1 On-Line Seminar Series December 15, 2000 LGO-SDM ALUMNI.

1

On-Line Seminar SeriesOn-Line Seminar SeriesOn-Line Seminar SeriesOn-Line Seminar Series

December 15, 2000

LGO-SDM ALUMNI

2

Supply Chain Modeling Supply Chain Modeling and Optimizationand Optimization

Supply Chain Modeling Supply Chain Modeling and Optimizationand Optimization

Stephen C. Graves

MIT, E40-439

Leaders for Manufacturing Program

[email protected]

http://web.mit.edu/sgraves/www/

3

OverviewOverviewOverviewOverview

• Supply chain management: definition & issues• Framework for supply chain modeling• Example: safety stock placement• Example: supply chain configuration• Example: inventory hedge for cyclic demand• Other examples and future plans

4

Supply Chain ManagementSupply Chain ManagementSupply Chain ManagementSupply Chain Management

• A process-oriented integrated approach to procuring, producing and delivering products and services to customers

• Includes suppliers, internal operations, channel partners and end customers

• Covers the management of materials, information and funds

5

Challenges and IssuesChallenges and IssuesChallenges and IssuesChallenges and Issues

• Process orientation, rather than functional• Process metrics and accounting• IT as an enabling technology• Need to cross organizational boundaries• Need to focus on system performance rather than

department or company performance

6

Network Representation of a Network Representation of a Supply ChainSupply ChainNetwork Representation of a Network Representation of a Supply ChainSupply Chain

US Suppliers

Off Shore Suppliers

US Factory

Singapore Factory

US DC

European DC

Asia/Pacific DC

Americas Demand

European Demand

Asia/Pacific DemandKit Suppliers

7

Framework for Supply Framework for Supply Chain ModelingChain ModelingFramework for Supply Framework for Supply Chain ModelingChain Modeling

• Decision Hierarchy– Strategic --- network design– Tactical --- flow planning; countermeasures for

variability and uncertainty– Operational --- scheduling and control

• My interest – tactical level

8

Strategic Safety Stock Strategic Safety Stock PlacementPlacementStrategic Safety Stock Strategic Safety Stock PlacementPlacement

• Intent - a tactical model to determine the amount and positioning of safety stocks in supply chains

• Intent - a tactical model to support supply chain improvement teams

9

KIMES 100KIMES 100KIMES 100KIMES 100

10


11


12

Supply Chain Supply Chain BEFOREBEFORESupply Chain Supply Chain BEFOREBEFORE

13

Supply Chain Supply Chain LEAD TIMESLEAD TIMESSupply Chain Supply Chain LEAD TIMESLEAD TIMES

14

Supply Chain Supply Chain COSTSCOSTSSupply Chain Supply Chain COSTSCOSTS

15

Supply Chain Supply Chain OPTIMIZEDOPTIMIZEDSupply Chain Supply Chain OPTIMIZEDOPTIMIZED

16

Supply Chain Supply Chain IMPLEMENTEDIMPLEMENTEDSupply Chain Supply Chain IMPLEMENTEDIMPLEMENTED

17

Key BenefitsKey BenefitsKey BenefitsKey Benefits

• Shows value from “holistic” perspective• Formalizes inventory-related supply chain costs, and

provides an optimal benchmark• Provides framework and terminology for cross-

functional debate• Shows the effectiveness of inventory, strategically

positioned in a few places to de-couple the supply chain

18

Key LearningsKey LearningsKey LearningsKey Learnings

• De-couple supply chain prior to a high-cost added stage; and prior to product explosion

• Substitute information for inventory• Postpone product differentiation step• Win-win from optimizing multi-company supply chain• Value of a standard terminology and a neutral tool

19

Supply Chain Configuration Supply Chain Configuration for New Productsfor New ProductsSupply Chain Configuration Supply Chain Configuration for New Productsfor New Products

• How to configure a new product’s supply chain?• Two vendors can deliver an identical product.

– A quotes 100 days at $1.00 per unit– B quotes 3 days at $1.10 per unit

• Which one do you pick?

20

Problem StatementProblem StatementProblem StatementProblem Statement

• A supply chain configuration is the set of options selected for each stage in supply chain

• Stages include procurement; production, assembly and test processes; distribution channels; and transportation modes

• Intent: develop a DSS for determining options in SC configuration, given a stable product design

21

Digital Camera ExampleDigital Camera ExampleDigital Camera ExampleDigital Camera Example

• Pro-summer model– Monthly demand has a mean of ~550 units and a

standard deviation of ~50• Three major subassemblies• Two customer markets: US and export

22

Relevant Supply Chain CostsRelevant Supply Chain CostsRelevant Supply Chain CostsRelevant Supply Chain Costs

• Cost-of-goods sold• Safety stock cost• Pipeline stock cost• Time-to-market cost• Quality cost

23

Current Practice = Target Current Practice = Target CostingCostingCurrent Practice = Target Current Practice = Target CostingCosting

• Target for unit manufacturing cost (UMC) set based on market price and gross margin

• UMC is then allocated to subassemblies• Designers independently source their portion of the

supply chain • Other factors (functionality, quality, flexibility)

considered based on thresholds

24

Digital Camera Supply ChainDigital Camera Supply ChainDigital Camera Supply ChainDigital Camera Supply Chain

25

Component/Process Product.Description Option Time CostRaw Silicate 1 60 $5

2 20 $8Wafer Fab 1 30 $800

8 $825Wafer Pkg. and Test 1 10 $200

5 $225CCD Assembly 1 5 $200

2 $250Miscellaneous Components 1 30 $200Parts w/ 8 Week LT 1 40 $105

2 20 $1083 10 $1094 0 $110

Parts w/ 4 Week LT 1 20 $1752 10 $1773 0 $179

Note: All data has been disguised by scaling

Digital Camera OptionsDigital Camera OptionsDigital Camera OptionsDigital Camera Options

26

Component/Process Product.Description Option Time CostParts w/ 2 Week LT 1 10 $200

2 0 $203Parts on Consignment 1 0 $225Circuit Board Assembly 1 20 $225

2 5 $300Camera Body 1 70 $650

2 30 $665Accessory Processing 1 40 $100Local Accessory Inv. 1 10 $60Camera Assembly 1 6 $420

2 3 $520Central Distribution 1 5 $180US Demand 1 5 $12

2 1 $25Export Demand 1 11 $15

2 2 $40


27

Three Solution ApproachesThree Solution ApproachesThree Solution ApproachesThree Solution Approaches

• Minimize unit manufacturing cost• Minimize production time• Minimize supply chain costs

28

$3,794$4,078$3,756$3,756Unit Manufacturing Cost

118 days45 days127 days127 daysLength of Longest Path

18.920.019.019.1Total Configuration Cost

0.9 0.6 1.2 1.3Inventory Cost ($MM)

18.019.417.817.8COGS ($MM)

Min SC Costs

Min Prod Time

Min UMCCurrent Policy

Solution ComparisonSolution ComparisonSolution ComparisonSolution Comparison

29

Component/Process Product.Description Option Time CostRaw Silicate 1 60 $5

2 20 $8Wafer Fab 1 30 $800

8 $825Wafer Pkg. and Test 1 10 $200

5 $225CCD Assembly 1 5 $200

2 $250Miscellaneous Components 1 30 $200Parts w/ 8 Week LT 1 40 $105

2 20 $1083 10 $1094 0 $110

Parts w/ 4 Week LT 1 20 $1752 10 $1773 0 $179


30

Component/Process Product.Description Option Time CostParts w/ 2 Week LT 1 10 $200

2 0 $203Parts on Consignment 1 0 $225Circuit Board Assembly 1 20 $225

2 5 $300Camera Body 1 70 $650

2 30 $665Accessory Processing 1 40 $100Local Accessory Inv. 1 10 $60Camera Assembly 1 6 $420

2 3 $520Central Distribution 1 5 $180US Demand 1 5 $12

2 1 $25Export Demand 1 11 $15

2 2 $40


31

15% 30% 45% 60%Raw Silicate 1 1 1 1Wafer Fab 1 1 1 2Wafer Pkg. and Test 1 1 1 1CCD Assembly 1 1 1 1Miscellaneous Components 1 1 1 1Parts w/ 8 Week LT 1 3 4 4Parts w/ 4 Week LT 1 2 3 3Parts w/ 2 Week LT 1 1 2 2Parts on Consignment 1 1 1 1Circuit Board Assembly 1 1 1 1Base Assembly 2 2 2 2Accessory Processing 1 1 1 1Local Accessory Inv. 1 1 1 1Digital Capture Device Assembly 1 1 1 1Central Distribution 1 1 1 1US Demand 1 2 2 2Export Demand 1 2 2 2

Role of Holding CostRole of Holding CostRole of Holding CostRole of Holding Cost

32

18.23,8252.5

18.13,8002.7

18.03,7942.8

17.93,7733.3

COGS($MM)

UMC($/unit)

Initial Investment($MM)

Inventory Investment and Inventory Investment and UMC InteractionUMC InteractionInventory Investment and Inventory Investment and UMC InteractionUMC Interaction

33

Key LearningsKey LearningsKey LearningsKey Learnings

• SC optimization saves three times the savings from SIP

• Optimization did not make some “obvious” choices• Increasing unit manufacturing cost by $37 is

significant. • As you move farther downstream in the supply chain,

higher cost options can be more attractive• More complex the supply chain, more likely

optimization will find opportunities

34

Next StepsNext StepsNext StepsNext Steps

• Verify/validate the model in practice• Software to disseminate expected, March 01• Incorporate side constraints, e.g. number of vendors• Extend to consider capacity options

35

Creating an Inventory Creating an Inventory Hedge for Cyclic DemandHedge for Cyclic DemandCreating an Inventory Creating an Inventory Hedge for Cyclic DemandHedge for Cyclic Demand

• How to plan long lead-time, custom product subject to cyclic demand?

• Motivations: Teradyne case• Intent: demonstrate effectiveness of planning tactic

(the hedging policy) and develop a stylized model for guiding its deployment

36

Teradyne CaseTeradyne CaseTeradyne CaseTeradyne Case

• Flagship product is Catalyst, for testing linear and mixed signal devices

• Tester sells for $1.5 - 2.0 M• Each sale is for a customized product• Product is complex, 100’s of PCB’s, 10000’s of

components• Customer delivery time << Manufacturing lead time• Demand is volatile with limited predictability

37

Product StructureProduct StructureProduct StructureProduct Structure

• Three levels to BOM: option level, PCB, component• 175 options; each option consists of 1 - 8 PCB’s• Each tester consists of about 50 options, plus work

station, test head, mechanical assembly

38

DemandDemandDemandDemand

• Aggregate demand very volatile, subject to bull whip, alternates between up cycle and down cycle

• About 30% of options have stable demand; and account for over 70% of material cost

• 50% of options are used infrequently, account for less than 15% of cost

39

Figure 1: ICD’s Sales by quarter for past 5 years

Quarterly Sales Data (Q2 1994 - Q1 1999)

0

0.5

1

1.5

2

2.5

3

1994

Q2

1994

Q3

1994

Q4

1995

Q1

1995

Q2

1995

Q3

1995

Q4

1996

Q1

1996

Q2

1996

Q3

1996

Q4

1997

Q1

1997

Q2

1997

Q3

1997

Q4

1998

Q1

1998

Q2

1998

Q3

1998

Q4

1999

Q1

Quarter

Sal

es

40

Production PlanningProduction PlanningProduction PlanningProduction Planning

• Testers are assembled to order• MPS at the option level to fill material pipeline• MPS assumes an aggregate demand rate (x per

week)• MPS assumes a planning bill to plan open testers• Scheduling group matches ‘potential’ and booked

orders to open testers, and re-schedules options as necessary

• Options not in planning bill are scheduled in ad hoc way

41

Hedging PolicyHedging PolicyHedging PolicyHedging Policy

• MPS for next L weeks assumes current aggregate demand rate, x per week.

• MPS beyond L weeks assumes a “maximal” demand rate, y per week, y > x

• This creates an intermediate-decoupling inventory, sized to meet a “maximal” demand rate

• This inventory permits ICD to ramp quickly when transition from down to up cycle

42

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

35 33 31 29 27 25 23 21 19 17 15 13 11 9 7 5

Lead-Time

Cos

tPossible Hedging Points

Cost Accrual Function -- Suggests hedging points

43

Implementation ResultsImplementation ResultsImplementation ResultsImplementation Results

• Considerably better performance: $50 - $100 million in incremental revenue was realized over recent upturn

• More systematic treatment of materials management than before

44

Research QuestionsResearch QuestionsResearch QuestionsResearch Questions

• The role of hedging stocks in environments subject to non-stationary demand– Where does one locate the hedging stock?– How is the hedging stock sized?– What kind of service performance will be attained

for a given hedging stock configuration

45

Hedging ModelHedging ModelHedging ModelHedging Model

• Principal issues to address– Need an effective strategy to manage the inventory

pipeline when demand is non-stationary and changes without predictability

– Need model to characterize inventory levels and service performance, and illuminate tradeoffs

46

t = 0

System ships

Replenishment orders

Cumulative value of material purchased as a function of time

t = m

Diagram of a supply pipelineDiagram of a supply pipelineDiagram of a supply pipelineDiagram of a supply pipeline

47

Intermediate-decoupling Inventory FGI

m-L L

• Where to locate the decoupling stock (how to set L)?

– decoupling stock is sized for the high-demand rate

• How does one characterize service level, given non-stationary demand?

48

Optimization ProblemOptimization ProblemOptimization ProblemOptimization Problem

• Decision variables L,SD,SH

• Objective function: minimize average expected holding costs over a cycle

• Subject to: Target fill rates during the low to high transient sub-cycle, the high demand and low demand sub-cycles

49

mLl

SS

L

LSLSLL

m

L

L

LS

m

Lm

L

LS

L

LS

OHE

DH

HDDH

DDHDDH

H

HH

HH

HH

D

D

DD

0,

)(

)(3)(3)()(

3

2)()(

)(

)(

:ToSubject

][ Minimize

2/12/12/32/3

50

0

20000

40000

60000

80000

100000

120000

140000

0 5 10 15 20 25 30 35

L

TC

*

Figure 2: The optimal objective function value as a function of L

51

Conclusions and LearningsConclusions and LearningsConclusions and LearningsConclusions and Learnings

• Can use model to understand the value of hedging policy

• Can use optimization model to determine the location of a decoupling inventory and base stock levels for both inventories

• Find that optimal value of L lies near the kink in the lead-time cost accrual profile

52

Other ExamplesOther ExamplesOther ExamplesOther Examples

• Seasonal planning with uncertainty (Monsanto)• Evaluation of flexibility benefits (GM)

53

Current/Future PlansCurrent/Future PlansCurrent/Future PlansCurrent/Future Plans

• Planning tactics for low-volume custom-engineered products (ABB)

• Evaluation of capacity options in supply chain design (TBD)

• Evaluation and utilization of POS data (Kodak?)

54

Further InformationFurther InformationFurther InformationFurther Information

• Papers and theses: http://web.mit.edu/sgraves/www/

• SIP model software: http://web.mit.edu/lgorg3/www/

• [email protected] 617.253.6602

1 On-Line Seminar Series December 15, 2000 LGO-SDM ALUMNI.

Documents

Transcript of 1 On-Line Seminar Series December 15, 2000 LGO-SDM ALUMNI.