Supply Chain Management Prashant Yadav, Professor of SCM, MIT-Zaragoza International Logistics Program
Tsukuba University, Jan 8-11, 2008
Why does inventory management get complicated in a disintegrated multi-agent supply chain ?
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 2
Number Numberof ofbabies babiesare areconstant! constant! How Howcome comethere thereis issuch suchaavariability variabilityin inthe thedemand demandfor forPampers? Pampers?
Order Size
Distributor Orders Retailer Orders
Customer Customer Demand Demand
Manufacturer’s Manufacturer’sOrders Orders
Time Source: Tom McGuffry, Electronic Commerce and Value Chain Management Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 3
This phenomenon of order variability amplification as you go upstream in the supply chain is commonly termed “bullwhip effect”
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 4
Examples: Procter & Gamble (P&G) and Diapers –
Reasonably stable market at consumer level
–
Substantial variability in orders that P&G placed on suppliers (such as 3M)
–
P&G coined the term “bullwhip effect”
Hewlett Packard (HP) and Printers
–
Some variability in sales
–
Bigger swings at the orders of resellers
–
Even bigger swings for orders from printer division to the IC division
Barilla and its distributors
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav
Order Variation
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 6
Distributor Performance
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 7
What causes the bullwhip effect ?
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 8
Demand Forecast Updating Orders as the only signals of product demand to upstream members Forecasting techniques rely heavily on recent demand observations
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav
Order Batching Economies of scale in ordering costs and manufacturing setups lead to
batching Periodic planning and ordering as part of MRP and periodic review
systems Economies of scale in transportation (full truckload rates) “Push” ordering motivated by short term financial performance measures
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav
Price Fluctuation and Trade Promotion Quantity discounts and advance purchases Variance of buying quantities bigger than variances in consumption
rate Hysteresis vs. Pantry Loading Marketing vs Supply Chain Management
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav
“Hysteresis” If you push on something, it will yield: when you release, does it spring back completely ? If not, it is exhibiting hysteresis
sales increase quickly as a result of promotion, then remain the same or decline slowly
Time Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav
Pantry Loading consumers buy in advance in response to a promotion, leading to post-decline in sales
Time Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav
Rationing and Shortage Gaming When demand exceeds manufacturing capacity, rationing of orders Customers exaggerate their real needs (“gaming”) Example, 1980’s DRAM chips When capacity constraints are removed, orders suddenly drop Example, HP and Laserjet III Printers
Motorola 1992, 1993 cellular phones
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav
Countering the bullwhip effect ?
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 15
Countering the Bullwhip Effect
Vendor Managed Inventory (VMI) Barilla’s JITD Continuous Replenishment Program (CRP) at Campbell Soup, Nestle, M&M, P&G, Scott Paper
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav
Countering the bullwhip effect with information sharing
Sell-through data Point-of-Sale (POS) data
Discount for Information Sharing ?
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav
POS ? Kmart Workbench, Wal-Mart RetailLink, Target Partners-On-Line In 1985, Walton and his chief lieutenant, David Glass, began developing a
program called Retail Link. The software, and the hardware that went along with it, took years to perfect, eventually costing $4 billion. Wal-Mart shared this revolutionary software with suppliers at no cost, in order to help them meet the retailer's needs more efficiently Through Retail Link, suppliers can:
Determine how many of their products were sold at Wal-Mart stores the previous day.
Download purchase orders from Wal-Mart.
Check the status of their invoices to Wal-Mart.
Examine the effects of markdowns or returns on their inventory.
Access reports on sales over a period of up to two years, as well as sales forecasts for their products for up to one year.
Upload reports and updates for Wal-Mart.
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav
Dirty POS ? Cleaning POS Stock-outs and Unconstraining IRI, AC Nielsen survey data for weekly POS Other means of obtaining customer demand data directly
Jewellery retail example
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav
Countering the bullwhip effect by reducing order batching
Internet buying exchanges and Computerized Ordering
Discount for Truck Load Assortment- Mixed SKUs
Shipment Consolidation
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav
Countering the bullwhip effect by eliminating price fluctuation
Everyday Low Price (EDLP)
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav
Thought Exercise Twolever is a large European food giant with strong brands in almost all food
product categories. Its products are distributed through major grocery chains in the UK and
Europe The global supply chain manager at Twolever, Jeff (unfortunately) studied at
a business school in UK and took a course in supply chain management. He has mandated a company wide program to reduce the bullwhip effect. He
believes order batching and forward buying at the retailers is causing the demand fluctuations to get amplified before they reach the factories. He has also compiled some aggregate information on how much does additional variation in demand cost Twolever’s supply chain. He initiates a special bonus and career rewards program for the category
supply chain manager who achieves maximum reduction in bullwhip effect in his/her product category. Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 22
Thought Exercise They have also started obtaining Point of Sales data from a retail chain
which controls 30% of the market and very soon start obtaining it from the other big ones. He is confused on the right measure to base his bonus/rewards program
around. Jeff is amazed by your superior and more analytical knowledge of supply
chains and has engaged you as a consultant to help him.
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 23
Thought Exercise Twolever’s ice-cream division produces many different variants of a special
frozen treat with a shelf life of 7 days Twolever uses its own cold chain distribution network for distributing these
frozen treats. They are sold on a delivered cost basis to the retail chains’ distribution centers by Twolever's trucks. For this product the lead time from the factory to the retail distribution
centers is less than a day and large chains typically receive two deliveries per week There is a high storage cost for frozen treats ($2 per box of 1000 per day) for
Twolever. This cost is likely to be even higher for the retailers. The ice-cream SC manager argues that Jeff’s scheme does not make sense.
At least not for his category. As the unbiased external consultant what would be your recommendations Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 24
Traditionally: You call, we haul
Retailer monitors places
inventory levels
orders
Manufacturer manufactures
product
routes
vehicles
makes
deliveries
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 25
Vendor Managed Inventory : You rely, we supply Retailer
lets the vendor to manage the inventory
Manufacturer
monitors customers’ inventory
controls inventory replenishment & decides – when to deliver – how much to deliver – how to deliver
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 26
VMI: more formally stated Supplier takes on the responsibility of managing the inventory at the
buyer’s site for the products it supplies, i.e., monitoring, planning, and directly replenishing the inventory at the buyer’s location(s) Supplier determines when stocks are to be replenished and in what
quantities, rather than it responding passively to orders placed by its buyers. Guided by a contract which specifies the financial terms, inventory
constraints, and/or performance targets such as service measures. VMI transfers inventory management (and in some cases ownership)
from the customer to the supplier Consignment is when supplier owns title of the goods
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 27
VMI in philosophy To produce things and to rear them,
but not to take possession of them, To act, but not to rely on one's own ability, This is called profound and secret virtue.
Lao Tzu, The Way of Lao-Tzu Chinese philosopher (604 BC - 531 BC)
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 28
VMI Benefits Buyer Relieved of the burden to specify, to place and to monitor purchase
orders, while enjoying guaranteed service levels Less resources for inventory management
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 29
VMI Benefits Supplier Significantly reduced forecast uncertainties, and hence safety
stocks Reduced logistics costs; better coordination of deliveries to
decrease transportation cost Freedom in when & how to manufacture product and make
deliveries - level production Better coordination of inventory levels at different customers
Both A renewed and reinforced link rises between the buyer and supplier
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 30
VMI Operational Implementation The supplier guarantees the continuity of a service level ,
usually a contracted service level agreement (SLA) The stock is stored in the buyer’s raw material inventory Raw materials are picked up daily, according to production
requirements or market demand, and paid to the supplier on the consumption, according to the agreement arranged The consumption trend information is constantly refreshed and
immediately transferred to the supplier
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 31
VMI Service Level Agreement ? Service level SL Fill rate FL stock level between a minimum level z and a maximum level Z Balanced scorecard attribute Time horizon to measure these and penalty clauses
Manufacturer-Retailer VMI vs. Supplier-Manufacturer VMI
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 32
Issues in VMI coupled with Consignment One time inventory buy back by suppliers to start the Consignment
inventory program Handling Shrink Pharmaceutical supply chain example
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 33
VMI Essentials TRUST
Accurate information provided on a timely basis
Confidential information kept confidential
TECHNOLOGY
Automated electronic messaging systems to exchange sales and demand data, shipping schedules, and invoicing
Source: Martin Savelsbergh, GA Tech
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 34
Power equation in the supply chain decides how easy or difficult it will be to implement vendor managed inventory or continuous replenishment programs So which direction is the power equation tilting ?
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 35
Retail power
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 36
Category Captainship The big dog not only eats first but also decides who will get the
plate scraps The limiting case of VMI is Retailers only sell shelf space to
manufacturers Slotting allowances
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 37
Multi-level inventory management and stock positioning
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 38
Typical Structure Central DC
Regional DC
Regional DC
Retailer
Retailer
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 39
Complexity Demand at central DC is dependent upon the demand and the stocking
decisions at the branch DCs Demand at branch DC is dependent upon the demand and the stocking
decisions at the retail locations
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 40
Pertinent questions ? Should inventory stocking and replenishments decisions be made centrally or
in a decentralized fashion ? Should inventory be held at central warehouses or in regional warehouses ?
How much ? Where should inventory be deployed ? Should it be held at a central location or
should it be pushed forward closer to the retailer ? If there is a shortage of stock how should it be allocated ?
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 41
Central vs. Decentralized Inventory Planning Centralized Takes an overall cost view and hence may result in optimal stock positioning Harder to implement and makes decision making less responsive Multi-party supply chains are the de facto standard in most industries – stages
are owned by different entities Decentralized Does not take a global view of costs involved in making individual stage
decisions
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 42
“From a very early age , we are taught to break apart problems , to fragment the world. This apparently makes complex tasks……more manageable , but we pay a hidden , enormous price. We can no longer see the consequences of our actions , we lose our intrinsic sense of connection to a larger whole. When we try to ‘see the big picture’, we try to reassemble the fragments in our minds, to list and organize all the pieces….the task is futile…..after a while we give up trying to see the whole altogether.” Peter Senge The Fifth Discipline
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 43
Complexity in multi stage inventory management Each stage does not realize the implications of inventory holding and shortages
at stages downstream and upstream to it. “If I do not hold the appropriate amount of inventory, my downstream stages
may have to hold even more”
Pooling effects
Upstream responsibility
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 44
Risk Pooling Consider these two systems:
Warehouse 1
Market 1
Warehouse 2
Market 2
Supplier
Market 1 Supplier
Centralized Warehouse Market 2
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 45
Sums of random variables normal distribution X
standard deviation σX
normal distribution Y
standard deviation σY
• assuming X and Y independent
variance σ2 for X+Y = σx2 + σY2
Simple simulation experiment
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 46
Supply chain implications if demand at n retailers have standard deviations σ1, σ2, …, σn demand at centralized warehouse that serves them has standard deviation and if demand at different retailers is independent
σ = σ 1 + σ 2 + ... + σ n 2
2
2
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 47
Safety stock location
goal: 95% demand filled from stock
total SS requirement if SS kept at retailers as compared to SS requirement if SS kept at centralized warehouse is
1.
more
2.
less
3.
the same
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 48
Safety stock location
SS kept at retailers
kσ1 + kσ2 +…+ kσn
SS kept at centralized warehouse?
k σ 12 + σ 2 2 + ... + σ n 2
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 49
Safety stock location
Assume all retailers have same demand variance
SS kept at retailers
σ1 = σ2 =…= σn = σ kσ1 + kσ2 +…+ kσn = n kσ
SS kept at centralized warehouse
k σ + σ 2 + ... + σ n = k nσ = nkσ 2 1
2
2
2
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 50
Risk Pooling For the same service level, which system will require more
inventory? For the same total inventory level, which system will have
better service? What factors affect these answers?
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 51
Aside on Stock Positioning Two types of demand correlation to be considered:
Auto-correlation: That is the correlation of the demand with itself in a time period. Example: You are less likely to buy a car tomorrow if you bought one today.
Inter-correlation(or cross-correlation): That is the correlation of demands that are realized by different retailers. Example:If you buy your car from one retailer, that means you won’t buy from another one in close future assuming you want to but only one car. Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 52
Risk Pooling
Centralizing inventory reduces both safety stock and average inventory level for the same service level
This works best for – High coefficient of variation, which reduces required safety
stock – Negatively correlated demand. Why?
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 53
Tradeoff centralization
lower inventories
operating efficiency
lower responsiveness
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 54
Stock Positioning Where to position safety stock ? Downstream Closer to customer Typically downstream units are the most expensive places to hold stock Pooling and aggregation of uncertainty effects are lost Upstream Farther from the customer May not provide the desired service level The inventory manager’s job is to create a fine balance
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 55
Safety Stock Positioning Autocorrelation of demands brings a tendency to hold stock at downstream
levels Intercorrelation of demands brings a tendency to hold SS at upper levels. An optimization approach for jointly correlated demands needs to be
introduced Unfortunately, your ERP systems don’t solve this problem Some vendors have started product offerings:
Optiant
Smart Ops
ToolsGroup
The question of optimally positioning stock is however a very context and
company specific issue. It requires specific models Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 56
Revenue Sharing as a tool to align incentives
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 57
Thought Experiment You are a retailer. You buy a widget at a wholesale price of $3.00 from a
supplier and sell it for $ 4.00. It costs your supplier $2.00 to make that widget Unsold widgets are marked down and sold at $1.50 Demand is uncertain and uniformly distributed between 5 and 10 It is equally likely that demand is 5,6,7,8,9 or 10 units
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 58
Quiz: Understocked vs. Overstocked supply chains Stockout for the manufacturer is not a stockout for the retailer. If
the customer does not find Coke he is willing to settle for Pepsi Manufacturer is offering quantity discounts and huge seasonal
promotions Retailer has small margins Retailer has to take huge markdown when the product does not sell Manufacturer is keen to meet analyst expectations and is pushing
product into the channel
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 59
The channel as a whole Marginal Revenue/Costs
c
Qint
Q
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 60
The channel as a whole Marginal Revenue/Costs
c
Qint
Q
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 61
Traditional setting Marginal Revenue/Costs
w
c
Qd
Qint
Q
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 62
With revenue sharing Marginal Revenue/Costs
w
c
w = λc
Qd
Qint
Q
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 63
Non Linear pricing motivation Marginal Revenue/Costs
c
Qint
Q
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 64
Order Minimums
To illustrate typical structure only. A detailed ordering matrix can be obtained from Rentrak website for each studio
Source: Rentrak Website
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 65
Terms Vary
••$$0.00 0.00upfront upfrontfees fees •• 65/35% 65/35%split split(you (youkeep keep65%) 65%)vs vs$1.05 $1.05minimum minimumtransaction transactionfee feefor forfull fullterm term
••Sell Sellon on29th 29thday day(VHS (VHScost cost= =$2.00 $2.00vs vs35% 35%//DVD DVDcost cost= =$3.50 $3.50vs vs35% 35%––can cansell sell50%) 50%) (VHS/DVD) (VHS/DVD) ••66month want purchase monthlease leaseterm term (cost (cost= =$0.50) $0.50)ififyou wantto to •you $0.00 up-front fee purchase • $0.00 up-front fee
••$$3.75 3.75upfront upfrontfees fees •• 44/56% 44/56%split split(you (youkeep keep56%) 56%)vs vs$1.15 $1.15minimum minimumtransaction transactionfee feefor forfull fullterm term
••Sell Sellon on41st 41stday day(VHS (VHScost cost= =$2.40 $2.40can cansell sellall allbut butone) one) ••66month monthlease leaseterm term (cost (cost= =$1.20) $1.20)ififyou youwant wantto topurchase purchase
Source: Rentrak Website
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 66
Terms Vary
••$$0.00 0.00upfront upfrontfees fees •• 61/39% 61/39%split split(you (youkeep keep61%) 61%)vs vs$1.15 $1.15minimum minimumtransaction transactionfee feefor forfull fullterm term
••Sell Sellon on28th 28thday day(VHS (VHScost cost= =$2.00 $2.00vs vs39% 39%//DVD DVDcost cost= =$3.00 $3.00vs vs39%) 39%) (VHS/DVD) (VHS/DVD) ••66month want purchase monthlease leaseterm term (cost (cost= =$1.50) $1.50)ififyou wantto to •you $0.00 up-front fee purchase • $0.00 up-front fee
••$$1.25 1.25upfront upfrontfees fees •• 58/42% 58/42%split split(you (youkeep keep58%) 58%)vs vs$1.25 $1.25minimum minimumtransaction transactionfee feefor forfull fullterm term
••Sell Sellon on31st 31stday day(VHS (VHScost cost= =$1.70 $1.70DVD DVDCost Cost= =$$3.35) 3.35) ••66month monthlease leaseterm term (cost (cost= =$1.75) $1.75)ififyou youwant wantto topurchase purchase
Source: Rentrak Website
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 67
Revenue Sharing vs. Fixed Price A titles
B titles
C titles
Source: Mortimer 2002, Working Paper
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 68
Please feel free to send questions, comments, critique, war-stories etc. to pyadav@zlc.edu.es pyadav @ mit.edu
Yadav –Supply Chain Management-Tsukuba Univ. © 2007 Prashant Yadav 69