Supply Chain Management

Chapter 2
Supply Network Design
Prof. Dr. Guido Voigt

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 0

Learning objectives

• Understand the role of network design in supply chains

• Identify factors influencing supply chain network design decisions
• Develop a framework for making network design decisions
• Use optimization for facility location and capacity allocation decisions

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 1

Agenda

2.1. Factors influencing network design decisions

2.2. Framework and OR tools
2.2.1. Phase 1: Hotelling
2.2.2. Phase 2: Mixed-integer programming
2.2.3. Phase 3: Gravity models
2.2.4. Phase 4: Factor rating method and mixed-integer programming

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 2

2.1. Factors influencing network design decisions
The role of network design

Classification
• Facility role:
– What role/Which processes should be assigned to each facility?
• Facility location:
– Where should the facility be located?
• Capacity allocation:
– How much capacity should be allocated to each facility?
• Market and supply allocation:
– Which supply sources should feed each facility and what markets should be served by
the facility?

 Strategic decisions with big influence on performance, flexibility and

responsiveness of the supply chain
 All network decisions affect each other (interdependency) resulting in trade-offs

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 3

2.1. Factors influencing network design decisions
Strategic and Technological Factors

Strategic Factors [Link]

• Strategies: Cost leadership vs. Responsiveness [Link]
– Discounters as cost leaders with large but few stores
– Convenience store chains focusing on responsiveness with small but many stores
• Balancing the aims: example Zara with facilities in Asia and Europe
– Asia: low cost, standardized and low-value products in large amounts
– Europe: responsive, innovative products with unpredictable demand

Technological Factors [Link]

• Production technology determines ratio of facility fix costs and transportation

costs
• High fix costs enabling high economies of scale (and vice versa)
– Relatively low transportation costs
– Build few high-capacity facilities

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 4

2.1. Factors influencing network design decisions
Macroeconomic Factors

Macroeconomic Factors
• Not depending on the individual company
– economic performance
– legal restrictions
– political circumstances
[Link]

a) Tariffs, tax incentives, subsidies

 Tariffs: customs to be paid if international borders are crossed
 High tariffs by a country: either no service in this country or plants in the country itself
 Tariffs used to decrease due to general tendency to free trade
 this changed in the recent past, e.g. “US-China-Trade War”(NBER, 2022)
 Tax incentives/subsidies: incentives on federal, state or communal level
aiming to encourage firms to build plants in the respective region
 Example: semi-conductor industries (Intel) or Green Investment in Inflation
Reduction Act (US)
 Indirect: limits on imports or local content requirements

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 5

2.1. Factors influencing network design decisions
Macroeconomic Factors

b) Exchange rate, costs, and demand risk

b.1) Exchange risk:
 Production costs assessed in currency α, sales profits in currency β
 If α increases relatively to β in value, costs rise while profits fall
 Exchange rate risk can be hedged (aside from financial hedge instruments) by
(some) overcapacity in every country enabling the firm to react on exchange rate
changings, thus even profit from them
 See also chapter on Sales and Operation Planning (S&OP)

b.2) Demand risk:

 cyclical downturns and progressions in countries, e.g. presence in growing
markets and high flexibility to divert resources

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 6

2.1. Factors influencing network design decisions
Macroeconomic and Politcial Factors

c) Freight and fuel costs

 Fluctuations in freight and fuel costs difficult to hedge on a SC basis
 Better to use financial instruments like options on commodity supply or long-
term agreements

Political Factors
• Political stability affects predictability of commercial development and ensures
ownership rights
• Especially important for developing countries with emerging markets
• Measured by the global political risk index (GPRI) considering the ability of a
market to withstand shocks in government, society, security and economy

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 7

2.1. Factors influencing network design decisions
Infrastructure and Competitive Factors

Infrastructure Factors
• Low labor and land costs can be equalized by bad infrastructure and vice versa
• Infrastructure comprising availability of sites and labor, proximity to highways,
distribution centers, train stations, airports and seaports

Competitive Factors (Consider competitors’ strategy, size and location)

a) Positive externalities between firms

 Presence of multiple competitors leads to improving infrastructure
(technology parks) or higher demand (shopping malls)

b) Locating to split the market (Hotelling model, see 2.2.1)

 If proximity to customers is the main factor, firms aim for the largest market
share
 Locate close to each other to split the market  Hotelling's Location Model

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 8

2.1. Factors influencing network design decisions
Competitive Factors

c) Customer response time and local presence

 Locate close to the customers, many small sites rather than wide spread large
ones
 Example: convenience stores vs. supermarkets vs. discounters

 Large sites plus rapid means of transportation

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 9

2.1. Factors influencing network design decisions
Competitive Factors

d) Logistics and Facility Costs

 Depending on number of facilities, their location and capacity allocation
 Logistics costs comprise inventory and transportation costs
 Increasing number of facilities
 leads to decreasing transportation costs but increasing inventory (in particular:
safety stock) and facility costs, see chapter 4
 Depending on the transformation done in a facility, the distance to the supplier is
to be decided:
 if a large volume and/or weight of commodities is transformed into a much
smaller and/or lighter product, the facility should be close to the supplier

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 10

Agenda

2.1. Factors influencing network design decisions

2.2. Framework and OR tools
2.2.1. Phase 1: Hotelling
2.2.2. Phase 2: Mixed-integer programming
2.2.3. Phase 3: Gravity models
2.2.4. Phase 4: Factor rating method and mixed-integer programming

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 11

2.2. Framework and OR tools
Framework for network design decisions
- Competitive Strategy - Global Competition
Phase I - Internal constraints
(capital, growth strategy, existing network)
Supply Chain
Strategy

- Production Technologies - Tariffs and tax incentives

(cost, scale/scope impact, support required, flexibility) - Regional demand
Phase II
- Competitive environment (size, growth, homogeneity, local specifications)
Regional Facility
- Aggregate Factor and Logistics costs - Political, exchange rate, demand risk
Configuration

- Production methods - Available infrastructure

(skill needs, response time)
Phase III
Desirable Sites

- Factor costs - Logistics costs

(labor, materials, site specific) (transport, inventory, coordination)
Phase IV - Production Technologies - Tariffs, tax incentives, regional demand,
Location Choices political, exchange rate, demand risk

Goal: Maximizing profits while satisfying customer needs in

terms of demand and responsiveness
SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 12
2.2. Framework and OR tools
Phase I – Supply Chain Strategy

Supply Chain Design

• Stages in SC
• Functions outsourced or in-house

Procedure
• Defining firm’s competitive strategy
• Matching capabilities with strategy
• Forecasting global competition and identify local or global market for relevant
competition
• Identifying financial constraints
• Determining growth strategy to use ec. of scope and scale
-> Acquiring existing facilities, building new ones or finding partners

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 13

2.2. Framework and OR tools
Phase II – Regional Facility Configuration

Procedure
• Forecasting demand by region (e.g. country)
– Demand per region, correlation of customer requirements across regions
-> Few large consolidated facilities if homogenous requirements
-> Smaller, localized facilities if heterogeneous requirements
• Potential of ec. of scope/scale for given production technology
-> significant ec. of scale/scope: few large consolidated facilities
-> Ec. of scale/scope insignificant: smaller, localized facilities
• Uncertainty: demand, exchange-rate and political risk by region
• Trade restrictions: tariffs, local production requirements, tax incentives, export or
import restrictions
• Competitors: decide to locate own facilities close to or far from competitor
• Determining target response time and logistics costs per region
-> Locate facilities and determine their potential roles
-> Supply several regions, number of products produced, capacity

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 14

2.2. Framework and OR tools
Phase III – Set of potential sites

Procedure
• Select a set of desirable potential sites based on an infrastructure analysis
– Hard infrastructure requirements:
• Suppliers
• Transportation services
• Communication
• Utilities
• Warehousing facilities
– Soft infrastructure requirements:
• Skilled workforce
• Workforce turnover (fluctuation)
• Community receptivity to business and industry

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 15

2.2. Framework and OR tools
Phase IV – Location Choices

Procedure
• Select a precise location (realty)
• Determine the capacity allocation for each selected facility

 Every single location is now chosen based on

• Phase I: its contribution to competitive strategy
• Phase II: considering the expected margin and demand per region, taxes,
tariffs and other trading restrictions at each region
• Phase III: various logistics and facility requirements.

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 16

Agenda

2.1. Factors influencing network design decisions

2.2. Framework and OR tools
2.2.1. Phase 1: Hotelling
2.2.2. Phase 2: Mixed-integer programming
2.2.3. Phase 3: Gravity models
2.2.4. Phase 4: Factor rating method and mixed-integer programming

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 17

2.2. Framework and OR tools
Modelling facility location and capacity allocation

Goals of modelling SC networks

• Maximize profit and satisfy customer needs (esp. response time)
• Long-run (1 to 5 years): Decide at which location facilities will be established and
assign capacity to each of those
• Mid-term (quarter to a year): Assign demand (e.g. annual or monthly basis) to the
available facilities and determine along which relations a product will be
transported considering changes in tariffs, exchange rates, prices and demand

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 18

2.2. Framework and OR tools
General methods overview

Qualitative Methods Quantitative Methods

Intuitive Approach Exclusive consideration of quantitative

(e.g. Checklist Method): (monetary) factors:
• Assessment based on checklists, e.g. • Broad variety of potential methods
site factor catalogue • Possibility and requirement to use
Operations Research methods

Formal Approach
(e.g. cost-utility analysis):
• Analytical assessment methods

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 19

2.2. Framework and OR tools
General methods overview: Quantitative Methods

Basic properties:
• Solely quantitative (monetary) factors considered
-> Vast variety of decisions possible
-> Decisions interact (facility location and transportation requirements)
-> OR methods/Quantitative Decision Support

• Alternative modeling approaches:

Location planning in networks Location planning in the field

( discrete location planning) ( continuous location planning)
• Finite number of possible sites • Infinite number of possible sites
• Heterogeneous geographic plane • Homogenous geographic plane
• Distances determined individually • Distances determined based on
• High effort planning methods generic calculations

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 20

2.2.1. Phase 1: Hotelling
Supply Chain Strategy, Hotelling's Location Model

Assumptions:
• Customers uniformly located along a line from 0 to 1
• Two firms compete only based on distance to costumers
• Customers go to the closer firm
• Customers equidistant to both are evenly split
• Total demand is 1
• Firm A locates at , Firm B at
• Example for competition on distance: Demand at A and B is respectively

𝟏 𝒃 𝒂 𝟏 𝒃 𝒂
𝒂 𝟐 𝒃 𝟐
𝟏 𝒃 𝒂
𝟐
a 1-b

0 1
𝟏−𝐛−𝒂
Watch:
𝟐 [Link]

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 21

2.2.1. Phase 1: Hotelling
Supply Chain Strategy, Hotelling's Location Model
Optimal location for fixed end-customer prices:
• Starting from two arbitrary points the solution is arrived iteratively
• Both parties are not able to unilaterally increase profits if
– In this state there is no incentive for either firm to move its location
Optimal location with price setting power:
• Assume a firm can set higher end-customer prices if the competitor is located
further away (e.g., customers are willing to accept higher prices in order to travel
less)
• Assume the contribution margin per unit is
• In this case, the optimal solution is
• Firms can maximize revenues by increasing prices. Note that the total demand per
firm stays constant.

Insight: Optimal location strategies critically depends on the firms ability to set higher
prices in the absence of firms located nearby
For exhaustive coverage see
Shy, O. (1997) Industrial Organization – Theory and Application, [Link], MIT Press
SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 22
Agenda

2.1. Factors influencing network design decisions

2.2. Framework and OR tools
2.2.1. Phase 1: Hotelling
2.2.2. Phase 2: Mixed-integer programming
2.2.3. Phase 3: Gravity models
2.2.4. Phase 4: Factor rating method and mixed-integer programming

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 23

2.2.2. Phase 2: Mixed-integer programming
Network Optimization Models

Example: SunOil
• Manufacturer of petrochemical products, worldwide sales
• Regions: Asia, Africa, Europe, South America, North America
• Options:
1) One facility per region
• Low transportation costs, avoids duties for imports
2) Few plants in some regions
• Higher transportation costs, duties for imports to regions without plants
• Exploits economies of scale far more

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 24

 2.2.2. Phase 2: Mixed-integer programming
Network Optimization Models
SunOil - Data
Inputs – Costs, Capacities, Demands

Demand Region
Fixed Cost &
Product, Inventory and Transportation Cost per
Capacity Plant
1000 Units
Supply Fixed
NA SA EU AS AF Capacity
Region Cost

NA 81 92 101 130 115 6000

SA 117 77 108 98 100 4500

EU 102 105 95 119 111 6500

AS 115 125 90 59 74 4100

AF 142 100 103 105 71 4000

Demand 12 8 14 16 7

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 25

2.2.2. Phase 2: Mixed-integer programming
Network Optimization Models

AS

NA AF
EU

SA

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 26

2.2.2. Phase 2: Mixed-integer programming
Simple Plant Location Model (PLM)
PLM (aka Simple Warehouse Location Problem)
Assumptions:
– Linear variable costs (no ec. of scale through variable)
– All demand must be satisfied
– No taxes on earnings
– Aim: minimizing costs
Input:
– = number of potential plant locations
– = number of markets or demand points
– = transportation cost per unit from location i to market j
– = annual demand from market
– = annualized fixed cost of keeping plant open
– = “Big M”: as small as possible, but sufficiently large, e.g.
Variables:
– = 1 if plant is open, 0 otherwise
– = quantity shipped from plant to market
SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 27
2.2.2. Phase 2: Mixed-integer programming
Simple Plant Location Model

for

for

for

for

“Big M” is ideally as small as possible but sufficiently large, e.g.,

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 28

2.2.2. Phase 2: Mixed-integer programming
GAMS - Code

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 29

2.2.2. Phase 2: Mixed-integer programming
GAMS - Code

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 30

2.2.2. Phase 2: Mixed-integer programming
GAMS - Code

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 31

2.2.2. Phase 2: Mixed-integer programming
GAMS - Solution

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 32

2.2.2. Phase 2: Mixed-integer programming
GAMS - Solution

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 33

2.2.2. Phase 2: Mixed-integer programming
GAMS - Solution

AS ,

,
,

, 14
NA AF
EU

,
SA

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 34

2.2.2. Phase 2: Mixed-integer programming
Network Optimization Models – Capacity Constraints

Example contd.: SunOil with Capacity Constraints

• Region is not necessarily supplied by nearest facility (if capacity is short)

• Additional Input:
– = capacity of potential plant at location

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 35

 2.2.2. Phase 2: Mixed-integer programming
Network Optimization Models – Capacity Constraints

Inputs – Costs, Capacities, Demands

Demand Region
Fixed Cost &
Product, Inventory and Transportation Cost per
Capacity Plant
1000 Units
Supply Fixed
NA SA EU AS AF Capacity
Region Cost

NA 81 92 101 130 115 6000

SA 117 77 108 98 100 4500

EU 102 105 95 119 111 6500 1

AS 115 125 90 59 74 4100

AF 142 100 103 105 71 4000

Demand 12 8 14 16 7

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 36

2.2.2. Phase 2: Mixed-integer programming
Capacitated Plant Location Model

for

for

for

for

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 37

2.2.2. Phase 2: Mixed-integer programming
GAMS - Code: Capacitated

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 38

2.2.2. Phase 2: Mixed-integer programming
GAMS - Code: Capacitated

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 39

2.2.2. Phase 2: Mixed-integer programming
GAMS - Code: Capacitated

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 40

2.2.2. Phase 2: Mixed-integer programming
GAMS - Solution: Capacitated

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 41

2.2.2. Phase 2: Mixed-integer programming
GAMS - Solution: Capacitated

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 42

2.2.2. Phase 2: Mixed-integer programming
GAMS – Solution: Capacitated

AS , 5

NA ,
, AF
EU

, , 3
,

SA

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 43

2.2.2. Phase 2: Mixed-integer programming
Network Optimization Models – Capacity Levels

Example contd.: SunOil with Capacity Constraints plus option to build up

different capacity levels

Additional Input:
– = capacity of potential low capacity plant
– = capacity of potential high capacity plant
– = annualized fixed cost of keeping low capacity plant open
– = annualized fixed cost of keeping high capacity plant open

Variables:
– = 1 if low capacity plant is open, 0 otherwise
– = 1 if high capacity plant is open, 0 otherwise

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 44

 2.2.2. Phase 2: Mixed-integer programming
Network Optimization Models – Capacity Levels

Inputs – Costs, Capacities, Demands

Demand Region
Fixed Cost &
Product, Inventory and Transportation Cost per High Capacity Plant
Capacity Plant
1000 Units
Supply Fixed Low Fixed High
NA SA EU AS AF
Region Cost Capacity Cost Capacity

NA 81 92 101 130 115 6000 9000 20

SA 117 77 108 98 100 4500 6750 20

EU 102 105 95 119 111 6500 1 9750 20

AS 115 125 90 59 74 4100 6150 20

AF 142 100 103 105 71 4000 6000 20

Demand 12 8 14 16 7

Note: better fixed costs/capacity ratio for large plants

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 45
2.2.2. Phase 2: Mixed-integer programming
Network Optimization Models – Capacity Levels

for

for

for

for

for

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 46

2.2.2. Phase 2: Mixed-integer programming
Gams-Code: Capacity Levels

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 47

2.2.2. Phase 2: Mixed-integer programming
Gams-Code: Capacity Levels

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 48

2.2.2. Phase 2: Mixed-integer programming
Gams-Code: Capacity Levels

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 49

2.2.2. Phase 2: Mixed-integer programming
Gams-Solution: Capacity Levels

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 50

2.2.2. Phase 2: Mixed-integer programming
Gams-Solution: Capacity Levels

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 51

2.2.2. Phase 2: Mixed-integer programming
Gams-Solution: Capacity Levels

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 52

2.2.2. Phase 2: Mixed-integer programming
Gams-Solution: Capacity Levels

AS ,

,
NA AF
EU

,
,

SA

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 53

Agenda

2.1. Factors influencing network design decisions

2.2. Framework and OR tools
2.2.1. Phase 1: Hotelling
2.2.2. Phase 2: Mixed-integer programming
2.2.3. Phase 3: Gravity models
2.2.4. Phase 4: Factor rating method and mixed-integer programming

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 54

2.2.3. Phase 3: Gravity models
Center of Gravity Method (Steiner-Weber-Modell)

Specification
• One period with given duration
• One object to localize
• Infinite number of homogenous potential sites (planning in the field)
• customers with given coordinates
• Given transport volume between object of planning and customers (in quantity
units)
• Proportional transport costs (regarding volume and distance) with rate (in
currency units per quantity units and distance units)
• No site specific fixed costs
• Planning objective: minimizing total transport costs

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 55

2.2.3. Phase 3: Gravity models
Center of Gravity Method (Steiner-Weber-Modell)

Modelling
Decision variable: coordinates of planning object

Objective function:

with =distance between planning object and customer

Equivalence:

: transport costs
: overall transport provision (= distance *quantity)

Equivalent objective function: with

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 56

2.2.3. Phase 3: Gravity models
Center of Gravity Method (Steiner-Weber-Modell)

Modelling with Euclidean distance metric

• Objective:

• Problem properties:
– No decomposability (i.e., cannot be solved for and separately)
– Convexity of in and

• Criteria of optimality (first order condition):

 necessary and sufficient condition

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 57

2.2.3. Phase 3: Gravity models
Center of Gravity Method (Steiner-Weber-Modell)

Modelling with Euclidean distance metric

• Rearrange first order condition :

 Consequence: entire isolation of not possible  non-linear problem

 Linear optimization tools not suitable  iterative algorithm by Miehle (1958)

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 58

2.2.3. Phase 3: Gravity models
Center of Gravity Method (Steiner-Weber-Modell)

Iterative Algorithm by Miehle

• Initiation: center of gravity

∑ ∑
∑
and ∑

• Iteration steps:
– Substitute last solution ) as right side of and
– Identify new solution ) by solving and for

• Stop criterion: Stop if is below a pre-determined bound:

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 59

2.2.3. Phase 3: Gravity models
Center of Gravity Method (Steiner-Weber-Modell)

Example:
• Locating one warehouse using Euclidean metric
• Six customer sites with demand :
1 2 3 4 5 6

•
•
• Center of gravity:

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 60

2.2.3. Phase 3: Gravity models
Center of Gravity Method (Steiner-Weber-Modell)

Example contd.
• Using this starting point, the algorithm leads to the following iteration steps:
Iteration
0 2.300 2.433 69.9684
1 2.228 2.560 69.6403
2 2.167 2.640 69.4716
3 2.125 2.697 69.3819
… … … …
21 2.031 2.881 69.2626
22 2.031 2.882 69.2625

•
•
 Stop! Solution:

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 61

Agenda

2.1. Factors influencing network design decisions

2.2. Framework and OR tools
2.2.1. Phase 1: Hotelling
2.2.2. Phase 2: Mixed-integer programming
2.2.3. Phase 3: Gravity models
2.2.4. Phase 4: Factor rating method and mixed-integer programming

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 62

2.2.4. Phase 4: Factor rating method and MIP
Location choice

Evaluating location alternatives: The factor-rating method

• Unlike former methods, this one uses the many qualitative and quantitative factors
which influence the locating decision.
• To take different levels of importance objectively into account, the factors are
weighted
• The advantages of this method are therefore the wide variety of factors (see
chapter 2.1), which can be objectively included

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 63

2.2.4. Phase 4: Factor rating method and MIP
Factor-Rating method

Procedure
1. Develop a list of relevant factors called key success factors
2. Assign a weight to each factor to reflect its relative importance in the company’s
objectives
3. Develop a scale for each factor (e.g. 1 to 10 or 1 to 100 relative points)
4. Have management score each location for each factor, using the scale from step 3
5. Multiply the score by the weights for each factor and total the score for each
location
6. Make a recommendation based on the maximum point score, considering the
results of other quantitative approaches as well

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 64

2.2.4. Phase 4: Factor rating method and MIP
Factor-Rating method

Example Five Flags

• Five Flags, a U.S. chain of 10 family-oriented theme parks, has decided to expand
overseas by opening its first park in Europe. It wishes to select between Dijon,
France and Copenhagen, Denmark
• Find key success factors, their weights (adding up to 1 by design) and their rating
in the following table:
Scores (out of 100)
Key success factors Weight France Denmark
Labor availability and 70 60
attitude .25
People-to-car ratio .05 50 60
Per capita income .10 85 80
Tax structure .39 75 70
Education and health .21 60 70
Total 1.00

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 65

2.2.4. Phase 4: Factor rating method and MIP
Factor-Rating method

Example Five Flags contd.

Scores Weighted
(out of 100) Scores
Key success factors Weight France Denmark France Denmark
Labor availability and 70 60 .25*70 .25*60
attitude .25 =17.5 =15.0
People-to-car ratio .05 50 60 2.5 3.0
Per capita income .10 85 80 8.5 8.0
Tax structure .39 75 70 29.3 27.3
Education and health .21 60 70 12.6 14.7
Total 1.00 70.4 68.0

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 66

2.2.4. Phase 4: Factor rating method and MIP
Factor-Rating method

Example Five Flags contd.

• Given 100 points reachable, the French location is preferable due to this method
and this weights
• Sensitivity analysis.
– changing the score for labor availability and attitude by 10 shifts the decision
to the Denmark location.
– Given the inevitable subjectivity of the weights and scores, this method,
although quantitative, can not be seen as an ‘exact’ approach
• Exercise: Dropping the weight for tax structure to .20 and increasing the weight for
education and health to .40, does the decision change?
• Answer: Denmark is now chosen, with a 68.0 vs. a 67.5 score for France

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 67

2.2.4. Phase 4: Factor rating method and MIP
Network Optimization Models

Example: TelecomOne & HighOptic

• Two manufacturers of fiber-optic telecommunication equipment
• TelecomOne focused on eastern USA
– Plants in Baltimore, Memphis, Wichita
– Markets in Atlanta, Boston, Chicago
• HighOptic focused on western USA
– Plants in Cheyenne, Salt Lake City
– Markets in Denver, Omaha, Portland
Merging both companies: TelecomOptic
• Merged company has 5 plants for 6 markets
• Management wants to shut down some plants to raise profits
• Assumptions:
– Taxes and duties do not vary among locations
– Just fixed and variable costs as relevant parameters
• Locating factories and allocating demand to those factories remaining open
• Using the CPLM from Phase II (with more detailed data regarding customers and
time)
SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 68
 2.2.4. Phase 4: Factor rating method and MIP
Network Optimization Models: Example - Data
Demand City
Production and Transportation Costs per 1000 Units

Monthly
Supply Monthly
Atlanta Boston Chicago Denver Omaha Portland Fixed
City Capacity
Cost

Baltimore 1675 400 685 1630 1160 2800 18 7650

Cheyenne 1460 1940 970 100 495 1200 24 3500

Salt Lake
1925 2400 1425 500 950 800 27 5000
City

Memphis 308 1355 543 1045 665 2321 22 4100

Wichita 922 1646 700 508 311 1797 31 2200

Monthly
demand 10 8 14 6 7 11
Dj

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 69

2.2.4. Phase 4: Factor rating method and MIP
Gams-Code: Example – Separate Operations

1.) Determine the cost for each company (without merger)

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 70

2.2.4. Phase 4: Factor rating method and MIP
Gams-Code: Example – Separate Operations

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 71

2.2.4. Phase 4: Factor rating method and MIP
Gams-Solution: Example – Separate Operations

Caution! These are just the variable costs. Adding the monthly costs of the sites
we get total cost of:
 Identical calculation can be done for TelecomOne
 Total cost TelecomOne: $28.836.009
 Total cost both companies: $50.201.009 (about 50 million USD)
SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 72
2.2.4. Phase 4: Factor rating method and MIP
Gams-Code: Example – Merged Operations

2.) Optimal decisions for the merged company

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 73

2.2.4. Phase 4: Factor rating method and MIP
Gams-Code: Example – Merged Operations

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 74

2.2.4. Phase 4: Factor rating method and MIP
Gams-Code: Example – Merged Operations

!!!!!!!

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 75

2.2.4. Phase 4: Factor rating method and MIP
Gams-Solution: Example – Merged Operations

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 76

2.2.4. Phase 4: Factor rating method and MIP
Gams-Solution: Example – Merged Operations

• Total costs of $47.401.000

• Total cost unmerged operations: $50.201.009
-> Savings of about $2.8 million per month compared to separate firm operation

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 77

2.2.4. Phase 4: Factor rating method and MIP
TelecomOptic – Single Sourcing

CPLM with Single Sourcing

• Each market is supplied from only one factory, a so called single source
• Advantages:
– Reduced complexity
– Less flexibility from each facility required
• Modifications of the original CPLM:
– Variables:
• 𝑖
• 𝐢𝐣
– Model:

s.t. for j = 1,…,m

for i = 1,…,n

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 78

2.2.4. Phase 4: Factor rating method and MIP
Gams Code – Single Sourcing

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 79

2.2.4. Phase 4: Factor rating method and MIP
Gams Code – Single Sourcing

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 80

2.2.4. Phase 4: Factor rating method and MIP
Gams Code – Single Sourcing

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 81

2.2.4. Phase 4: Factor rating method and MIP
Gams Solution– Single Sourcing

$49.717.000 is about $2.300.000

higher than without single sourcing
constraint

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 82

2.2.4. Phase 4: Factor rating method and MIP
Multiple Supply Chain Stages

Locating Plants and Warehouses simultaneously

• Designing the entire SC from the supplier to the customer
• Suppliers send material to factories which supply warehouses which in
turn supply markets
• Location and allocation decisions have to be made for factories as well as
warehouses
• It is assumed that one product needs one unit of input along the whole SC
 the units are appropriately adjusted

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 83

2.2.4. Phase 4: Factor rating method and MIP
Multiple Supply Chain Stages

Locating Plants and Warehouses simultaneously (contd.)

𝐾 𝑊
𝑓 𝑓
𝑆 𝐷
𝑐̂ 𝑐 𝑐̅
ℎ=1 𝑖=1 𝑒=1
𝑐 𝑐̅ 𝑗=1
𝑐̂ 𝑊
𝐾 𝑓 𝑐̅ 𝐷
𝑆 𝑐̂ 𝑓
𝑖=2 𝑒=2
ℎ=2 𝑐̅ 𝑗=2
𝐾
𝐷
𝑖=3 𝑓
𝑆 𝑗=3
ℎ=3
𝐷

𝑗=4

Supplier: Factories: Warehouses: Markets:

n

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 84

2.2.4. Phase 4: Factor rating method and MIP
Multiple Supply Chain Stages

Specification
•
•
•
•
• 𝑗
• 𝑖
• ℎ
• 𝑒
• 𝑖
• 𝑒
• ℎ𝑖
• 𝑖𝑒
• 𝑒𝑗

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 85

2.2.4. Phase 4: Factor rating method and MIP
Multiple Supply Chain Stages

Variables
• 𝑖
• 𝑒
• ℎ𝑖
• 𝑖𝑒
• 𝑒𝑗

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 86

 2.2.4. Phase 4: Factor rating method and MIP
Multiple Supply Chain Stages

ℎ𝑖 ℎ𝑖 𝑒𝑗 𝑒𝑗

s.t.
𝑥 ≤𝑆 for

𝑥 − 𝑥 ≥0 for

𝑥 ≤𝐾𝑦 for

𝑥 − 𝑥̅ ≥0 for

𝑥̅ ≤𝑊𝑦 for

𝑥̅ =𝐷 for

, ,
SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 87
2.2.4. Phase 4: Factor rating method and MIP
Multiple Supply Chain Stages

Locating Plants and Warehouses simultaneously (contd.)

example: surfboard manufacturer

𝐾 =40 𝑊 =60
𝑆 =30 𝑓 =300 𝑓 =630
𝐷 =10
ℎ=1 𝑐̂ =324 𝑖=1 𝑐 =197 𝑒=1 𝑐̅ =495 𝑗=1
Bremen Kassel Nürnberg Dort-
𝑐̂ =386 𝑐̅ =503 mund
𝑐 =235
𝐾 =46 𝐷 =8
𝑖=2 𝑒=2 𝑐̅ =544
ℎ=2 𝑐̂ =255 Erfurt 𝑓 =400 Stuttgart 𝑗=2
Hanno- Essen
ver 𝑊 =45 𝑐̅ =475
𝑆 =40
𝑖=3 𝐾 =28 𝑓 =520 𝐷 =15
Göttin- 𝑗=3
gen 𝑓 =380
Bonn
ℎ=3 𝑆 =50
Berlin 𝐷 =17
𝑗=4
Bochum

Suppliers: Factories: Warehouses: Markets:

n

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 88

2.2.4. Phase 4: Factor rating method and MIP
Multiple Supply Chain Stages

Factory
Production and Transportation Costs
per units

Kassel Erfurt Göttingen

Supplier Capacity
(i=1) (i=2) (i=3)

Bremen
324 386 255 30
(h=1)
Hannover
290 301 265 40
(h=2)
Berlin
275 350 290 50
(h=3)

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 89

2.2.4. Phase 4: Factor rating method and MIP
Multiple Supply Chain Stages

Warehouse
Production and Transportation
Costs per units

Nuremberg Stuttgart Fixed Cost Capacity

Factory
(e=1) (e=2) F(i) K(i)

Kassel
197 235 300 40
(i=1)
Erfurt
205 290 400 46
(i=2)
Göttingen
175 300 380 28
(i=3)

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 90

2.2.4. Phase 4: Factor rating method and MIP
Multiple Supply Chain Stages

Demand
Production and Transportation Costs
per units

Dortmund Essen Bonn Bochum Fixed Cost Capacity

Warehouse
(j=1) (j=2) (j=3) (j=4) F(e) W(e)

Nuremberg
495 503 544 475 630 60
(e=1)
Stuttgart
520 499 460 500 520 45
(e=2)

Demand 10 8 15 17

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 91

2.2.4. Phase 4: Factor rating method and MIP
Multiple Supply Chain Stages: GAMS Code

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 92

2.2.4. Phase 4: Factor rating method and MIP
Multiple Supply Chain Stages: GAMS Code

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 93

2.2.4. Phase 4: Factor rating method and MIP
Multiple Supply Chain Stages: GAMS - Code

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 94

2.2.4. Phase 4: Factor rating method and MIP
Multiple Supply Chain Stages: GAMS Code

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 95

2.2.4. Phase 4: Factor rating method and MIP
Multiple Supply Chain Stages

Locating Plants and Warehouses simultaneously (contd.)

solution with minimized total cost

ℎ=1 𝑖=1 𝑥 =7 𝑒=1 𝑐̅ =10 𝑗=1

Bremen Kassel Nürnberg Dort-
𝑥 =15 𝑐̅ =8 mund
𝑥 =28
𝑖=2 𝑒=2
Erfurt Stuttgart 𝑗=2
𝑐̅ =15 Essen
ℎ=2
𝑥 =28
Hanno- 𝑖=3
ver Göttin- 𝑗=3
gen 𝑐̅ =17 Bonn

𝑥 =22
𝑗=4
Bochum
ℎ=3
Berlin

Suppliers: Factories: Warehouses: Markets:

n m

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 96

2.2.4. Phase 4: Factor rating method and MIP
CPLM model extension

Accounting for taxes, tariffs and customer requirements

• Network design should maximize profits after taxes and tariffs while meeting
customer service requirements
• Maximizing profits after taxes:
– rj = revenue from selling one unit in market j (in common currency)
– Leading to a new objective function:

– To maximize the profit we have to relax the demand satisfaction constraint:

for m

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 97

Phase IV: Location Choices
CPLM model extension

Accounting for taxes, tariffs and customer requirements

• Considering customer preferences and requirements
– E.g. desired response time, choice of transportation mode or transportation provider
– Example: two different modes of transportation available between plant and market
• Mode 1 is sea, mode 2 is air
• New decision variables and
• To consider the desired response time in the model there must be a constraint
which allows transports just if the time taken is less than the desired response time
• Decision among different transportation providers may be modeled similarly

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 98

Summary

• Qualitative and quantitative factors needs to be considered in network design

• Factors that are measured on different scales (e.g., infrastructure good/bad vs.
cost positions) may be compared via factor rating methods
– But there are also other – more sophisticated – methods such as Multi Attribute
Decision Making methods (MADM), Analytical Hierarchy Process, Promethee etc.
• If we only compare monetary factors, in particular logistics cost (transportation,
inventory, fixed facility cost), Operations Research methods can help to find
optimal solutions (given the model’s assumptions)

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 99

Summary

• Several tools can be helpful in several phases, starting from network models
(discrete) and gravity models (continuous) to game-theoretic models (Hotelling)
• It was shown how different approaches can be used in different decision phases
– However, note, network design is an active research field, and methods are not solely
suitable for the phases in which they are presented here
– The assignment of methods to phases is just one example how decision support tools
might help in each phase

• Note: Please read Chapter 5.5. in Chopra/Meindl ( edition) for additional

factors to be considered in practice (“Making Network Decisions in Practice”)

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 100

References
NBER (National Bureau of Economic Research), 2022: How the US-China Trade
War Affected the Rest of the World,
[Link]
world, last accessed on 12th Dec 2023

SC D ESIGN P ROF. D R . GvUIDO VOIGT S LIDE 101