Supply-Chain Management:A View of the Future

Leroy B. SchwarzKrannert School of Management

Purdue University

Supported by e-Enterprise Center at Discovery Park

Outline

• Supply-Chain Management of “Yesterday” – How Modeled

– How Practiced

• Supply-Chain Management of “Today” – How Practiced

– How Modeled

Outline (cont.)

• Introduce Paradigm called:

“IDIB Portfolio”

• Describe My Vision of the “Future”of SCM

• Provide an Overview of 2 Projects

• Collaborative Decision-Making and Implementation

• Secure Supply-Chain Collaboration

SCM Models of “Yesterday”

• Took Centralized Perspective– Assumed Single, Systemwide Objective

Function: F(x1, x2, x3, ...)

– Assumed System Information was:• Available• Omnipresent

– Assumed Implementation was “Contractible”

• Typical Results: – Characteristics of the Optimal Policy for

Special Structures• Clark & Scarf, ‘60• Schwarz, ‘73

– Examination of Heuristics for More General Structures

• Clark & Scarf, ‘62• Roundy, ‘85

SCM Practice of “Yesterday”

• Single-Owner Chains Took a Centralized Perspective– Single Objective Function: F(x1, x2, x3, ...)

– De-Centralized Decision-Making

– Information: Not Available or, at best, “Asymmetric”

– Implementation: De-Centralized; NOT Contractible

• Consequently: – “Supply Chains” Managed as Separate Entities,

regardless of their ownershipEx.: Local Objective Functions: F1(x1), F2(x2), ...

• Examples– USAF Logistics Command Consumable

Inventory System

– IBM Service-Parts Inventory System

Consequences of this:

==> Huge Buffers– Raw, WIP, and Finished-Goods Inventories

– Capacity Buffers (e.g., understated capacity)

– Leadtime Buffers (e.g., overstated leadtime)

“Yesterday’s” Relationship: “Mismatched”

• Models– Too Specialized

– Required More Information than Practice Had

• Practice– Inexperienced with Models & Computers

– Confused by Models

– Suspicious of Models

SCM Practice “Today”

• The Beginnings of “Real” SCM for Single-Owner Chains– Ex: Wal-Mart’s Retail Link

Target’s Partners OnLine

• Capabilities– Broadcast SKU-level Data Across the Chain

– Observe Status ==> Implemetation “Contractible”

• Results:– Huge Reductions in Buffers ==> Lower

Operating Costs

– Improved Competitiveness• Lower Prices• More Customization• Higher Availability

• Development of Technologies to Support Multiple-Owner SCM

• Internet is Providing Experience• E-Markets

– Providing Buyer-Supplier Linkages• Data Standardization; e.g. RosettaNet

• Beginnings of SCM for Multiple-Owner Supply Chains– VMI, Quick Repsonse

– VICS’ CPFR Campaign

• Huge Challenges for Multi-Owner Chains– Multiple — often Conflicting — Objective

Functions

– Technical Difficulties in Sharing Information• SKU Identification• Time-Frame

– Fear about Information Sharing• Vertical “Leakage”• Horizontal “Leakage”

SCM Models of “Today”• Models with Multi-Ownership, Competing

Objective Functions, and Asymmetric Information– Roots in Economics

– 1980’s Work of Monahan, Pasternak

– Contemporary Work• “Supply-Chain Coordination with Contracts”, G.

Cachon (forthcoming)• “Information-Sharing and Supply-Chain

Coordination”, F. Chen (forthcoming)

• Models for Assessing the Impact of Decentralized Decision-Making and/or Asymmetric Information– Ex: Lee, et al. “Bullwhip” Paper (MS 43:4)

• Results:– Assessments of “Agency Loss”

• Non-bathtub Shaped Loss Functions

– Contracting Mechanisms to Improve/Optimize Performance

Relationship “Today”: “Out of Step”

• Models beginning to include ownership and private-information issues, but

– Little Work on How to Share Information or How to Collaborate on Decision-Making or Implementation

– Ignoring the Development of More Sophisticated “Centralized” Models

Relationship “Today”: “Out of Step”

• Practice ready to “Dance” but No Model “Partner”

– Using simple models based on “pull down” menus in ERP systems

– “Swimming” in Data, but uncertain about how to use it

What About the Future

of SCM?

First.......

The IDIB Portfolio

a.k.a.The Information, Decision-Making,

Implementation, Buffer Portfolio

“Managing” anything can be viewed as 4 related activities:

• Getting Information• Making Decisions• Implementing Decisions• Buffering against Imperfections in

information, decision-making, or implementation

Every “Management System” is, in fact, 4 Sub-Systems

• The Information System provides information

• The Decision-Making System makes decisions

• The Implementation System implements decisions

• The Buffer System copes with imperfections in information, decision-making, or implementation

Each Sub-System has Cost and Quality Characteristics

The Information System– Quality Characteristics

• Accuracy• Leadtime• Aggregation Level• Horizon• Etc.

– Cost: Increasing and Marginally-Increasing with Quality

Each ... Characteristics (cont.)

The Decision-Making System– Quality Characteristics

• “Optimality”; i.e., “how good”?• Leadtime; i.e., “how long to make”?• Etc.

– Cost: Increasing and Marginally-Increasing with Quality

Each ... Characteristics (cont.)

The Implementation System– Quality Characteristics

• Accuracy; i.e., conformance to decision • Leadtime; i.e., “how long to implement”• Etc.

– Cost: Increasing and Marginally-Increasing with Quality

Each ... Characteristics (cont.)

The Buffer System– Quality Characteristics

• Form • Robustness• Etc.

– Cost: Increasing and Marginally-Increasing with Quality

IDIB “Portfolio”?

• Like a Financial Portfolio, the IDIB System requires an investment of Dollars

• Like a Financial Porfolio, each Subsystem’s Characteristics Should Complement the Characteristics of the Others– Ex: Robust Buffer System Complements an

Inaccurate Information System

– Ex: Tradeoffs Among Buffer Sub-Systems

Managing the IDIB Portfolio....

.... means changing the nature and quality of its 4 sub-systems so that total portfolio cost — which includes the cost of imperfect buffering — is minimized

This is NOT Rocket Science!

Most Operations-Research Models Ignore the IDIB Portfolio

• Example: The Newsvendor Model– Information-System Quality Assumed

– Implementation is Ignored

– Select Decision-Rule to Minimize Buffer-System Cost

IDIB Portfolio View of Newsvendor “Problem”

• The “Problem” is that acquistion/production decsion must be made before demand occurs

• What if:– Production was instantaneous?

– Production Decision and Implementation Leadtime ≤ “Horizon” of Known Demand?

What is the Value-Added of the IDIB Paradigm?

• Vantage Point on the Majority of Operations-Research Models

• Vantage Point on Past/Present Practice

• Vantage Point on the Future

1st Axiom of the IDIB Portfolio:

Given an existing IDIB Portfolio, increasing the quality of one of its components typically facilitates decreasing the quality of at least one of its other three components while maintaining the same level of customer service

“the Tradeoff Axiom”

Examples:

• In a (Q,r) system:– If all leadtimes are fixed, then the information-

system, decision-making, and implementation leadtimes tradeoff one-for-one

– If any of these leadtimes are variable, then reducing their variance facilitates reducing safety stock (buffer) inventory

Examples from Practice:

• Schneider National– Increasing Quality of I, D, and I; Reducing B;

improving service

• Manufacturer Making Transition from a “Push” (e.g., MRP) to “Pull” (e.g., JIT)– Reducing Buffer Inventory, increasing Buffer

Capacity

• Domestic Manufacturer Outsourcing to Off-Shore Supplier– Reducing Implementation Quality (Leadtime);

Increasing Buffer Inventory

The IDIB Perspective on State-of-the-Art Practice in SCM

• Involves the sharing of past, present, and future-oriented information between buyer-supplier pair; and/or

• Involves delegation of decision-making or implementation to the supplier

.....So, then what is the future.......?

2nd Axiom of the IDIB Portfolio:

Investment to improve the quality of any single component of the IDIB Portfolio will, over some range, decrease total cost of the Portfolio; but, beyond some quality level, increase total cost of the Portfolio

“Do-Nothing-in-Excess Axiom”

The Future of Supply-Chain Management Involves

Collaborative Decision-Making and/or Implementation

Why?

• For Supply Chains that already share information, the returns from additional information sharing are diminishing

• For Supply Chains that are already delegating some decision-making, the returns from additional delegation are marginally diminishing

Two Personal Projects

• Models for Collaborative Decision-Making– How to Improve Decision-Making and

Implementation Based on Shared Information

• Protocols for Secure Supply-Chain Management– How to Improve Decision-Making and

Implementation without Sharing Information

Models for Collaborative Supply-Chain Decision-Making

withVinayak Deshpande

&Jennifer Ryan

Starting Point is “Collaborative Planning, Forecasting, and Replenishment” (CPFR)

What is CPFR?

• A process model, shared by the buyer and supplier, through which inventory status-, forecast-, and promotion-oriented information are shared and replenishment decisions generated

The 9 Process Steps: Step 1: Develop Front-End Agreement: Roles,

Measurement, Readiness

Step 2: Create Joint Business Plan: Strategies and

Tactics

Step 3: Create Sales Forecast: Buyer or Supplier

Step 4: Identify Exceptions for Sales Forecast

The 9 Process Steps: Step 5: Resolve/Collaborate on Exception

Items

Step 6: Create Order Forecast

Step 7: Identify Exceptions for Order Forecast

Step 8: Resolve/Collaborate on Exception Items

Step 9: Order Generation

CORNINGConsumer Products

MeadSchool & Office

FederatedFederated Department StoresDepartment Stores

Schnuck Markets

JCPenney Staples

QRSBenchmarkingPartners

FIELDCREST CANNON

CPFR: Who’s Behind it?

CPFR History:

• ‘95/96: Wal-Mart Warner-Lambert “CFAR” Pilot

• ‘97: VICS Develops CPFR Initiative• ‘98: VICS CPFR Guidelines Published• ‘99: Pilots Between

– Kimberly-Clark & K-Mart, – P&G & Meier, Target, Wal-Mart– Nabisco & Wegman’s, etc.

• ‘00:1st Production Rollout: K-Mart

CPFR’s Future:

• “n-Tier” Collaboration– Extension to Include Master-Scheduling

Decisions

– Include Transportation

Research Topics in CPFR:

• Process Model: How and Where does the CPFR model (e.g., forecast collaboration) fit into the supply-chain process?

• Front-End Agreements: How Should agreements be structured, performance measured, and benefits shared?

• Data Sharing: How should data be shared (aggregation/disaggregation issues)?

• Exception Processing: What constitutes an exception?

Secure Supply-Chain Collaboration

withMikhail Atallah

&Vinayak Deshpande

The Starting Point....

“Information Asymmetry” is one of the major sources of inefficiency in Managing Supply Chains

==> Wrong Investment in Capacity==> Misallocation of Resources==> Distorted Prices==> Reduced Customer Service==> Unnecessary Additional Costs

.... there are Very Good Reasons for Keeping Private Information

Private

• Fear that Supply-Chain Partner will Take Advantage of Private Information

• Fear that Private Information will Leak to a Competitor

So, then, the Obvious Question...

Is it possible to enjoy the benefits of Information-Sharing without Disclosing Private Information?

It Depends

If the Value of Private Information is the Information

Itself, then..

...obviously, information must be disclosed for value

to be created

But, if the Value of Private Information is a Decision .......

...then it is possible to create value without Disclosing Private

Information

Example

In CPFR: Determine agreed-upon planned orders without sharing forecasts, etc.

Secure Multi-Party Computation

• SMC is Decades Old• Elegant Theory• General Results w.r.t. Existence,

Complexity, etc.• Recently, Practical Protocols for Specific

ProblemsEx. Electronic Voting

Information Retrieval

SMC Paradigm

• Alice has Private Information: XA

• Bob has Private Information: XB

• Want to Determine f(XA, XB)

• f(XA, XB) is well defined

• No Trusted Third Party

• Provide f(XA, XB) to Alice, Bob, both, or Neither

We are Developing Secure Multi-Party Protocols for Supply-

Chain Management:

“Secure Supply-Chain Collaboration”

More Specifically...

...we are developing protocols to enable Supply-Chain Partners to

Make Decisions that Cooperatively Achieve Desired System Goals without Revealing

Private Information

Our Goals:

• Develop and Apply SSCC Protocols to Some Well-Known SCM Problems

• Simple e-Auction Scenarios• Simple Capacity-Allocation Scenarios• Bullwhip Scenarios

• Compare Effectiveness of Protocols vs. non-cooperative decision-making

Our Goals (cont.):

• Develop Proof-of-Concept Software

• Examine Security versus Cost Tradeoffs

Ex: Capacity Allocation • Single Supplier; N Retailers; Single Sales

Period• Supplier has constant marginal production

cost, but fixed capacity, K• Retailers operate in non-competing markets;

each retailer i has private information, i, about its market that influences its order to the supplier; Supplier has prior Pr()

• If Ordersi > K, Supplier Uses Pre-announced Allocation Mechanism

Cachon and Lariviere (MS, ‘99)

• Examine this scenario from perspective of the retailers in non-cooperative setting

• Linear Demand: Market-Clearing Price, r(q)

r(q) = i - q • Several Very Interesting Results

– Retailers will over-order even if Pareto allocation mechanism is used

– Supplier and Supply-Chain Profit can increase if a truth-telling mechanism is replaced by manipulable one.

Deshpande & Schwarz (‘02)

• Examine this scenario — and a newsvendor scenario — from perspective of maximizing Supply-Chain Profit assuming truth-telling

• Derive conditions under which two commonly-used allocation mechanisms maximize supply-chain profit

• Our SSCC Protocols use these mechanisms without revealing the retailers’ i’s

Allocation Mechanisms

• Supplier has Capacity K• Retailers place orders: q1, q2, q3,..qN

• Assume qi > K

• Linear Allocation: qi’ = qi - (qi- K)/N’

• Proportional Allocation: qi’ = qi • (K/ qi )

Proportional Allocation Protocol

1.Retailers choose a random R;2.Every retailer sends its R•qi to Supplier3.System computes:

D = (R•qi/K) and sends it to all the retailers

4.Every retailer computes its allocation: qi’ = R•qi/D

and sends to supplier

Notes:

• We are assuming that retailers will tell the truth; i.e., reveal the quantity they truly want; (one that is consisent with their i)

• Supply Chain Profit will be reduced if they don’t

• Contracting Mechanisms will be Required

Notes:

• The Supplier Learns each Retailer’s qi, but not i

• Supplier Might be able to Infer i

• Shipping Proxies

We Have Only Just Begun...• Tough Issues to Deal with:

– SMC Complexities; e.g.,• How to Deal with Collusion• Computational Complexity (e.g., simultaneity)

– Supply-Chain Modeling Complexities; e.g.• Contracting/Incentive Issues

– SSCC Complexities; e.g., • Inverse Optimization• Bob’s Objective is fB(xA, xB); Alice’s is fA((xA, xB)

Discussion....