Chapter 6: Supply Chain Management (SCM)

IE 3265 – POMR. R. Lindeke

UMD-MIE

Topics For Discussion: Defining the issues of SCM Major Players

How we can work in this new model Vender relationships Data Management

Major Issues: Bullwhip effect Transportation problems Location issues

What is SCM?

“Supply Chain management deals with the control of materials, information, and financial flows in a network consisting of suppliers, manufacturers, distributors, and customers” (Stanford Supply Chain Forum Website)

“Call it distribution or logistics or supply chain management... In industry after industry . . . executives have plucked this once dismal discipline off the loading dock and placed it near the top of the corporate agenda. Hard-pressed to knock out competitors on quality or price, companies are trying to gain an edge through their ability to deliver the right stuff in the right amount of time” (Fortune Magazine, 1994)

Growing Interest in SCM – Why?

As manufacturing becomes more efficient (or is outsourced), companies look for ways to reduce costs

Several significant success stories: Efficient SCM at Walmart, HP, Dell Computer

SCM considers the broad, integrated, view of materials management from purchasing through distribution

The huge growth of interest in the web has spawned web-based models for supply chains: from “dot com” retailers to B-2-B business models

Mass Customization:Designing Final Choices into Supply Chains

Several companies have been able to cut costs and improve service by postponing the final configuration of the product until the latest possible point in the supply chain. Examples: Hewlett Packard printer configuration Postponement of final programming of

semiconductor devices – all routines loaded, only certain ones activated

Assemble to order rather than assemble to stock (Dell Computer)

Design For Logistics: Many firms now consider SCM issues in

the design phase of product development

One example is IKEA whose furniture comes in simple to assemble kits that allows them to store the furniture in the same warehouse-like locations where they are displayed and sold

Shipping container designs for FedEx and UPS – airfreight

Dunnage control in Big Auto

Efficient Design of the Supplier Base

Part of streamlining the supply chain is reducing the number and variety of suppliers

The Japanese have been very successful in this arena (they’re an Island – so getting materials there has always been a problem)

In the mid 1980’s Xerox trimmed its number of suppliers from 5,000 to 400.

Overseas suppliers were chosen based on cost Local suppliers were chosen based on delivery

speed In 1996, Ford Motor reduced their supplier

count by more than 60%

Dell Designs the Ultimate Supply Chain!

Dell Computer has been one of the most successful PC retailers. Why? To solve the problem of inventory becoming obsolete, Dell’s solution: Don’t keep any inventory! - All PC’s are

made to order and parts shipped directly from manufacturers when possible.

Compare to the experience of Compaq Corporation – initial success selling through low cost retail warehouses but they did not garner web-based sales

Data Exchange – A Critical Idea

EDI: Electronic Data Interchange Involves the Transmission of

documents electronically in a predetermined format from company to company. (Not web based.)

The formats are complex and expensive. It appears to be on the decline as web-based systems grow.

Data and Products – E-Tailing E-tailing: Direct to customer sales on the

web – the so-called Click & Mortor retail model Perhaps best known e-tailer is Amazon.com,

originally a web-based discount book seller Today, Amazon.com sells a wide range of

products (we can think of many, many similar organizations)

Amazon and others spawned so called “dot com” stock explosion in the NASDAQ (1997 to April, 2000)

Today, many traditional “bricks and mortar” retailers also offer sales over the web, often at lower prices

Dealing with Data – the modern way

B2B (business to business) supply chain management: While not as visible and “sexy” as E-tailing, it

appears that B2B supply chain management is the true growth industry!

Web searches yield over 80 matches for supply chain software providers. Some of the major players in this market segment include:

Agile Software based in Silicon Valley. i2 Technologies based in Dallas. Ariba based in Silicon Valley

Data Transfer in Supply Chains: Vendor Managed Inventory (the real solution?)

Walmart and P & G Target and Pepsi/Coke But … Barilla SpA. An Italian pasta producer

pioneered the use of VMI (Vendor Managed Inventory)

They obtained sales data directly from distributors and decide on delivery sizes based on that information

This is in opposition to allowing distributors (or even retailers) to independently decide on order sizes!

Order Growth – The Bullwhip Effect – An Important issue

Information Transfer in Supply Chains: cause of ‘The Bullwhip Effect’

First noticed by P&G executives examining the order patterns for Pampers disposable diapers.

They noticed that order variation increased dramatically as one moved from retailers to distributors to the factory.

The causes are not completely understood but have to do with batching of orders and building in safety stock at each level

Problem: increases the difficulty of planning at the factory level

There has been a Revitalization in the Analytical Tools needed to Support SCM

Inventory management and demand forecasting models such as those discussed in this course

The transportation problem and more general network formulations for describing flow of goods in a complex system

Analytical methods for determining delivery routes for product distribution – optimal location of new resources

Focusing on the Distribution Problem:

The Goal is to reduce total transportation costs throughout the supply chain

Usually solved with some approach to the “Transportation Problem”

Our approach will be the Balanced Matrix model

Lets do one, by example:

\ToFrom

Albuquerque

Boston Cleveland Capacity

Des Moines$5 $4 $3

100

Evansville$8 $4 $3

300

Ft. Lauderdale

$9 $7 $5300

Demand 300 200 200/700

700/

Cost of moving a unit of product from Row to column location

In the Transportation Problem:

We must have a Supply/Demand balance to solve In this problem that requirement is met

If it is not met, we must create “Dummy” sources (at $0 move costs) or Dummy Sinks (also at $0 move costs) to achieve the require S/D balance

The Transportation problem: Goal is to minimize the total cost of

shipping We will allocate products to cells – any

allocation means the row resource will ship product to the column demand

The process is an iterative one that requires a feasible starting point Can start by using NW Corner approach Can start using a more structured VAM (Vogel

Approximation method)

Starting with a VAM Solution technique: Determine Row Penalty number (PNi) –the difference

between lowest and 2nd lowest cost in row Here: R1 is 1; R2 is 1: R3 is 2

Determine Column Penalty Number (PNj) – the difference between the lowest and 2nd lowest col. Cost

Here: C1: 3; C2: 0; C3: 0 Choose R or C with greatest penalty cost – here is C1

If there is a tie, break tie by choosing C or R with smallest costs

Max out the allocation in chosen C or R at lowest cost cell then x-out the C or R

And so on after allocation (after we recompute PN’s!)

Phase 1 of VAM:

Step A B C PNr1 PNr2 PNr3

DM $5100

$4 $3 1 X-out -- --

E $8 $4200

$3100

1 1 5 x-out

FL $9200

$7 $5100

2 2 4

PNc1 3 0 0

PNc2 1 3 xout 2

1 -- 2

Costing The model: Current:

100*5 + 200*9 +200*4 + 100*3 + 100*5 = $3900

Before proceeding, check if the Feasible solution is (or isn’t) degenerate: Number of allocation must be at least: m +n -

1 = 3 + 3 - 1 = 5 (we have 5 is the above set so the solution is not degenerate! See next slide if it was)

Now, we must determine if it’s optimal? We must continue to a second phase to

determine this!

Dealing with Degeneracy (when it is found)

We must allocate a very small amount of material movement (call it ) to any independent cell

An independent cell is any one where we can not complete a “stepping motion” of only horizontal and vertical movements through filled cells to return to the originating cell

We call this the -path. (this would be done by alternating adding or subtracting assignments of material to any filled cell we step on)

Note any cell were a path can be build is a dependent cell We would add sufficient ’s to reach allocated cells

count of m + n – 1 number (make the solution non-degenerate)

Here since R1C1 is independent – check for yourself – we will fill it with units – this makes our solution non-degenerate – 5 cells are allocated!

Entering Phase II: Determining Optimality We will explore the MODI (modified

distribution) algorithm After finding a non-degenerate initial

solution, add a row of Kj’s and a column of Ri’s to the Matrix

To begin, Assign a zero value to any R or K position

For each allocated cell, the following expression must be satisfied: Ri + Kj + ci,j = 0

Starting MODI with a possible R/K allocation:

Kj

Ri

A B C

-5 -2 -1

DM0

$5 100

$4 +2

$3 +2 100

E-2

$8 +1

$4 200

$3100

300

FL-4

$9200

$7 +1

$5100

300

Demand:

300 200 200

Indicator cost for this cell (= R+K+c)

MODI continued: Examine all indicator values for empty cells – if

all are non-negative the solution is optimal If some are negative then develop a -path

beginning at most negative cell (here is R3C3) Complete the -path by stepping only to filled

cells (and pivoting) while alternatively subtracting then adding allocation

After completing the path, determine the “-” cell with the smallest quantity and choose its value for “” – substitute it along the whole path

Note here: all indicators are positive thus we have the optimal solution!

Forming the -Path (starts in R2C3)

Kj

Ri

A B C

-5 -4 -3

DM0

$5 +

$4 -100

$3Ind: 0

100

E0

$8Ind: +3

$4 + 100

$3 -200

300

FL-4

$9 -300

$7Ind: -1

$5 +Ind: -2

300

Demand:

300 200 200

Smallest “-” Allocated amount

– if we had erroneously allocated as seen below and requiring an addition

After 100 unit re-allocation – now recompute R’s and K’s & Indicators

Kj

Ri

A B C

Cap.

-4 -1 0

DM-1

$5 +100

$4 +2

$3 +2

100

E-3

$8 +1

$4200

$3100

300

FL-5

$9200

$7 +1

$5100

300

Demand:

300 200 200

Looking at this Matrix All indicators are now positive – this

indicates an optimal solution! Note this agrees with optimal solution found

earlier!!! Relax value to zero – makes cell 1,1

allocation 100 units Optimal transportation cost is:

5*100 + 4*200 + 3*100 + 9*200 + 5*100 = 3900

Lets try one:

/ToFr/

D E F G Cap.

A8 6 4 2

4

B10 6 6 2

3

C4 2 3 8

6

Demand 3 3 3 4/13

13/

But the Transportation Problem can be solved by LP! Define an

Objective Function:

Subject to:

1 1

:

is cell cost and is amount moved

m n

ij iji j

ij ij

c X

where

c X

1

1

i

j

for 1 i m

for 1 j n

where:

a are all shipment from a source (capacity)

b are all shipments into a "sink" (Demand)

n

ij ij

m

ij ji

X a

X b

Our Example (by LP Solver)

Variables XDM-A XDM-B XDM-C XE-A XE-B XE-C XFL-A XFL-B XFL-C

Values: 0 0 0 0 0 0 0 0 0

V*C 0 0 0 0 0 0 0 0 0 OBJ.Fn.

Costs: 5 4 3 8 4 3 9 7 5 0

CapC1 1 1 1 0 100

CapC2 1 1 1 0 300

CapC3 1 1 1 0 300

DemC1 1 1 1 0 300

DemC2 1 1 1 0 200

DemC3 1 1 1 0 200

Applying Solver:

Examining Results:

Optimal Value = $3900 (as we found ‘by hand’!)

Ship: 100 DM-A; 200 FL-A; 200 E-B; 100 E-C; 100 FL-C

All as we found using the VAM Heuristic

Much Faster and easier using LP!

Expansion to Transshipment Problem

When a system is allowed to use intermediate ‘warehousing’ sites – they are Source sites or even Sink sites in regular transportation problem – for reducing the total cost of transportation we call the problem the transshipment network problem

We require more costs to be obtained but typically, in most complex S. Chains, companies find savings of from 7 to 15% (or more) in implementations that allow transshipment

Expansion to Transshipment Problem

In the general Transshipment problem the transport network is expanded to allow movement between

sources and between sinks (and even back to other sources)

Expansion to Transshipment Problem

Extracted from: J. P. Ignizio, Linear Programming in Single- & Multiple-Objective Systems, Prentice Hall 1982The original

transportation problem

Another Level of Transport – the delivery route problem

This problem is usually one of very large scale (classically called the Traveling Salesman Problem)

Because of this, we typically can not find an absolutely optimal solution but rather only near optimal solution – as seen in our textbook

Here, the knowns are the costs of travel from point to point throughout the network and we try to save costs by “ganging up” trips

Solution typically follows along a line of attack based on the “Assignment Problem”

See Handout, focus on the Shortest Route Problem

Delivery Optimization: Realistically, a delivery vehicle

can only carry so much – so this may reduce effectiveness of solutions

Delivery’s take Real Time – again this must be considered during scheduling and routing

Loading of vehicles is very critical to control step 2 time – load in reverse delivery order!

Looking at the Locating of New Facilities:

Considerations: Labor Climate Transportation issues:

Proximity to markets Proximity to suppliers & resources Proximity to parent company (sharing

expertise, purchasing, drop routing) – could be plus or minus!

Quality of transportation system

Looking at the Locating of New Facilities:

Consideration, cont. Costs to operate (utilities, taxes, real

estate costs, construction) Expansion considerations

Room available for growth? Construction to modify structure?

Any local incentives to re-locate? Quality of Life (schools, recreational

possibilities, health care – cost, availability)

Looking at the Locating of New Facilities:

Most organization compare several alternatives They identify weighting factors for the

characteristics then narrow choices 1st consider regions 2nd narrow search to communities 3rd consider specific sites Done by collecting data addressing the various factors

under study After data is collected and weighted, make

selection (typically by starting with quantitative decision follow with qualitative analysis)

Location Decisions in SCM: In the final analysis the decision

typically comes down to a “Center of Gravity Solution” that minimizes the total travel distance between the Facility and all possible Contact facilities Contact facilities may be sources of Raw

Materials or other Suppliers or they may be destinations for Product stored in or made at the new facility under design

Lets Consider an Example:

DC 1

Store C

Store E

Store G

Store FDC 2

Store D

Store A

Store B

(where should we put our new Distribution Center?)

Given this information:

StoreLocation (X,Y)

C. Sales

A (2.5, 2.5) 5

B (2.5, 4.5) 2

Ca (5.5, 4.5) 10

D (5, 2) 7

E (8, 5) 10

Fb (7, 2) 20

G (9, 3.5) 14a Site of Possible DC 1; b Site of Possible DC 2

Solution is a Type of Transportation Minimization:Using Either Euclidian or Recta-linear offsets

Euclidean Distance:

Recta-Linear (RL) Distance:

2 2

: X's or Y's are map coordinates

of stores or Distribution Center

j jeuclid i DC i DCd X X Y Y

here

j jRL i DC i DCd X X Y Y

Now What? Best Location is the one that minimizes

the sum of the total needs of all Demands times the travel distances involved:

i

for all Sinks

and each possible new Source

ji euclidall

D d

Leading to this analysis:

DC 1 5.5 4.5

DC 2 7 2

STORE X Location Y Location D Euclid 1 D Euclid 2 D RL 1 D RL 2 Demand D*DE 1 D*DE 2 D*Drl 1 D*Drl 2

A 2.5 2.5 3.605551 4.527693 5 5 5 18.02776 22.63846 25 25

B 2.5 4.5 3 5.147815 3 7 2 6 10.29563 6 14

C 5.5 4.5 0 2.915476 0 4 10 0 29.15476 0 40

D 5 2 2.54951 2 3 2 7 17.84657 14 21 14

E 8 5 2.54951 3.162278 3 4 10 25.4951 31.62278 30 40

F 7 2 2.915476 0 4 0 20 58.30952 0 80 0

G 9 3.5 3.640055 2.5 4.5 3.5 14 50.96077 35 63 49

176.639 142.711 225 182

From the Analysis: DC 2 minimizes costs

But is this the Optimal location?

Perhaps we could place the ‘Center’ at the Median of all the current demand locations?

In an ‘RL’ sense, form the Cumulative Weighting (Cum. Demand)

1st: order each target location in increasing level of X and then Y

Determine the average of this CumWt for X & Y and place location the is the same as the site that first exceeds this value

C. Wt. Average is 34 Store

Dem.

X coor

C. Wt.

Sel. Store

Dem.

Y coor

C. Wt.

Sel

A 5 2.5 5 D 7 2 7

B 2 2.5 7 F 20 2 27

D 7 5 14 A 5 2.5 32 *

C 10 5.5 24 * G 14 3.5 46 **

F 20 7 44 ** B 2 4.5 48

E 10 8 54 C 10 4.5 58

G 14 9 68 E 10 5 68

Locate at: 7 (in X) and 3.5 (in Y) as a rule*

Optimization using Euclidian Distances:

1st Compute the Center of Gravity:

for each location i

i i

i

i i

i

D xX

D

D yY

D

Optimization: Start with X*, Y* determined above as Xcur,

Ycur

Compute:

Stop iteration when X and Y stop changing

2 2,

:

,

,

,

,

ii

cur i cur i

i inew

i

i inew

i

Dg x y

x x y y

then

x g x yx

g x y

y g x yy

g x y

Trends in Supply Chain Management

Outsourcing of the logistics function (example: Saturn outsourced their logistics to Ryder Trucks. Outsourcing of manufacturing is a major trend these days)

Moving towards more web based transactions systems

Improving the information flows along the entire chain

Global Concerns in SCM

Moving manufacturing offshore to save direct costs complicates and adds expense to supply chain operations, due to: increased inventory in the pipeline Infrastructure problems Political problems Dealing with fluctuating exchange rates Obtaining skilled labor