Applications in Supply Chain Analytics

Facility Location

1
What is a facility?
• Facilities are “fixed” points in a logistics network

• Their locations are major determinants of an organization’s logistics system efficiency

and its ability to successfully implement its competitive strategy

2
Factors Influencing the Location Decision

3
Expansion or Relocation
• Expand into new markets
• Development of new product lines
• Technological changes
• Competitive pressures
• Shifts in population that affect markets or workforce composition
• General growth of the company
• Improve competitive position or profitability by reconfiguring the logistics network,
and/or projecting a new image to customers and suppliers

4
Consolidation or Closing
• Made with increasing frequency in the past decade
• Declining competitiveness in world markets, a relaxed regulatory climate encouraging
mergers and acquisitions are contributing factors
• Cost-saving measure, spurred by erosion of market share, reduced profit margins,
increased competition, or a preemptive desire to increase logistical efficiency and
enhance competitive position, usually in a mature market
• Labor contracts frequently stipulate “givebacks” such as wage concessions, in return for
management guarantees against plant closings. Such considerations have become
significant determinants in facility location decisions.

5
Industry Type
• Labor-intensive, heavy industries place more emphasis on labor considerations
• Availability of a suitable workforce
• Management-worker climate
• Labor rates
• Degree of unionization
• Skill level
• Work ethic and productivity of the work force
• Attitudes of local public officials toward labor issues.

• The trend toward highly automated operations will shift emphasis toward other factors such
as the availability of a smaller, but more highly skilled, workforce.

6
Industry type (contd.)
• High-tech industries, which are generally innovation-oriented and R&D-intensive, place
greater emphasis on factors that help attract and keep a highly skilled workforce.
• Proximity of major universities and research centers, access to airports, and high quality-of-life
ratings are key factors attracting high-tech companies to specific locations.

• In industries where logistics costs are considerable (e.g., bulk packaging products and
beverages), logistical issues assume the greatest importance.
• Emphasize proximity to finished goods markets and raw material sources above all other
factors

7
Product Life Cycle
• In the earliest stages of product development, facilities are often located in major R&D
centers.
• High costs of facilities in these areas can be justified by the need to develop and introduce the
new product quickly.

• As product demand increases, transportation costs and delivery times become more
critical, often spurring a move to regionalized plants located near major markets.
• The rush to add capacity during the growth stage creates a cost structure which may
become the basis for success or failure when a product reaches maturity and the
competitive emphasis shifts to cost.

8
Product Life Cycle
• As a product matures, logistical considerations like facility location depend on a firm’s
competitive strategy.
• The decision to compete on innovation, service, or cost leadership, for instance, involves
different facility location considerations.
• For example, companies anticipating increased price sensitivity may relocate plants to areas
where production costs are lower.
• Before a product enters a decline, its period of profitability can often be extended through
careful locational adjustment.

9
Facility Type
• Survey by Fortune Magazine (1989) • Plants
• More likely to consider (re)locating new • Easy access to transportation services
regional headquarters, plant site or back- (especially trucking), domestic markets, and
office/processing facility. customers;
• Less likely to consider (re)locating a • Ample area for future expansion
distribution center or warehouse, and least • Ready availability of electrical utilities
likely to consider moving corporate
• Community receptivity to business and
headquarters.
industry
• Reasonable state and/or local tax structures
• Property costs
• Availability of skilled workers, as well as
their extent of unionization

10
Facility Type
• Distribution Centers and Warehouses
• The most important considerations are logistical
• Inbound and outbound transportation access and costs
• Level of customer service achievable from the facility

• Corporate Headquarters and R & D Facility

• Decisions are currently dominated by quality-of-life concerns due to the demand for a highly
skilled and educated workforce at these types of facilities.

11
Obnoxious Facilities
• Prisons and noxious facilities such as trash disposal plants and hazardous waste
incinerators
• Local residents often strongly oppose allowing these facilities in their area because they
produce relatively few new employment opportunities and limited additional tax
revenues vis-a-vis their perceived negative impact.
• Developers often offer incentives in the form of specific benefits (e.g., paid health care or
the development of a park or recreation area) to encourage the community to accept the
facility.

12
Models
• Single factor models

• Multi-attribute decision models

13
Single Factor Model – Distance 𝑥! , 𝑦!

Minimization
𝑥$ , 𝑦$
• There are 𝑛 existing facilities (could be suppliers,
customers, distribution centers, etc.) located at
coordinates (𝑥! , 𝑦! ).
𝑥, 𝑦
• The number of trips per unit time to each facility
represent the weight (𝑤! ) assigned to that facility.
• Locate new facility in relation to the existing 𝑥# , 𝑦#
facility such that the total distance travelled is
minimized. 𝑥" , 𝑦"
• Assume that distance is measured in a straight line. min ∑𝑤% 𝑑%
• The optimal location is the weighted average of 𝑑% = 𝑥% − 𝑥 " + 𝑦% − 𝑦 "
the 𝑥 and 𝑦 coordinates of all the existing
facilities. ∑𝑤% 𝑥% ∑𝑤% 𝑦%
𝑥= 𝑦=
∑𝑤% ∑𝑤%

14
Single Factor Model – Distance Minimization
• Alternative distance measurement methods can be considered.
• E.g. rectilinear distance.

• Alternative objectives may also be considered – minimize the maximum distance to

existing facilities.
• May be relevant where response time is important.
• E.g. location of emergency services such as fire stations, hospitals, etc.

15
Single Factor Model – Minimize Rectilinear Distance
𝑥! , 𝑦!

𝑥$ , 𝑦$

𝑥, 𝑦

𝑥# , 𝑦#

𝑥" , 𝑦"

16
The Fixed Charge Problem

17
The Fixed Charge Problem in Supply Chains
• Companies have several potential warehouses (facilities) from which they can serve
customer demand across different regions.
• Each warehouse has a fixed opening cost (e.g., rent, utilities, staff) that must be paid
whether the warehouse is used at full capacity or not.
• Additionally, there are variable transportation costs depending on the amount of goods
transported from a warehouse to customers.
• The goal is to decide which warehouses to open and how much inventory to transport
from each open warehouse to minimize the total cost (sum of fixed and variable costs)
while meeting customer demand.

18
Fixed Charge Problem
• The fixed-charge problem deals with • Given that 𝐹 is the fixed charge, 𝑐 is the
situations in which the economic activity variable unit cost, and 𝑥 is the level of
incurs two types of costs production, the cost function is expressed
• A fixed cost needed to initiate the activity as:
and a variable cost proportional to the level
𝐹 + 𝑐𝑥 𝑖𝑓 𝑥 > 0
of the activity. 𝐶 𝑥 =&
0 otherwise
• For example, the initial tooling of a machine
prior to starting production incurs a fixed • The function 𝐶(𝑥) is intractable analytically
setup cost regardless of how many units are because it involves a discontinuity at 𝑥 = 0.
manufactured. Once the setup is done, the
cost of labour and material is proportional to
the amount produced.

19
The Fixed Charge Problem
• Certain activities, such as operating a facility, using a vehicle, or starting production, incur
a fixed cost regardless of the amount of activity, in addition to a variable cost that
depends on the level of activity.
• If a facility undertakes a non-zero level of an activity (decision variable takes on a non-zero
value, i.e. the facility is open), it will incur a fixed cost regardless of the level of the activity plus
a variable cost proportional to the level of the activity.
• If the activity level (decision variable) associated with that facility is zero (i.e. the facility is
closed), the cost incurred at that facility will also be zero.

• The goal is to minimize the total cost, which is the sum of all fixed and variable costs,
while satisfying a set of constraints.

20
Common Applications
• Facility Location Problems
• Deciding where to open facilities to minimize fixed facility costs and transportation costs.

• Production Planning
• Determining whether to start production at a plant (incurs a fixed cost) and how much to
produce (variable cost).

• Transportation and Logistics

• Choosing which routes to use (fixed cost for starting a route) and how much to transport along
each route (variable cost).

21
The Fixed Charge Problem – Costs and Decisions
Costs Decisions
• Fixed Costs • Binary Decisions
• A cost incurred once a facility is opened, a • Decisions often involve binary variables
machine is started, or a service is (e.g., 0 or 1) to represent whether the
launched, regardless of the output or fixed cost is incurred (e.g., a facility is
usage level. open or closed).

• Variable Costs • Continuous Decisions

• Costs that increase with the level of • The variable cost depends on continuous
activity, such as production quantities, variables, such as the amount of product
distances travelled, or units consumed. shipped or the distance travelled.

22
The Transshipment Model Extended
• The objective function now becomes:

Total Cost = * 𝑓! 𝑦! + * * 𝑐"! 𝑥"! + * * 𝑐!# 𝑥!#

! " ! ! #

𝑓# = fixed cost of opening DC

(warehouse) j and Fixed cost 𝑓" incurred if DC 𝑗
is open (𝑦" = 1), else (𝑦" = 0)
𝑦" = 1 if warehouse j is opened, 0 it is zero.
otherwise (binary variable).

23
The Fixed Charge min 1 𝑓 𝑦 + 1 1 𝑐 " " !" 𝑥!" + 1 1 𝑐"$ 𝑥"$ Total cost of operating the
facilities + Total cost of
" ! " " $ transportation
Problem Formulation subject to
Supply constraint: the total (summed over all DCs) quantity
1 𝑥!" ≤ 𝑎! , ∀𝑖
supplied from each source to the DC should be less than the
" available quantity.
Consider the transshipment
model considered earlier.
1 𝑥"$ ≥ 𝑏$ , ∀𝑘 Demand constraint: the total (summed over all DCs) quantity
Let supplied to each destination should be more than the
"
demand at that destination.

𝑓# = fixed cost of opening DC

1 𝑥!" = 1 𝑥"$ , ∀𝑗 ∈ {1, … , 𝑛} Balance constraint
(warehouse) j and
! $
Big-M constraint: Flow can take place from a facility j only if it
𝑦" = 1 if warehouse j is opened, 0 is open (𝑦 = 1). 𝑀 is an arbitrarily large value.
𝑥!" ≤ 𝑀𝑦" , ∀𝑖, 𝑗
otherwise (binary variable).
𝑥!" , 𝑥"$ ≥ 0, 𝑦" ∈ {0, 1}

24
The 𝑝-Median Problem

25
The p-median problem
• Given:
• A set of potential facility locations (denoted as 𝐹).
• A set of demand points (denoted as 𝐷).
• A distance or cost metric 𝑑%- between each demand point and each facility location.

• The objective is to select 𝑝 facilities from the available set 𝐹 such that the total
"weighted" distance (or cost) between the demand points and their assigned facilities is
minimized.
• Each demand point is assigned to one facility, typically the one closest to it (in terms of
distance or cost).

26
The p-median problem – decision variables
• Let 𝑥"! be a binary variable that equals 1 if demand point 𝑖 is assigned to facility 𝑗, and 0
otherwise.
• Let 𝑦! be a binary variable that equals 1 if facility 𝑗 is selected, and 0 otherwise.

27
The p-median problem – objective
• Minimize the total cost of assigning each demand point 𝑑𝑖 ∈ 𝐷 to one of the selected
facilities 𝑓! ∈ 𝐹.
• The cost can be calculated as the sum of the distances from each demand point to its
assigned facility

Minimize 𝑍 = 4 4 𝑑%- ⋅ 𝑥%-

%∈/ -∈0

28
The p-median problem – constraints
• Each demand point is assigned to exactly one facility.

1 𝑥!" = 1 for all 𝑖 ∈ 𝐷

"∈&

• A demand point can only be assigned to an operational facility.

𝑥!" ≤ 𝑦" for all 𝑖 ∈ 𝐷, 𝑗 ∈ 𝐹

• Exactly 𝑝 facilities must be selected.

1 𝑦" = 𝑝
"∈&

29
The p-median problem – numerical example
• Number of facilities to select (𝑝): 2 • Distance matrix (𝑑"! )

A B C D E F
• Potential facility locations (𝐹): 1, 2, 3, 4, 5
1 4 3 2 7 6 5
2 6 5 3 8 7 4
• Demand points (𝐷): A, B, C, D, E, F 3 3 4 1 6 5 7
4 8 7 6 2 3 4
5 5 6 4 3 2 8

30
The p-median
param p; # Number of facilities to be selected
problem param
param
n; # Number of potential facilities
m; # Number of demand points
param d {1..n, 1..m}; # Distances between demand points and facilities

AMPL model (.mod) # Variables

var x {1..m, 1..n} binary; # x[i,j] = 1 if demand point i is assigned to facility j
var y {1..n} binary; # y[j] = 1 if facility j is selected

# Objective function: Minimize the total distance

minimize TotalDistance: sum {i in 1..m, j in 1..n} d[j,i] * x[i,j];

# Constraints:
# Each demand point is assigned to exactly one facility
subject to AssignDemand {i in 1..m}: sum {j in 1..n} x[i,j] = 1;

# A demand point can only be assigned to an open facility

subject to FacilityAssignment {i in 1..m, j in 1..n}: x[i,j] <= y[j];

# Exactly p facilities should be selected

subject to FacilitySelection: sum {j in 1..n} y[j] = p;

31
The p-median problem
# Data
param p := 2; # Number of facilities to be selected
param n := 5; # Number of potential facilities
param m := 6; # Number of demand points

# Distances between demand points and facilities

# For each row (facility) and column (demand point),
AMPL data (.dat)
# the distances are listed
# in the format:
# facility demand_point distance
# For example, 1 1 4 means the distance from facility
# 1 to demand point A is 4.

param d :=
1 1 4 1 2 3 1 3 2 1 4 7 1 5 6 1 6 5
2 1 6 2 2 5 2 3 3 2 4 8 2 5 7 2 6 4
3 1 3 3 2 4 3 3 1 3 4 6 3 5 5 3 6 7
4 1 8 4 2 7 4 3 6 4 4 2 4 5 3 4 6 4
5 1 5 5 2 6 5 3 4 5 4 3 5 5 2 5 6 8;

32
The p-median problem
ampl: include "p-median.run";
Gurobi 11.0.2: optimal solution; objective 17
13 simplex iterations
1 branching node
y [*] :=
1 0
2 0
Solution
3 1
4 1
5 0
;

x [*,*]
: 1 2 3 4 5 :=
1 0 0 1 0 0
2 0 0 1 0 0
3 0 0 1 0 0
4 0 0 0 1 0
5 0 0 0 1 0
6 0 0 0 1 0
;

TotalDistance = 17

33
The p-Centre problem
Aspect P-Centre Problem P-Median Problem

Minimize the maximum distance from any Minimize the total distance from all
Objective
customer to the nearest facility. customers to their nearest facility.

Focuses on the worst-case scenario (the most Focuses on the overall cost or distance (all
Focus
distant customer). customers).

Aiming to limit the maximum travel time (or Aiming to minimize the total travel time (or
Goal
distance) for any customer. distance) for all customers.

Location of emergency services (e.g., fire Location of retail stores or distribution centres
Example
stations) to minimize response time for the to minimize the total travel distance for
Application
farthest person. customers.

34
Multi-attribute Model

35
Multi-attribute Model – Categorization of Factors
• Critical factors • Objective factors
• Absolutely essential for a particular location to be • Can be measured in monetary terms
considered • Converted to a scale of 0 to 1 after consolidation
• Evaluated on a scale of 0 (not present) or 1 • E.g. transportation cost, land cost, etc.
(present/available)
• May be assessed objectively or subjectively if • Subjective factors
present (evaluated to 1) • Cannot be measured in monetary terms (but may
• E.g. regulatory permissions be measured quantitatively)
• Requires subjective assessment by the person
evaluating the factor
• Converted to a scale of 0 to 1 after giving
appropriate weightage to each factor
• E.g. political stability, weather, etc.

36
The Brown-Gibson Model
Multi-attribute decision making framework

37
Illustration
A company that manufactures pumps is currently located in the industrial district of a large
industrial city. Several problems have led to the decision to relocate. First, and most important,
the present site is old and inefficient and there is no room left for easy expansion. A contributing
factor is high labor rates, which translate almost directly into high labor costs in the products.
The company has been able to operate with an independent union, and the president feels that
relationships have been excellent. Recently, there has been a great deal of pressure for the local
union to associate with a larger national level union. Although this is not the major reason for
moving, the president hopes that a carefully planned move may avoid future labor problems,
which would have an adverse effect on the company. As the search for a new site continues, the
president continues to place more and more emphasis on this labor relations factor.

38
Illustration (contd.)
The search has finally narrowed to four cities and comparative data are shown in the table below. Objective
factor costs that can be measures and that seem to be affected by alternative locations are shown in table
1. The president is surprised by the range in these costs, city D having costs 1.9 times those of city A.

Seven subjective factors have been isolated as having importance and the search staff has made a
preliminary rating of each factor in table 2. The first six factors were rated on a scale of “excellent, plentiful,
very good, good, adequate or fair.” The seventh factor, union activity, was rated “active, significant,
moderate or negligible.”
When the summary data were presented to the president, he was impressed, but he expressed concern
about how to equate the various objective and subjective factors. Based on that data collected what
decision would you make? How sensitive is your decision to the objective factor weight (x)?

39
Illustration Data
!"#$%F'
("#)* !*"+,-)*. /*)-%*.0F 23%$ 4.".%F!"1%, 5$%6.*76F 9".%*
!"1%, /)8%*
!"#$%F G() *(+ *(*, *(*- *(*. *(*- *(*.
!"#$%/ G(+ *(0 *(*) *(*+ *(*- *(*0 *(*,
!"#$%! G(1) *(+ *(*+ *(*+ *(*) *(*) *(*,
!"#$%2 ,(-) *() *(G *(*+ *(*1 *(*+ *(*)

!"#$%F:
("#)*F !0-%F);F <..7.3=% <--%"*"+6% !*"+,-)*. L%6*%".7)+ ?+7)+F
43--$0 ("#)* <6.7@7.0
!"#$%F F34567#4 8993 8993 :7"; 8993 8993 <"=>"?"@7>#
!"#$%/ AB4>#"?6B CD@4BB4># N4;$%8993 8993 N4;$%8993 N4;$%8993 F4=B"="GB4
!"#$%! AB4>#"?6B CD@4BB4># N4;$%8993 8993 8993 N4;$%8993 F4=B"="GB4
!"#$%2 AB4>#"?6B CD@4BB4># 8993 8993 N4;$%8993 N4;$%8993 F@#"H4

40
 Objective Factors

!"#$%F'($)*+%F,-.
/+",- 0-+1.2,-F 3-,2$-F4) *6$7 8F+F$)0+5$. 97$%F-'%) ;+F$- 0,F+7)!*< !*=
0+5$. 3,:$-
!"#$%& '() *(+ *(*, *(*- *(*. *(*- *(*. .(.) L?@@
!"#$%/ '(+ *(0 *(*) *(*+ *(*- *(*0 *(*, .()) @?AB
!"#$%! '(1) *(+ *(*+ *(*+ *(*) *(*) *(*, .(0 @?CA
!"#$%2 ,(-) *() *(' *(*+ *(*1 *(*+ *(*) -(, @?@@

41
!"#$%F'()%*+,F'-./
0,#-.* 431%*-5* 6''('"7% 611%,.,8F% 4.,8/1-.' 9%F.%,'(-8 :8(-8*
!"1123 0,#-. 6F'()('3
!"#$%F FG()*+#( ,--G ,--G .+"/ ,--G ,--G 0"12"3"4+2#
!"#$%5 67(2#"3*7 894(77(2# :(/$%,--G ,--G :(/$%,--G :(/$%,--G ;(17"1"<7(
!"#$%! 67(2#"3*7 894(77(2# :(/$%,--G ,--G ,--G :(/$%,--G ;(17"1"<7(
!"#$%= 67(2#"3*7 894(77(2# ,--G ,--G :(/$%,--G :(/$%,--G F4#">(

0,#-.* 431%*-5* 6''('"7% 611%,.,8F% 4.,8/1-.' 9%F.%,'(-8 :8(-8*

!"1123 0,#-. 6F'()('3
;%(<='/ L?@ L?@ L?@ L?@ L?@ L?@ L?A
!"#$%F ?@A ?@B ?@B ?@C ?@B ?@B ?@M
!"#$%5 ?@N C ?@F ?@B ?@F ?@F C
!"#$%! ?@N C ?@F ?@B ?@B ?@F C
!"#$%= ?@N C ?@B ?@B ?@F ?@F ?

FG()*+#( ?@A F4#">( ?

894(77(2# C G-G(/+#( ?@F
.+"/ ?@C ;(17"1"<7( C
,--G ?@B 0"12"3"4+2# ?@M
67(2#"3*7 ?@N
:(/$%,--G ?@F

42
Location Measure x !"#
𝒊 $%F '%F (F
!"#$%& '()) )(*' )(+,
Find Location Measure (LM).
!"#$%- )(./ )(.) !")*
Select the facility with the highest LM.
!"#$%! )(+. )(+. )(+.
!"#$%0 )()) )(*. )('/

𝐿𝑀7 = 𝑥 ⋅ 𝑂𝐹𝑀7 + 1 − 𝑥 ⋅ 𝑆𝐹𝑀7

43