Tema 4: Supply Chain Management Logistics 1 Supplier Manufacturer Warehouse Retailer Customer Tema basado en: Supply Chain Management: Strategy, Planning, and Operations, Sunil Copra and Peter Meindl (Editors). Notas personales
Bullwhip Effect 2 In many industries, it has been observed that the variability in orders is higher in upper echelons of the supply chain than it is in the lower echelons. Manufacturer orders to supplier Wholesaler orders to manufacturer Retailer orders to wholesaler Customer demand at retailer Orders Time
Bullwhip Effect 3 Lee et. Al. (1997) examine four causes of the bullwhip effect: 1. Reactions to demand signals 2. Order batching 3. Price fluctuations 4. Shortage gaming
Bullwhip Effect 4 1. Reactions to demand signals Consider the infinite-horizon, periodic review inventory model we discuss earlier. Suppose a retailer uses the order-up-to S policy as discussed in that model. Furthermore, as a change to the earlier model, suppose the demand across periods is known to be positively correlated (as opposed to being independent across periods). Question: The retailer observes a slight increase in consumer demand this period. How should the retailer adjust S?
Bullwhip Effect 5 1. Reactions to demand signals Discussion: How can we mitigate the bullwhip effect due to demand signal processing? Demand information sharing VMI: vendor managed inventory Both of these measures will help the wholesaler observe the changes in real customer demand as opposed to the exaggerated responses of the retailer, which helps the wholesaler make better inventory decisions.
Bullwhip Effect 6 2. Order Batching Clarification regarding terminology: Order batching refers to the number of periods of demand batched in an order. For example, if a retailer is ordering once every R periods, then the retailer is batching R periods of demand in a single order. The more batched the orders are, the higher the variability in orders observed by the upstream party. Companies may like to increase R because: of: Transportation costs Volume discounts Fixed cost of ordering
Bullwhip Effect 7 2. Order Batching Companies may order at the same time as others because of: Institutional practices Use of MRP systems Discussion: How can we mitigate the bullwhip effect due to order batching? Reduce volume discounts Reduce fixed cost of ordering Use 3PL
Bullwhip Effect 8 3. Price Fluctuations Suppose the random demand for a product has the same distribution throughout time, but a retailer faces two different prices when buying the product from the wholesaler. The retailer orders will be large whenever the price decreases, will go down after a price increase, go back up when price decreases again, and so on One important reason for wholesale discount price fluctuations is the use of discounts as promotional tools. When a manufacturer gives a discount to the retailer, it is usually on the condition that the retailer will also give a discount on the product to the end customer (or, promote the product in some other way, through displays, flyers, etc.)
Bullwhip Effect 9 3. Price Fluctuations Discussion: How do we mitigate the bullwhip effect caused by the price fluctuations? EDLP Give discounts but face out the delivery of orders
Bullwhip Effect 10 4. Rationing Game Suppose a manufacturer sells a product to N retailers. The manufacturer has limited capacity, i.e., the total amount ordered by the retailers is likely to exceed the manufacturer s supply of the product. If the manufacturer has the policy of allocating supply in proportion to orders, what would the retailers do? In such case, the orders will show more variability than demand, leading to the bullwhip effect.
Bullwhip Effect 11 4. Rationing Game Discussion: How do we mitigate the bullwhip effect due to rationing game? Flexible contracts (faced in time) Ration capacity in proportion to earlier sales (or market share)
Postponement 12 Postponement is delaying the differentiation of the products in a production/distribution process
Postponement 13 Example: Suppose a bike manufacturer in Asia is producing a given bike in only two varieties: with disc brakes or V-brakes. In status quo, the distribution of bikes in the US works as follows: Manufacturer Bike w/ DB Bike w/ VB 6 weeks US Distribution Center Bike w/ DB Bike w/ VB Demand for Bikes w/ DB Demand for Bikes w/ VB In status quo, the Us DC has to keep inventories of each bike to cover demand for 6 weeks.
Postponement 14 Example: Suppose now the bike is redesign so that the US DC can easily mount the brakes on the bike. How can we take advantage of this redesign? Manufacturer US Distribution Center Bike 6 weeks Bike Demand for Bikes w/ DB DB VB DB VB Demand for Bikes w/ VB
Postponement 15 Example: Now, the US DC has to keep inventories of the generic bike and each kind of brake to cover demand for 6 weeks. Allocation of total production to V-brakes and disc brakes can be done by the US DC, when there is more information on the customer demand. Question: Change of safety stock levels at US DC Before Redesign After Design Bikes (both x < x types) V-brakes 0 + Disc Brakes 0 + Question: why will postponement be desirable in this case?
Postponement 16 Example: (cont) Numerical example: Suppose, over a 6-week period, the demand in US for bikes with V-brakes is N(μ=10,σ=3) and the demand for bikes with disc brakes is N(5,2). The US DC has a policy of setting the safety stock so that stock-outs occur in only 5% of the ordering cycles. Compare the safety stock levels for bikes at the US DC before and after redesign. What are the safety stock levels for each brake type after redesign? Note: Assume Dc uses a Q-R model. Also z 0.95 =1.645 Calculations: Safety Stocks Before redesign Bikes w/ V-brakes = σ(z)= 3 ( 1.645) = 4.935 Bikes w/ D-brakes = σ(z)= 32 ( 1.645) = 3.290 Total safety stock of bikes= 8.225
17 Postponement Example: (cont) Safety Stocks After redesign First find the distribution of total demand for bikes: N(10+5, [3 2 +2 2 ] 1/2 ) Note: this reduction in total standard deviation is usually Bikes = σ(z)= 3.61( 1.645) = 5.938 V-brakes = σ(z)= 3 ( 1.645) = 4.935 D-brakes = σ(z)= 32 ( 1.645) = 3.290 Total safety stock of bikes = 5.938 called risk pooling
Postponement 18 Discussion : What will be the additional benefits of postponement? Product variety more attractive Discussion : What are some reasons that may make postponement hard to implement? Additional responsibility of DCs Product or production process needs to be redesigned
Component Commonality 19 Component commonality refers to the use of a single component in different varieties of the product. Example: A manufacturer of audio/video equipment has a warehouse in Japan where they keep inventories of TVs to meet demand from regional warehouses all over the world. The manufacturer produces a popular high-definition TV in two different varieties: One with 110V power supply (for US market) and the other with 220V power supply (for EU market).
Component Commonality 20 Example: Factory Japan WH Demand for TV w/110v TV w/ 110V 5 weeks TV w/ 110V TV w/ 220V TV w/ 220V Demand for TV w/220v Suppose now the HDTV is redesigned so that a universal power supply is used. Now, the distribution is simpler.
Component Commonality 21 Example: Factory Japan WH Demand for TV w/110v TV w/ Power supply 5 weeks TV w/ Power supply Demand for TV w/220v Question: How SS changes at warehouse Before Redesign level? After Redesign TVs (both types) x < x
Component Commonality 22 Example: Numerical example: Suppose, over a 5-week period, the demand at the warehouse for TVs with 110V power supplier is N(25,3) and the demand for TVs with 220V power supplies is N(30,4). The warehouse has a policy of setting safety stock so that stockouts occur only 10% of the time. Compare the safety stock levels for TVs before and after redesign. Note: z 0.90 =1.28. Safety Stock: Before redesign TVs with 110V PS = 3 (1.28) = 3.84 TVs with 220V PS = 4 (1.28) = 5.12 Total SS for TVs = 8.96
Component Commonality vs. Postponement 23 Discussion: Essentially, both are the means to the same end; take advantage of the risk pooling. However, there are some (potentially) important differences in implementation. Postponement Component Commonality Redesign cost Increases Increases Unit cost Same Present Component Same Decreases inventory Warehouse responsibility Increases Same
24 SC Coordination Consider the following model of a SC: A retailer faces a single period, stochastic demand for a product. Let F be the cdf of the demand for the product. In the beginning of the period, the retailer has to decide how many units of the product to stock. The retail price of the product is p per unit, and the wholesale price paid to the manufacturer is w per unit. The manufacturer s unit production cost is c. Once the retailer determines the stock level, she places an order with the manufacturer. The manufacturer builds to order, and delivers to the retailer before the selling season starts. c Manufacturer w Retailer p
25 SC Coordination The retailer faces the newsvendor problem The underage cost of the retailer is: p-w The overage cost of the retailer is: w Then, the retailer will choose stock level y R that satisfies the critical fractile: p w F( y R ) p c Manufacturer w Retailer p
26 SC Coordination Let S(y) be the expected sales to the end customer, i.e., S(y) = E[min(D,y)] Then, in the decentralized supply chain, the retailer s profit is: p S( y ) w R y R The manufacturer s profit is: c Manufacturer w ( w c) y R Retailer The total SC profit is: p S( y ) c R y R p
27 SC Coordination Now consider the same supply chain, but assume now that the manufacturer and the retailer are owned by a single firm. In other words, the supply chain is now vertically integrated. The underage cost for the firm is: p-c The overage cost of the retailer is: c Then, the firm will choose stock level y I that satisfies the critical fractile solution, i.e., p c F( y I ) p c Manufacturer w Retailer p The firm s profit is: p S( y ) c I y I
28 SC Coordination Observation: y I minimizes the integrated SC s profit. This is the maximum value total SC can take. We will denote it by π I. When y R units are stocked, the total supply chain profit will be less than π I. The inferior performance of the decentralized SC is due to double marginalization: In the decentralized (non-integrated) SC, the profit margin of a unit sale is divided between the manufacturer and the retailer, whereas in the integrated SC a single owner enjoys all the profit margin from a unit sale. c Manufacturer w Retailer p
SC Coordination 29 Question, can we induce the retailer to choose y I instead of y R? c Manufacturer w Retailer p
30 SC Coordination Buyback Now, after the end of the selling season, the manufacturer buys all unsold inventory at a price b per unit. The underage cost of the retailer is: p-w The overage cost of the retailer is: w-b Then, the retailer will choose stock level y B that satisfies the critical fractile: p w F( y B ) p b Question: What should b be so that y B =y I? c Manufacturer w Retailer p Solution: set b so that: p p w b p c p