network structure Pros Cons Direct shipping No intermediate warehourse Simple to coordinate High Inv (large lot sizes) Significant receiving expense Direct shipping wt milk Lower transp. for small lots runs Lower inventories Increased coordination complexity Via central DC with inventory Lower inbound transp. by consolidation Increased inventory cost Increased handling at DC Via central DC with cross-dock Lower transp cost through consolidation Increased coordination complexity Shipping via DC using milk runs Lower outbound transp. for small costs Further increase in coordination complexity Tailored network Transp choice best matches needs of individual product and store Highest coordination complexity 8 2 Types of Strategic Alliances Why direct shipping has higher transportation or/and higher inventory cost? • Example: Retailer’s weekly demand = 1/2 truckload Shipping cost from manufacturer to retailer = $100 • Third Party Logistics (3PL) • Retailer-Supplier Partnerships (RSP) Direct Shipping – Quick response (QR) – Continuous replenishment (CR) – Vendor managed inventory (VMI) Shipping via Warehouse 1/4 truckload per 1/2 wk Full truckloads per 1/2 wk 1/2 truckload per wk per 1/2 wk Full truckloads 1/2 truckload per wk Full truckloads 1 truckload per two wks per wk 1 truckload per two wks 1/4 truckload per 1/2 wk per wk per 2 wks per 2 wks 9 3PL 10 Why 3PL can achieve economy of scale & provide better service? • 3PL = Some or all of a firm’s logistics functions is taken over by an independent logistics service provider (LSP) • Involve long term commitments and multiple function instead of traditional transaction-based, single function logistic supplier relationships. • Consolidation is the key! Example: 2 independent firms, 2 independent supply chains Consider 2 scenarios • Example: Ryder Integrated Logistics Scenario 1: The firms performs their own logistics functions – Annual revenues around US$ 1.5 billion – Offers everything from transportation to network design and consulting – A five year agreement to design, manage and operate all of Whirlpool’s inbound logistics. Firm 1 2 warehouses 2 separate distribution networks Firm 2 • Advantages of 3PL: – focus on core strengths, providing technological flexibility (IT & equipment), providing other flexibilities (warehousing, small retailers, ...) Scenario 2: A 3PL takes care of both firms’ logistics functions Firm 1 • Issues with 3PL: – know your own cost, customer orientation of the 3PL, specialization of the 3PL, asset-owning vs non asset-owning 3PL 11 Firm 2 3PL 1 warehouse 1 distribution network 12 3 Major 3PLs Company 3PL in Practice Revenues ($ million) Ryder Integrated Logistics $1,300 Penske Logistics 959 Schneider 875 Tibbet & Britten Group 659 Americold 650 North American Logistics 650 Fritz Companies 578 UPS Logistics 488 APL Logistics 420 Federal Express • “3PL Study: Results and Findings of 2001 Annual Study” by Cap Gemini Ernst & Young – 93 companies – Covering automotive, chemical, computer, consumer products, & electronics – Most prevalent among large companies, 52% with sales revenues over $1B, 10% between $500M to $1B 360 Source: Logistics Magazine (07/00) 13 How many companies use 3PL? From “3PL Study: Results and Findings of 2001 Annual Study” by Cap Gemini Ernst & Young 14 What 3PL functions do companies use? 15 From “3PL Study: Results and Findings of 2001 Annual Study” by Cap Gemini Ernst & Young 16 4 Retailer-Supplier Partnerships (RSP) Benefits of 3PL • Quick Response (QR) – Supplier receives POS data from retailers – Supplier use it to improve its own forecasting and production scheduling – but retailer still prepares its own orders • Continuous Replenishment (CR) – Supplier replenishes retailers – Supplier receives POS data and use it to prepare shipments at previously agreed upon intervals to maintain specific levels of inventory Increasing trust level • Vendor Managed Inventory (VMI) – Supplier replenishes retailers – Suppliers have the total control over replenishment decisions From “3PL Study: Results and Findings of 2001 Annual Study” by Cap Gemini Ernst & Young 17 Information sharing Continuous replenishment Joint forecasting & planning Supplier control of inventory decision Advantages • Improved _forecast_ • Decreased _cost_ VMI VMI with supplier inventory ownership – Lower _inventory__ – Lower _stockout_ increasing _trust_ level _local_ inventory control 18 RSP Issues and Advantages Continuum of RSP Relationships Quick Response Inventory decision-making increasingly global Issues _inventory_ ownership IT (heavy investment) Mutual _trust_ Suppliers have more responsibility • Sharing benefits • Confidentiality • • • • _global_ inventory control 19 20 5 Example of RSP Success Why does RSP have those advantages? • VF Corporation’s Market Response System: Without RSP: sequential, myopic optimization Supplier Retailer Information flow – The VF Corporation, which has many well known brand names (including Wrangler, Lee, Girbaud, and many others), began its VMI program in 1989. Retailer optimizes its operations first. Then supplier optimizes its operations subject to the constraints imposed by the retailer. – Currently, about 40 percent of its production is handled using some type of automatic replenishment scheme. – This is particularly notable because the program encompasses 350 different retailers, 40,000 store locations, and more than 15 million replenishment levels. With RSP (particularly, VMI): Joint optimization Supplier Customer Supplier optimizes its operations and the retailer’s. This is system-wide. – VF’s program is considered one of the most successful in the apparel industry. Information flow 21 22 Other Kinds of Partnerships: Third Party e-Fulfillment (3eF) Example of RSP Failure Spartan Stores (grocery chain) • 3eF = the outsourcing of the back-end logistics of e-business including: the integration with front-end Internet operations, order capture and processing, fulfillment of individual orders, and return logistics. – Shut down its VMI effort about one year after its inception – Buyers were not spending any less time on reorders than they did before • Differences between 3PL & 3eF? – Issue: buyers didn’t trust suppliers -- continued to carefully monitor inventories and deliveries and to intervene at hint of trouble. S M W R Logistics: physical flow from suppliers to manufacturers, or/and from manufacturers to retailers – Suppliers did little to allay these fears; suppliers did not deal well with promotions -- delivery levels were often unacceptably low during these periods of peak demand. M W R C E-Logistics: “last mile”, i.e. logistics of order fulfillment of e-businesses 23 24 6 Why is 3eF different from 3PL? 3eF Examples 1. Fingerhut Business Services Traditional Supply Chain – – e-Supply Chain Supply Chain Strategy Push Push-Pull Shipment Type Bulk Parcel Information Flow Unidirectional Bi-directional Reverse Logistics Simple Highly Complex 2. OrderTrust – Destination A major provider of e-fulfillment service Wal-Mart’s cyberstore is managed by Fingerhut Manages SkyMall.com’s order fulfillment Small Number of Stores Highly Dispersed Customers Lead Times Depends Short 25 26 3eF: Why Need for Reverse Logistics? • Return Percentage in the Offline World (Online world has much higher percentages) Industry Magazine Publishing (50%) Book Publishers (20-30%) Book Distributors (10-20%) Greeting Cards (20-30%) Catalog Retailers (18-35%) Computer Manufacturers (10-20%) CD-ROMs (18-25%) Consumer Electronics (4-5%) Source: Rogers and Tibben-Lembke 27 7 01/10/2013 1 2 Les fonctions de gestion à couvrir Long terme, gestion intégrée Gestion du réseau logistique mois Planif des Planif Planification distribution appros production jour Temps réel Gestion Ordo Ordo des appros Production des stocks Gestion transports suivi des conduite Conduite achats production entrepôts Conduite transports CLIENTS semestre Vendre Prévision des ventes Chapitre 7. Les outils informatiques du supply chain management Acheter Fabriquer Stocker Livrer Dimensionnement du réseau logistique Administration des ventes FOURNISSEURS ans Allocation stocks, Gestion commandes Court terme, gestion localisée Le domaine « classique » ERP et APS Acheter Fabriquer Stocker Livrer FOURNISSEURS trimestres Plan industriel et commercial Planif appros mois Ordo des appros Prévision des ventes Planif production Planification distribution Administration GPAO ERP des ventes Gestion Gestion Ordo production des stocks transports MES Temps réel APS conduite Conduite production entrepôts SCE Conduite transports Gestion commandes 4 Objectifs : CLIENTS Planification du réseau logistique L’approche ERP classique : objectifs Résulte du mélange de logiciels de GPAO, de comptabilité, de prévision, de finance, de distribution Vendre Simulation du réseau logistique ans jour 3 Transactionnel : réaliser l’échange d’information et gérer les informations => base de donnée unique Financier : permettre l’évaluation financière de toutes les activités : production, distribution, administratif, commercial Méthodes de calcul issues des logiciels de support PIC, MRP, DRP, gestion de stocks, ... 1 01/10/2013 Acheter Fabriquer Stocker Livrer 5 Vendre ans Plan Industriel et Commercial multi-site Prévision de marchés Planif appros MRP II DRP mois jour Ordo appros Temps réel Ordo Gestion production des stocks Suivi Suivi production entrepôts Prévision de vente Administration des ventes Gestion transports Conduite transports CLIENTS FOURNISSEURS trimestres DRP <=> Calcul des besoins de MRP La nomenclature DRP est le réseau de sites à traverser pour distribuer une demande Besoins Brut Besoins Brut Stock, Attendu, Besoin Net Stock, Attendu, Besoin Net Loi de gestion Loi de gestion taille de lot de production taille de lot d ’approvisionnement Délai d ’obtention Délai de livraison Besoin net Jalonné Besoin à lancer lancement en production lancement au transport Calcul de charge DRP : charge de réception de livraison charge de préparation de commandes charge de transport Suivi commandes L’approche APS : objectifs 6 Comparaison MRP/DRP 7 Résulte de la volonté de proposer des outils d’optimisation pour l’aide à la décision Aide à la décision : fournir des procédures évoluées pour aider à décider Aider au suivi : réalisation ≠ prévu ou planifié Supports Procédures d’optimisation ou heuristiques évoluées Récupère les données dans les bases de données de l’entreprise Récupère les informations de suivi d’exécution Acheter ans Fabriquer Stocker Livrer trimestres mois Vendre Dimensionnement du réseau logistique Planification stratégique du réseau logistique FOURNISSEURS Objectifs : 8 L’approche APS : démarche Planification tactique du réseau logistique Planif production + appros Prévision de marché et de vente Planif distribution jour Ordo production Gestion transports Temps réel Suivi Suivi production entrepôts suivi transports Promesse de vente CLIENTS L’approche ERP classique : démarche suivi commandes 2 01/10/2013 9 10 Module de prévision Rôle : consolider des prévisions de nature différente : Prévisions moyen ou long terme de marchés, de comportement, de réseaux, de familles de produits, de filiales. ≠ prévisions court terme de commandes de clients Le supply chain Management Analyse statistique pour repérer : tendance d’évolution de la demande effets saisonniers impact de promotions Les APS : …. des modules valeurs + analyse de corrélations Modèles de comportement (marchés, réseaux, produits, pays, familles produits) Demand Planning 11 Sur quel niveau réaliser les prévisions 12 méthodes de prévisions statistiques incorporation de facteurs externes support à la collaboration sur les prévisions Simulation : études What If Calcul de stocks de sécurité Dimension temps Prévision de vente Dimension géographique Définition de hiérarchie sur les 3 niveaux et aide à la consolidation 3 01/10/2013 Demand Planning 13 Les méthodes statistiques demande sporadique ventes perdues ou ventes retardées Système de sélection des paramètres des modèles prévision de produits à faible durée de vie Analyse de l ’erreur de prévision facteurs externes : intégration de l ’expert revised judgment : prevision d ’expert puis statistique puis expert met à jour sa prévision combined forecast : poids fixe à expert et statistique revised extrapolation forecast : événements pré-définis rule based forecast : rêgle expertes de combinaison de prévision statistiques econometric forecast : expert choisi et valide les séries explicatives Objectif : aider à dimensionner une chaîne logistique Faut-il un nouveau site de production : où ? Faut-il un nouvel entrepôt régional : où ? quels produits ? Quelle usine fournit quelle région ? Récupération de modèles de marché du module de prévision + simulation de planifications stratégiques APS Evaluation des coûts d’une chaîne logistique donnée C’est un module de simulation de chaînes logistiques. 14 Problématiques particulières Moyenne mobile lissage exponentiel triple : Holt Winters ARIMA / Box Jenkins regression linéaire de séries expicatives Module de dimensionnement de chaîne logistique Demand Planning Dimensionnement de stocks de sécurité • 1 ou plusieurs niveaux 15 Module de planification multi-site 16 Objectif : planification stratégique en 1 passe Résolution du modèle précédent • Horizon : 1 ou 2 ans • Période : 1 mois ou trimestre Permet d’obtenir un PIC validé à capacité finie • Taille de période grande => peu de problème de taille de lot Optimisation des coûts sur toute la chaîne 4 01/10/2013 Module de planification locale 17 Objectif : planification tactique sur des sites Souvent : DRP pour faire remonter les besoins puis planification locale multi-site de la production. • Problématique de taille de lot + problématique de lissage de charge Résolution du modèle APS par des heuristiques Master planning detecte l ’incapacité à faire la demande allocation de volume selon la hiérarchie géographique de la demande • Par rang de priorité à certaines régions • en proportion de la prévision faite • coefficient fixe 18 Objectif : jugés critiques ATP et pénurie de produits Demand Fulfilment et Available to promise(ATP) répartir les pénuries donner une date de disponibilité à une commande client Principe : Master planning donne un profil de stock dans le temps ATP réserve ces stocks pour des demandes Production sur stock => stock de produit fini Production à la commande => composants ou MP Configuration à la commande => Capable TP réservation de capacité d ’assemblage réservation de composants 19 ATP et promesse de date 20 Actions possible si commande > prévision réserver sur une autre zone géographique réserver plus tôt ou plus tard réserver un produit de rechange Modules ATP (I2 technologies) paramétrer les règles pour sélectionner la dimension sur laquelle choisir. Les forces de vente ont besoin de ce retour d ’information pour planifier leurs actions 5 01/10/2013 21 Autres modules Ordonnancement : aide interactive à la mise au point d’un diagramme de Gantt des ordres de fabrication Transport : aide à la définition de tournées en fonction d’un parc de camions Peu standardisés = dépendent des éditeurs Problèmes posés par le modèle APS 22 1) Taille des problèmes Pour une rapidité de résolution, limiter la planification globale 2) Taille des lots les quantités produits sont multiples de ces tailles de lots Introduction de variables entières (nbre de lots) 3) Compromis charge/capacité difficile à ramener à un coût Pour la planification tactique : intérêt à définir des stratégies heuristiques DRP + MRP est une stratégie. Identifier les éléments critiques Planification tactique APS sur ces éléments Planification classique DRP/MRP sur les autres 23 Comparaison modèle de planification / MRP + DRP 24 Démarche MRP + DRP : Calcul charge Calcul besoins Production PDP Calcul charge Calcul besoins livraison Prévision vente Cas APS : Consolidation des prévisions de vente des magasins Planification à capacité FINIE en 1 passe pour tous les sites de production et de distribution moins de « bidouilles » sur taille de lot, délai de production ou d’approvisionnement Moins de stocks + flux plus tendu Le supply chain management Le marché des logiciels 6 01/10/2013 Les éditeurs ERP : 20 % de croissance en 1998 25 Pourcentage de part de marché des éditeurs ERP en France en 1998 Les éditeurs APS : 45 % de croissance en 1998 2,4% 8% 5% 14,1% 2,9% 12% 5,1% 7,7% SAP Oracle Baan 7,8% 7% 38% 12% 45,5% 3,3% Parts du marché mondial 1998 en % Parts du marché français 1998 en % 2,7% 18% 3,6% 3,1% 3,0% 2,5% 4,5% 6,8% 44,8% 7,8% 23,9% 8,4% Intentia People Soft SSA JD Edwards Arès Autres QAD Source IDC France 1999 Les tendances du marché logiciel 1) Les ERP intègrent les modules APS SAP, BAAN, Oracle => leur propre APS JDE => rachat de Numetrix 2) Les APS cherchent à optimiser un maximum de fonctions =>descente dans le court terme 3) le mouvement vers Internet 2.1) Planification partagée dans l’entreprise et avec clients et fournisseurs :e-chain, Collaborative Planning Forecosting and Replenishment 2.2) Définition de sites www portail entre fabricants et distributeurs : Marchés d’échange de produits => cf. Trade Matrix.com 27 Manugistics I2 Technologies Synquest 26 Dynasys Logility BAAN Numetrix AspenTech Ilog SAP Autres (source IDC France 1999) (source Benchmarking Partners) Tendances du marché méthodologie 28 Le Customer Relation Management Différentes methodes pour répondre à un besoin client <=> Où se situe le stock de sécurité, Quel délai promettre • Pick to Order : stock = produit fini : Réservation de stock et livraison sur commande • Assemble to Order : Stock = produit intermédiaire : Configuration produit + Assemblage + livraison sur commande • Make to Order : Stock = matière première Configuration produit + charge + Assemblage + livraison sur commande • Engineer to Order : Stock = matière première Configuration produit + gamme + charge + assemblage + livraison sur commande. 7 01/10/2013 Efficient Consumer Response (ECR) 29 Démarche introduite aux USA en 1992 Objectif : comment améliorer l ’efficacité de la chaîne logistique dans le secteur de la grande distribution. Liste de bonnes pratiques généralisables à d ’autres secteurs : Par action sur l ’organisation des flux physiques Par action sur l ’organisation des flux d ’information 11 thèmes = 11 types de bonnes pratiques ECR : bonnes pratiques (2) 31 4/ Production synchroniséé : Utiliser l ’information des ventes pour réduire les tailles de lots et améliorer réactivité => baisse des stocks, meilleure tension du flux, réduction des délais 5/ Production flexible et fiable Réduction des stocks exige livraison fréquente et fiable + production adaptable => augmenter la flexibilité des sites de production pour s ’adapter aux fluctuations 6/ Intégrer les fournisseurs Partager les informations avec Ss-traitants et fournisseurs en conception et production ECR : bonnes pratiques 30 1/ optimiser les unités de conditionnement : unité de conditionnement fournisseur = unité des magasins de distribution => réduction des places en magasins + réorganisation des unités de transport 2/ Cross docking, GPA : limiter au maximum opérations logistiques Palettes livrées par fabricant limitent les opérations à réaliser chez distributeur => éclatement facile de palettes, rassemblement de commandes, codes barres, … 3/Mise à jour continue : Information Tps réel des ventes en magasins à tous les fournisseurs, fabricants. => fournisseur s ’engage sur un taux de service et Stock sécurité ECR : bonnes pratiques (3) 32 7/ Catégories de produits d ’après la vente finale Établir des familles de produits et objectifs par famille sur l ’ensemble de la chaîne d ’après les implications pour les clients finaux. 8/ Réduire les erreurs lors du lancement d ’un nouveau produit par une coordination sur l ’ensemble de la chaîne D ’autant plus important que rythme d ’innovation s ’accélère 9/ Réapprovisionnement de stock automatique Utiliser les ventes réelles pour définir des points de réapprovisionnement automatique => limité délai administratif de commande + facture par période. 8 01/10/2013 ECR : bonnes pratiques (4) 33 10/ gestion des coupons automatisée Informatiser complètement la gestion (émission, traitement, enregistrement) des bons de réduction 11/ échanges automatisés dans les contrats Simplifier la structure des contrats avec fournisseurs. Demander un engagement sur certains objectifs moyennant échange d ’information Les objectifs du CPFR Les intégrateurs : Collaborative Planning, Forecasting and Replenishement (CPFR) 34 Le CPFR est un processus global de collaboration industrie-commerce visant l’alignement de l’offre et de la demande dans le secteur des produits de grande consommation. 35 Processus du CPFR 36 Aligner les objectifs commerciaux du client et du fournisseur pour une (des) catégorie(s) de produits donnée(s) Intégrer les plans commerciaux dans les plannings opérationnels Fiabiliser les prévisions Augmenter l’efficacité des promotions Réduire les stocks dans la chaîne d’approvisionnement Dynamiser les ventes 9 01/10/2013 GPA / CPFR, deux processus complémentaires 37 En tant que système de calcul de commande efficace, la GPA peut se substituer en l’état aux étapes 6, 7, 8 et 9 du CPFR Maintenir les 2 systèmes à part : les prévisions concertées sont faites au plan national entre les fonctions commerciales, marketing, merchandising et opérationnelles la GPA fonctionne à l’échelle locale pour le calcul de commandes optimisées Donner de la visibilité à la GPA concernant les promotions concertées à 3 mois Les résultats sont substantiels à la fois pour les industriels et les distributeurs : Bénéfices : Amélioration des prévisions de vente 10 – 40% Réduction des stocks 10 – 15% Amélioration du taux de service 0.5 – 4.0% Augmentation des ventes 2 – 25% Source: Derived from VICS published CPFR pilots and Transora participant company results 38 Johnson & Johnson UK Collaboration aval avec Superdrug Par article / semaine / entrepôt distributeur Extension à d’autres produits et distributeurs Marks & Spencer Collaboration quotidienne avec les fournisseurs de sandwiches Fenêtre de visibilité : 14 jours + 10 semaines Extension à l’ensemble des produits et fournisseurs de la catégorie Kimberly Clark France Collaboration sur les promotions avec un distributeur (75 produits ; 3 entrepôts) Extension progressive du périmètre Procter & Gamble CPFR et VMI (Vendor-Managed Inventory) avec différents distributeurs et fournisseurs de matières premières en Europe Pfizer and Unichem Collaboration sur les prévisions de vente et de livraison (50 produits, 11 entrepôts) Donner de la réactivité aux prévisions grâce à l’analyse des premiers jours de vente afin de corriger le tir, si besoin est. Exemples de résultats obtenus avec une démarche CPFR Quelques projets CPFR en Europe Par article / semaine / entrepôt distributeur 39 Synthèse sur le CPFR 40 Le CPFR n’est pas un mythe, mais une réalité On peut l’assimiler à un puissant « système de contrôle » de la bonne marche des affaires internes et externes d’une entreprise Par essence, il stimule la croissance et le reengineering orienté marché Il n ’y a pas un mais plusieurs CPFR selon les scénarios Il s’applique particulièrement aux produits dont la demande est irrégulière (nouveaux produits, promotions, produits saisonniers...) Ses performances sont telles que les entreprises qui réussissent le mieux sont celles qui en parlent le moins ! Mais sa mise en œuvre requiert une forte motivation de la direction générale et la dynamique d’une équipe de projet pluridisciplinaire 10 Strategic fit 1. How would you characterize the competitive strategy of a high-end department store chain such as Nordstrom ? What are the key customer needs that Nordstrom aims to fill? 2. What level you place the demand faced by Nordstrom on the implied demand uncertainty spectrum? Why? 3. What level of responsiveness would be most appropriate for Nordstrom’s supply chain ? What should the supply chain be able to do particularly well? 4. How Nordstrom expand the scope of strategic fit across its supply chain? 5. Reconsider the previous four questions for other companies such as Amazon, a supermarket chain, an auto manufacturer, and a discount retailer such as Wal-Mart. 2. Consider the purchase of a can of soda at a convenience store. Describe the various stages in the supply chain and the different flows involved. 3. Why should a firm like Dell take into account total supply chain profitability when making decisions? 4. What are some strategic, planning, and operational decisions that must be made by an apparel retailer like The Gap? 5. Consider the supply chain involved when a customer purchases a book at a bookstore. Identify the cycles in this supply chain and the location of the push-pull boundary. 6. Consider the supply chain involved when a customer orders a book from Amazon. Identify the push/pull boundary and two processes each in the push and pull phases. 7. In what way do supply chain flows affect the success or failure of a firm like Amazon? List two supply chain decisions that have a significant impact on the supply chain profitability. See www.emse.fr/~xie/MasterGI/Ch1_supplement.pdf for questions below. A. Quelles sont les principales caractéristiques de Supply Chain Management ? B. Décrire les principaux flux d’une supply chain et leurs caractéristiques. C. Décrire le rôle de la frontière Push/Pull dans une chaîne logistique. Expliquer son importance dans une supply chain. D. Quels sont les leviers de performances des supply chains. E. Quels sont les principaux processus identifiés par le modèle SCOR ? Qu’apporte une analyse de son supply chain par la méthode SCOR ? F. Quels sont les avantages et les désavantages d'un réseau de distribution centralisée par rapport à un réseau de distribution décentralisé. G. Pour faire face à la diversité des produits en automobile, les constructeurs développent un nombre limité des modules génériques pour couvrir la plupart des demandes spécifiques. Quels sont les avantages et les inconvénients de cette méthode? H. Dans d'une chaîne logistique composée d'un producteur et des distributeurs appartenant à des entreprises différentes, traditionnellement, chaque distributeur détermine ses approvisionnements en fonction des demandes des clients finaux, le producteur considère les distributeurs comme des clients finaux et établit son plan de production pour répondre aux ordres d'approvisionnement de ces derniers. Quels sont les problèmes posés par cette pratique? Que peut apporter le partage des informations sur la demande client? Quel est l'avantage de laisser producteur gérer les stocks de distributeur? Ces stratégies sont-elles réalisables? 1. (maîtrise de concepts, soyez bref et clair) Chapitre 1. Introduction Exercices Usine 4 wk 4 wk DC2 DC1 Marché 2 Marché 1 Usine 4 wk DC Marché 2 Marché 1 On vous demande de : 1) Calculer la quantité économique de chaque DC 2) Déterminer le stock de sécurité de chaque DC 3) Déterminer le niveau de stock moyen ainsi que le coût de stockage annuel de chaque DC (suggestion : dessinez la courbe de l’évolution du niveau de stock) 4) Déterminer le niveau de stock maximal, la capacité de stockage de chaque DC. On suppose que chaque produit a un volume de 0,1 m3 et la capacité d’entreposage est de 3 fois le volume des produits (pourquoi ?). 5) Déterminer le coût total de stockage et de construction des DC sur un horizon de 5 ans. On suppose que le coût de construction est de 1000€ par m3 pour un DC de moins de 2000 m3 et de 900 € par m3 pour un DC de plus de 2000 m3. 6) Quelles conclusions en tirez vous ? Supply chain design Une entreprise souhaite savoir si il est possible d’optimiser son réseau de distribution. Actuellement, elle distribue ses produits via deux centres de distribution (DC1 et DC2). Elle cherche à savoir le gain en regroupant DC1 et DC2 en un centre de distribution central (DC). La demande hebdomadaire du marché 1 est de moyenne (1000) et d’écart-type (500). La hebdomadaire demande du marché 2 est de moyenne (1600) et d’écart-type (700). Les deux demandes sont indépendantes. Le délai d’approvisionnement est de 4 semaines dans tous les cas. Chaque produit a une valeur de 1000€, le coût de stockage annuel est de 13% de la valeur du produit, le coût de commande est de 10000 € par commande. Les centres de distribution sont gérés pour niveau de service de 95%. SC drivers and metrics 1. How could a grocery retailer use inventory to increase the responsiveness of the company’s supply chain ? 2. How could an auto manufacturer use transportation to increase the efficiency of its supply chain? 3. How could a bicycle manufacturer increase responsiveness through its facilities? 4. How could an industrial supplies distributor use information to increase its responsiveness? 5. Motorola has gone from manufacturing all its cell phone in-house to almost completely outsourcing the manufacturing. What are the pros and cons of the two approaches? 6. How can a home-delivery company like Peapod use pricing of its delivery services to improve its profitability? 7. How has globalization made strategic fit even more important to a company’s success? 8. What are some industries in which products have proliferated and life cycles have shortened? How have the supply chains in these industries adapted? 9. How can the full set of logistical and cross-functional drivers be used to create strategic fit for a PC manufacturer targeting both time-sensitive and price-conscious customers ? 6. Give arguments to support the statement that Wal-Mart has achieved very good strategic fit between its competitive and supply chain strategies. a) Considérons le réseau actuel. Déterminer le stock de sécurité à constituer dans chaque magasin. Déterminer le coût de stock de sécurité de l'ensemble des magasins. b) Déterminer le niveau de stock de sécurité de chaque magasin dans le nouveau réseau. c) Déterminer la variance de demande du DC du nouveau réseau et son stock de sécurité. Dans le deux cas, le niveau de service exigé est de 95%. Afin d'améliorer la chaîne logistique, l'entreprise procède à une refonte de son réseau de distribution. La nouvelle structure comporte un Centre de Distribution DC permettant d'alimenter les magasins en 1 semaine. Le DC s'approvisionne seul auprès du fournisseur avec un délai d'approvisionnement de 8 semaines (voir la figure de la page suivante). 1. (Lead time pooling) Une entreprise possède 300 magasins en France. Les magasins sont localisés de telle sorte que leur demande est similaire. La demande hebdomadaire de chaque magasin est une variable aléatoire de distribution normale N(1000, 5002) de moyenne 1000 et de l'écart-type 500. Le coût de chaque produit est de 100 euros par unité. Le coût de stockage annuel d'un produit est 13% de son coût. Chaque magasin s'approvisionne auprès d'un même fournisseur et le délai d'approvisionnement de 8 semaines. Risk pooling 3. (Aggregation with capacity constraint) WW Grainger sources from hundreds of suppliers and is considering the aggregation of inbound shipments to lower costs. Truckload shipping costs 500$ per truck along with 100$ per pickup. Average annual demand from each supplier is 10000 units. Each unit costs 50$ and Grainger has an annual holding cost of 20%. What is the optimal order frequency and order size if Grainger decides to aggregate 4 suppliers per truck? What is the optimal order size and order frequency if each truck has a capacity of 2500 units? 2. (aggregating multiple products in a single order) Consider 4 different product modes of Best Buy store each with data of problem 1 and all four products are sourced from the same source. a) what is the total cost and order quantity if each product model is managed independently. b) What if the four product managers coordinate their purchasing to ensure that all four products arrive on the same truck, i.e. shared the same fixed order cost? c) What do you think about the fixed cost structure? Is it reasonable? 1. (EOQ) Demand for Despro Computer at Best Buy is 1000 units per month. Best Buy incurs a fixed order placement, transportation, and receiving cost of 4000$ for each order. Each computer costs Best Buy 500$ and the retailer has an annual holding cost of 20%. a) Evaluate the order size, order frequency, cycle inventory, annual ordering cost, annual holding cost, average flow time. b) Evaluate the total inventory cost of a lot size of 1100 units. What observation can you make? (Robustness) c) Determine the order quantity, cycle inventory, flow time if the demand of Best Buy increases to 4000 computers per month (demand increased by a factor of 4). What observations can you make? d) For the situation D = 1000, what if the manager would like to reduce to lot size to Q = 200 units? What if the fixed cost is reduced to 1000$ per order? e) How much should the fixed order cost be reduced to in order to reduce the optimal order size to 200? Ch5 : Cycle inventory store 300 store 1 Supplier 8 wk LT retail DC 1 wk LT store 300 store 1 1 16 40 1 43 67 1 64 99 t demande 1 demande 2 t demande 1 demande 2 t demande 1 demande 2 2 94 72 2 84 8 2 92 59 3 70 20 3 54 44 3 11 18 4 52 67 4 77 30 4 3 12 5 8 42 5 42 55 5 7 14 6 56 68 6 74 6 6 42 5 7 19 62 7 56 26 7 34 44 8 7 41 8 6 66 8 30 22 9 50 93 9 50 88 9 33 37 10 39 17 10 10 93 10 78 73 2. Risk pooling Pour chacun des trois historiques de demande de deux produits similaires, (i) vérifions pour chacun si il y a une corrélation entre les deux produits, (ii) déterminer la moyenne et l'écart type de chaque demande i et de l'ensemble de deux demandes 1+2; (iii) déterminer pour les paramètres des politiques de gestion de stock (s, S) pour deux scénarios (a) les demandes sont satisfaites à partir de deux stocks différents et (b) les demandes sont satisfaites à partir d'un seul stock en utilisant les paramètres suivants: coût de commande K = 100 euros/ordre, coût de stockage h = 1 euro /produit/période, niveau de service = 95% et le délai d'approvisionnement L = 2, (iv) quelles conclusions en tirez vous? Supplier 8 week LT Déterminer le coût de stocks de sécurité du nouveau réseau. Quelle conclusion en tirer? Que se passe-il si il y ait que 2 magasins? Que se passe-il avec toujours 300 magasins mais avec 7 semaines de délai pour l'approvisionnement des magasins auprès du DC dans le nouveau réseau? h) Que pensez vous des coûts de transport? d) e) f) g)