Example 14.

Introduction to LP Modeling
Linear Programming
 Linear programming (LP) is a method of
spreadsheet optimization.

 LP is used in all type of organizations and it is used

to solve an extremely wide variety of problems.

 The simplex method is a systematic, arithmetic

intensive search through the set of all possible
solutions for the solution that optimizes a given
objective.

 We will not be using the simplex method; instead, we

will learn to formulate problems as LP models.
14.1a | 14.2 | 14.3
Background Information
 The Monet Company produces four types of picture
frames, which we label 1, 2, 3, and 4. The four types
of frames differ in respect to size, shape, and
materials used.

 Each type requires a certain amount of skilled labor,

metal, and glass as shown here. The table also lists
the unit selling price Monet charges for each type of
frame.
Skilled Labor Metal Glass Selling Price

Frame 1 2 4 6 $28.50
Frame 2 1 2 2 $12.50
Frame 3 3 1 1 $29.25
Frame 4 2 2 2 $21.50

14.1a | 14.2 | 14.3

Background Information --
continued
 During the coming week Monet can purchase up to
4000 hours of skilled labor, 6000 ounces of metal,
and 10,000 ounces of glass.

 The unit costs are $8.00 per labor hour, $0.50 per
ounce of metal and $0.75 per ounce of glass.

 Also, market constraints are such that it is impossible

to sell more than 1000 type 1 frames, 2000 type 2
frames, 500 type 3 frames, and 1000 type 4 frames.

 The company wants to maximize its weekly profit.

14.1a | 14.2 | 14.3

Solution
 In the traditional algebraic solution method, we first
identify the decision variables.

 In this small problem they are the number of frames

of type 1, 2, 3, and 4 to produce.

 We label these x1, x2, x3, and x4.

 Next, we write total profit and the constraints in terms

of x’s.

 Finally, since only nonnegative amounts can be

produced, we add explicit constraints to ensure that
the x’s are nonnegative.
14.1a | 14.2 | 14.3
Solution -- continued
 The resulting algebraic formulation is shown below:
maximize 6x1 + 2x2 + 4x3 + 3x4 (profit objective)
subject to 2x1 + x2 + 3x3 + 2x4 < = 4000 (labor constraint)
4x1 + 2x2 + x3 + 2x4 < = 6000 (metal constraint)
6x1 + 2x2 + x3 + 3x4 < = 10,000 (glass constraint)
x1 < = 1000 (frame 1 sales constraint)
x2 < = 2000 (frame 2 sales constraint)
x3 < = 500 (frame 3 sales constraint)
x4 < = 1000 (frame 4 sales constraint)
x1, x2, x3, x4 > = 0 (nonnegative constraint)

14.1a | 14.2 | 14.3

Solution -- continued
 To understand this formulation, consider the profit
objective first.
– The profit from x1 frames of type 1 is 6x1 because each
frame contributes $6 to profit. This $6 is calculated as the
unit selling price minus the cost of the inputs that go into a
single type 1 frame:
Unit profit = 28.50 - [2(8.00) + 4(0.50) + 6(0.75)] = $6

– Profits for the other three types are obtained similarly. Their
unit profits are $2.00, $4.00, and $3.00, respectively.

– Then the total profit is the sum of the profits from the four
products.
14.1a | 14.2 | 14.3
Solution -- continued
 Next, consider the skilled labor constraint.
– The right hand side, 4000, is the number of hours available.
On the left-hand side each type 1 frame uses 2 hours of
labor, so x1 units require 2x1 hours of labor.

– Similar statements hold for the other three products, and the
total number of labor hours used is the sum over the four
products.

– Then the constraint states that the number of hours cannot

exceed the number of hours available.

14.1a | 14.2 | 14.3

Solution -- continued
 The constraints of metal and glass are similar.

 Finally, the maximum sales constraints and the

nonnegativity constraints put upper and lower limits
on the quantities that can be produced.

 For many years all LP problems were formulated this

way.

 Because of this, many commercial LP computer

packages are written to accept LP problems in this
format.
14.1a | 14.2 | 14.3
Solution -- continued
 In the past decade a more intuitive method of
expressing LP problems has emerged.

 The method takes advantage of the power and

flexibility of spreadsheets.

 Actually, LP problems could always be formulated on

spreadsheets, but now with the addition of Solver
add-ins, spreadsheets have the capability of solving
LP problems as well.

 There are many ways to develop an LP spreadsheet

model.
14.1a | 14.2 | 14.3
Spreadsheet Elements
 The common elements in all LP spreadsheet model
are the following:
– Inputs. All numerical inputs - that is, the data needed to
form the objective and the constraints - must appear
somewhere in the spreadsheet. It is not necessary but it
often helps to enclose inputs in a blue border with shading.

– Changing cells. Instead of using variable names, such as

x’s, there is a set of designated cells that play the role of the
decision variables. These values in these cells can be
changed to optimize the objective. In Excel these cells are
called the changing cells. To designate these we often
enclose them in a red border.
14.1a | 14.2 | 14.3
Spreadsheet Elements --
continued
– Target (objective) cell. One cell, called the target cell or
the objective cell, contains the value of the objective. Solver
systematically varies the values in the changing cells to
optimize the value in the target cell. Our convention is to
enclose the target cell within a black double line border.

– Constraints. Excel does not show the constraints directly on

the spreadsheet. Instead we specify constraints in a Solver
dialog box.

– Nonnegativity. Normally we want the decision variables -

that is, the values in the changing cells - to be nonnegative.
For certain versions of Excel these nonnegativity constraints
might need to be specified explicitly.

14.1a | 14.2 | 14.3

Solution -- continued
 In general, the complete solution to the problem
involve three stages.

 The first stage is to enter all the inputs, trial values for
the changing cells, and formulas relating these in a
spreadsheet. We call this formulating the model.
– This stage if the most crucial because it is here that all of the
“ingredients” of the model are included and related
appropriately.
– In particular, the spreadsheet must include a formula that
relates the objective to the changing cells, so that if the
values in the changing cells vary, the objective value varies
automatically.
14.1a | 14.2 | 14.3
Solution -- continued
 After the model is formulated, we can proceed to the
second stage: invoking Solver. At this point we
formally designate the objective cell, the changing
cell, and the constraints, and we tell solver to find the
optimal solution.
– If the first stage is done correctly, the second stage is usually
very straightforward.

14.1a | 14.2 | 14.3

Solution -- continued
 The third stage is sensitivity analysis. In most
model formulations of real problems, we make “best
guesses” for the numerical inputs to the problem.
There is typically some uncertainty about quantities
such as the unit prices, forecasted demands, and
resource availability.
– When we use Solver to solve he problem, we use our best
estimates of these quantities to obtain an optimal solution.

14.1a | 14.2 | 14.3

 PRODUCTMIX.XLS
 This file illustrates the solution procedure for Monet’s
product mix problem.

 The first stage is to set up the spreadsheet, as

explained in step-by-step fashion.

14.1a | 14.2 | 14.3

14.1a | 14.2 | 14.3
Developing the Spreadsheet
Model
 Inputs. Enter the various inputs in the range B4:B6,
B9:E12, B18:E18, and D21:D23.

 Production levels. Enter any four values in cells

B16:E16. These do not have to be the values shown
in 14.1. These cells are the changing cells; that is,
the cells where the decision variables are placed. Any
trail values can be used initially; Solver will eventually
find the optimal values. Note that the four values
shown in the initial solution cannot be optimal
because they do not satisfy all of the constraints.

14.1a | 14.2 | 14.3

Developing the Spreadsheet
Model -- continued
 Resources used. Enter the formula
=SUMPRODUCT(B9:E9,Produced)
in cell B21 and copy it to the range B22:B23. These
formulas calculate the units of labor, metal, and glass
used by the current product mix. The SUMPRODUCT
function is particularly helpful: it multiplies each value
in the range by the corresponding value in the
Products range and then sums these products.

14.1a | 14.2 | 14.3

Developing the Spreadsheet
Model -- continued
 Revenues, costs and profits. The area from row 25
down shows the summary of monetary values.
Actually, all we need is the total profit in cell F32, but
it is useful to calculate the ingredients of this total
profit (that is, the revenues and costs associated with
each product). To obtain the revenues enter the
formula =B12*B16 in cell B27 and copy this to the
range C27:E27. For the costs, enter the formula
=$B$4*B$16*B9 in cell B29 and copy this to the
range B29:E21.

14.1a | 14.2 | 14.3

Developing the Spreadsheet
Model -- continued
 Revenues, costs and profits - continued. Then
calculate profits for each product by entering the
formula =B27-SUM(B29:B31) in cell B32 and copy
this to range C32:E32. Finally, calculate the totals in
column F by summing across each row with the SUM
function.

 The next step is to specify the changing cells, the

objective cell, and the constraints in a Solver dialog
box, and to instruct Solver to find the optimal solution.
Before we do this, it is useful to try a few guesses in
the changing cell.
14.1a | 14.2 | 14.3
Developing the Spreadsheet
Model -- continued
 Trying a few guesses has two purposes.
– First, by entering different sets of values in the changing
cells, we can confirm that the formulas in the other cells are
working properly.

– Second, is to provide a better understanding of the model

 The resulting solution appears next.

 The corresponding profit is $8750.

14.1a | 14.2 | 14.3

14.1a | 14.2 | 14.3
Solution -- continued
 We have now produced as much as possible of the
three frame type with the three highest profit margins.
Does this guarantee that this solution is the best
possible product? Unfortunately, it does not!

 The solution is not optimal. Even in this small model it

is difficult to guess the optimal solution, even when
we use a relatively intelligent trial and error
procedure.

 The problem is that a frame type with a high profit

margin can use up a lot of the resources and
preclude other profitable frames from being
produced.
14.1a | 14.2 | 14.3
Using Solver
 To invoke Excel’s Solver, select the Tools/Solver
menu item. This dialog box appears.

14.1a | 14.2 | 14.3

Using Solver -- continued
 It has three important sections that you must fill in:
the target cell, the changing cells, and the
constraints.

 For the product mix problem, we can fill these in by

typing cell references or we can point, click and drag
the appropriate range in the usual way.
– Objective. Select TotProfit (cell F32) as the target cell, and
click on the Maximize button.

– Changing cells. Select the Produced range (B16:E16), the

numbers of frames to produce, as the changing cells.
14.1a | 14.2 | 14.3
Using Solver -- continued
– Constraints. Click on the Add button to add the following
constraints:
Used <= Available
Produced < = MaxSales
The first constraint says to use no more of each resource
than is available. The second constraint say to produce no
more of each product than can be sold.

– Nonnegativity. Although negative production quantities

obviously make no sense, we must tell Solver explicitly to
make changing cells nonnegative. There are two ways to do
this:

14.1a | 14.2 | 14.3

Using Solver -- continued
• First, we can add another constraint in Step 3:
Produced >=0.

• Second, Excel 97 we can click the options button in the Solver

dialog box and we can check the Assume Non-negative box in
the resulting dialog box (see here).

14.1a | 14.2 | 14.3

Using Solver -- continued
– Linear model. There is one last step before clicking on the
Solve button. Solvers uses one of the several numerical
methods to solve various type of problems. Linear problems
can be solved most efficiently by the simplex method. We
must check the Assume Linear Model in the Solver options.

– Optimize. Click on the Solve button.

 At this point, Solver searches through a number of

possible solutions until it finds the optimal solution.
When it finishes it displays this message.

14.1a | 14.2 | 14.3

Using Solver -- continued

 You can tell it to return the values in the changing

cells to their original values or retain the optimal
values found by Solver.

 In some cases Solver is not able to find an optimal

solution, in which case one of several error
messages will appear.
14.1a | 14.2 | 14.3
Using Solver -- continued
 For now, clicking on the OK button to keep the Solver
Solution. You should see the solutions shown on the
next slide.

 The optimal plan is to produce 1000 type 1 frames,

800 type 2 frames, 400 type 3 frames, and no type 4
frames.

 This is close to the production plan, but the current

plan earns $450 more profit.

 Also, it uses all available labor hours and metal, but

only 8000 of the 10,000 ounces of glass available.
14.1a | 14.2 | 14.3
14.1a | 14.2 | 14.3
Using Solver -- continued
 Finally, in terms of maximum sales, the optimal plan
could produce more of frame types 2, 3, 4 (if there
were more skilled labor and/or metal available).

 This is typical of an LP solution.

 Some of the constraints are met exactly; that is, as

equalities, while others contain a certain amount of
“slack”.

14.1a | 14.2 | 14.3

Experimenting
 If we want to experiment with different inputs to this
problem - the unit revenues or resources available,
for example - we can simply change the inputs and
then rerun Solver.

 As a simple what-if example, consider the modified

model in the output on the next slide.

14.1a | 14.2 | 14.3

Solution to Product Mix with
New Inputs

14.1a | 14.2 | 14.3

Experimenting -- continued
 Here the unit price for frame type 4 has increased
from $21.50 to $26.50, and all other inputs have
remained the same.

 By making type 4 frames more profitable, we might

expect them to enter the optimal mix.

 This is exactly what happens.

 The new optimal plan discontinues production of

frame type 2 and 3, and instead calls for production
of 1000 units of frame type 4.
14.1a | 14.2 | 14.3
Experimenting -- continued
 There is one technical note you should be aware of.

 Because of the way the numbers are stored and

calculated on a computer, the optimal values in the
changing cells and elsewhere can contain small
roundoff errors.

 For example for all practical purposes

.000000008731 can be be treated as 0.

14.1a | 14.2 | 14.3