ISOM 2700: Operations Management
Session 8: Resource Management III
Dongwook Shin
Dept. ISOM, HKUST Business School
Learning Objectives: Session 8
• Solving LP using Excel
• Sensitivity analysis using Excel
1
Why Worry About LPs?
• Many real life problems can be modeled as LPs
• Linear programs are scalable
• With the number of decision variables and the number of
constraints of the order of up to a 100,000, we can solve them
on our computers
• LPs allow for sensitivity analysis
• What happens when we change the constraints slightly?
• What happens when we change the objective function slightly?
2
Why Excel?
• Advantages of Excel
• User interface: can see all calculations
• Mathematical, plotting, and database functionality
• Extendable functionality using VBA and C/C++
• Most widely-used software for different industries
• Disadvantages of Excel
• Not designed for large computations
• Not designed for repetitive tasks (e.g., looping)
• All results are stored (inefficient use of memory)
• Difficult to audit/verify/debug spreadsheets
3
Shelby Shelving Case
• A small company that produces two types of shelves for
grocery stores: Model S and Model LX
• Three steps in manufacturing process:
• Stamping, forming, and assembly
• Capacity of Stamping and Forming machines: 800 hours a
month
• Capacity of model S assembly department: 1900 units/month
• Capacity of model LX assembly department: 1400 units/month
• Machine Requirements (Hours/per unit):
• Stamping: 0.3h/u for model S and 0.3h/u for model LX
• Forming: 0.25h/u for model S and 0.5h/u for model LX
• Selling Price: Model S $260/unit; Model LX $245/unit
4
Shelby Shelving: Problem Elements
• Shelby Shelving company would like to determine how
many items to produce for each type to maximize
revenue, given constraints in resources
• Decision variables
• S: # units of Model S type
• LX: # units of Model LX type
• Objective function
• Revenue = 260 S + 245 LX
• Constraints
• (S assembly) S £ 1900
• (LX assembly) LX £ 1400
• (Stamping) 0.3 S + 0.3 LX £ 800
• (Forming) 0.25 S + 0.5 LX £ 800
• (Nonnegativity) S, LX ³ 0 5
Shelby Shelving: LP Formulation
max 260 S + 245 LX (Revenue)
subject to:
(S assembly) S £ 1900
(LX assembly) LX £ 1400
(Stamping) 0.3 S + 0.3 LX £ 800
(Forming) 0.25 S + 0.5 LX £ 800
(Nonnegativity) S, LX ³0
6
Shelby Shelving: Feasible Region
LX model
2500
0.
3S S=1900 (S assembly)
+0
2000 .3
LX
=8
00
(S
1500 ta
m
pi
ng LX=1400 (LX assembly)
)
1000
0.2
5S
Feasible Region for +0
.5 L
500 Shelby Shelving X=
800
(Fo
rmi
ng )
0 500 1000 1500 2000 2500 3000 S model
7
Shelby Shelving: Iso-profit Line
LX model
2500
2000 Profit for (S=1000, LX = 500)
= 260 ´ 1000 + 245 ´ 500 = 382,500
1500
1000
(S=1000, LX=500)
500
Profit = 382,500
0 500 1000 1500 2000 2500 3000 S model
8
Shelby Shelving: Optimal Solution
LX model Direction of increasing profit
2500
Optimal solution:
2000 This is as far as we can
go in the direction of
increasing profit!
1500
(S=1,900, LX=650)
1000 Pr
Pr of
Pr of it
of it =
= 65
it 49 3,
= 25
500 38 4,
2, 00 0
50 0
0
0 500 1000 1500 2000 2500 3000 S model
9
The Initial Spreadsheet
10
The Solver Dialog Box
Load solver in Mac
11
The Solver Dialog Box Explained
Add objective function cell here
and specify minimize or maximize
objective
The decision variable cells are
listed here
This is where you list constraints:
• Use the “Add” button to add
The “Solve”
new constraints
button will
• Use the “Change” button to
optimize the
edit an existing constraint
model
• Use the “Delete” button to
delete a constraint
Checking this automatically adds
the constraint that all decision
variables must be non-negative.
Simplifies the job of adding non-
negativity constraints
Select the algorithm here. For
now, use “Simplex LP”
12
Adding Constraints
1. Click on the blank space for “Cell Reference” and select
the total usage quantities, cells E11:E14
2. Click on the drop box in the middle and select “<=”
3. Click on the blank space for “Constraint” and select the
raw material availability values in cells G11:G14
13
The Solver Results Dialog Box
After pressing “Solve” the model will run and if successful
you will see this dialog box
Click Sensitivity in “Reports” column and press “OK”
14
The Spreadsheet after Optimizing
15
Shelby Shelving: Optimal Solution
max 260 S + 245 LX (Net Profit)
subject to:
(S assembly) S £ 1900
(LX assembly) LX £ 1400
(Stamping) 0.3 S + 0.3 LX £ 800
(Forming) 0.25 S + 0.5 LX £ 800
(Nonnegativity) S, LX ³0
Optimal solution: (S, LX) =(1900, 650), Net Profit = $653,250
16
Learning Objectives: Session 8
• Solving LP using Excel
• Sensitivity analysis using Excel
17
Shelby Shelving: Sensitivity Analysis
If the RHS is nicreased from 1900 to 1901, the revenue is increased by $137.5
Only increase by this amount, the shadow price remains the same
• Should we increase the production of Model S?
• Yes,
Yes, because
shadow the=shadow
price $137.5 price of increasing Model S is $137.5 per unit
per unit
• How much should we increase and how would profit increase?
We can
• 233
By increase
units Model
of Model S by 233
S, profit units, and
increased bythe profit would
$137.5*233 increase by $137.5
= $32037.5
x 233 = $32,037.5
18
Shelby Shelving: Sensitivity Analysis
LX model
2500
S=1900 S=1900 + ∆
2000
1500
The shadow price for
1000 increasing Model S is 137.5,
when ∆ ≤ 233.3
500
0 500 1000 1500 2000 2500 3000 S model
19
Shelby Shelving: Sensitivity Analysis
• General rule: If there are small changes in the objective coefficients,
the optimal solution does not change
• Optimal production plan would not change if
• The
•The price
price for for Model
Model S rangesfrom
S ranges from122.5
122.5(=260-137.5)
(=260-137.5)to
to infinity
infinity
• The
•The price
price for for Model
Model LXLXranges
rangesfrom
from 00 (=245-245)
(=245-245) to
to 520
520 (=245+275)
(=245+275)
20
Shelby Shelving: Sensitivity Analysis
• What is the “increase” in the optimal objective value if
the selling price of the LX model is changed from $245 to
$545?
1. It cannot be determined from the data
2. $300
3. $195,000
4. -$300
5. -$195,000
21
Shelby Shelving: Sensitivity Analysis
LX model
2500
Sl
Slope = -∞/245
op
2000
e
=
-2
S lo
60
pe
= -1 The optimal production
/2
1500 22.
45
5/2 plan would not change if
45 the price for Model S
range from 122.5 to ∞
1000
500
0 500 1000 1500 2000 2500 3000 S model
22
Shelby Shelving: Sensitivity Analysis
• A new product, GX, is proposed with the following characteristics
• Selling price: $150 No.
If you produce one unit of GX, then existing models (S&LX)
• Hours stamping: 2
can only use the forming machine for 799.5 hours.
• Hours Forming: 0.5
• No assembly needed
How much the revenue from the 2 models decreased by
producing one unit of GX?
• Is it worth producing? —> Shadow Price!!!
= -0.5 x 490 = -245 > 150 23
Ad Budget Allocation
• Quicker Oats must determine how much of its $200,000 advertising
and promotional budget should be spent in the following mediums
• TV
• Radio
• Magazines
• Prize promotion
• Each dollar spent on TV, radio, magazines, and prize promotion
increases sales $10, $5, $5, and $20, respectively
• Television advertising cannot exceed one-half of the total budget
• Total radio advertising must be at least 20% of total TV advertising
• At least $20,000 must be spent on magazine ads
• No more than $25,000 can be spent on the prize promotion
• Goal: Maximize the total increase in Quicker’s sales volume
24
Ad Budget Allocation: LP Formulation
• Decision variables
• XT = dollars spent on TV ads
• XR = dollars spent on radio ads
• XM = dollars spent on magazine ads
• XP = dollars spent on prize promotions
• Objective function: Total increase in sales volume
• 10XT +5XR +5XM +20XP
• Constraints
• (Budget) XT +XR +XM +XP ≤200,000
• (TV) XT ≤100,000
• (TV and radio) XR ≥0.2XT
• (Magazine) XM ≥20,000
• (Prize) XP ≤25,000
25
Ad Budget Allocation: LP Formulation
Max 10XT +5XR +5XM +20XP
s.t. XT + XR + XM +XP ≤ 200,000(Budget)
XT ≤ 100,000 (TV)
0.2XT − XR ≤0 (TV and radio)
− XM ≤−20,000 (Magazine)
XP ≤ 25,000 (Prize)
26
Ad Budget Allocation: Spreadsheet After
Optimizing
27
Ad Budget Allocation: Sensitivity Analysis
• Suppose a dollar spent on TV can increase the sale by $5, instead of
$10. Does the optimal solution change?
No,• because the optimal
No, because solution
the optimal does
solution notnot
does change asaslong
change longas
asthe
theobjective
objective coefficient
coefficient
for TV is withinfor TVrange
the is within the range [5, ∞]
[5,1E+30)
28
Ad Budget Allocation: Sensitivity Analysis
• How much does the optimal value increase if the total budget
increases by $10,000?
• The
The shadow
shadow price
price of the
of the Budget
Budget constraintisis5.5.Therefore,
constraint Therefore, if the
thebudget
budget increases
increases
by $10,000, byobjective
the $10,000, the objective
function function
increases byincreases by 5=x$50,000
5 x $10,000 $10,000 = $50,000
29
Takeaways
• Use Excel Solver to solve LP and to perform sensitivity
analysis
• Why do we need to use the sensitivity report, instead of
solving the problem many times using Solver?
• SR is a compact way of storing information
• Solving multiple times is not practical if there are thousands of
constraints
• It is needed to understand LP
• It will be on the midterm
• Next class: Service Management
30
