You are on page 1of 50

Multifactor Decision

Making
Jaco Wijnmaalen
jaco@quartzconsult.c

Todays Agenda
Decision Hierarchy
Objectives

Pair-wise comparisons

Engineering
vs
Management

Consistency
Selection

FIAP 271
Decision Making
MFEP Explained
AHP Explained

Opening Screen
Workspace
Pair-wise comparison
Consistency
Final Decision

Application

Objectives
Define and describe 2 types of scoring
models.
Use multi-factor scoring methods to make
decision choices.
Describe how Analytical Hierarchy Process
can be used to make decision choices.
Understand and use consistency statistics
to support AHP results.
Understand how spreadsheets can be used
to model AHP decision problems.

Management or Engineering?

Decision Making
Many decisions may involve a number of
factors to consider.
E.g. applying for a new job has many things
to consider.
Salary,
Career opportunities,
Location, etc

Or purchasing a new PC has many factors.


Processor speed,
Memory,
Backup service, etc

Multi-Factor Decision Making


Multifactor Evaluation Process (MFEP)
A more quantitative approach.
Important factors are given an appropriate
weight (relative importance).
Each alternative is then evaluated against
these factors.
Totals are added up.
The alternative with the highest score wins.

Example: Job Application


Alternatives to consider:
Alternative Action Co
Engineering Systems Design Ltd
Product Wiring Inc

Important Factors:
Starting salary
Career opportunities
Location

Factor Weights
Various factors are weighted according to
importance.
Salary: 0.3
Career advancement opportunities: 0.6
Location: 0.1

Important: weights must add up to 1


Factor

Importance (Weight)

Salary

0.3

Career Advancement

0.6

Location

0.1

Factor Evaluations
Each alternative is given a rating for each
factor considered on a scale of 0 to 1.

Factor

AA Co

EDS Ltd

PW Inc

Salary

0.7

0.8

0.9

Career Advancement 0.9

0.7

0.6

Location

0.8

0.9

0.6

Factor and Alternative


Comparison
With this information it is possible to determine
total weighted evaluation for each alternative.
Each company is given a factor evaluation for
each factor.
Factor weights are multiplied by factor
evaluations and added up to get a total
weighted evaluation for each company.
The highest score wins.

Alternative Evaluations
Evaluation for AA Co
Factor

Factor Weight

Factor Evaluation

Weighted Evaluation

Salary

0.3

0.7

0.21

Career

0.6

0.9

0.54

Location

0.1

0.6

0.06

Total

0.81

Evaluation for EDS Ltd


Factor

Factor Weight

Factor Evaluation

Weighted Evaluation

Salary

0.3

0.8

0.24

Career

0.6

0.7

0.42

Location

0.1

0.8

0.08

Total

0.74

Evaluation for PW Inc


Factor

Factor Weight

Factor Evaluation

Weighted Evaluation

Salary

0.3

0.9

0.27

Career

0.6

0.6

0.26

Location

0.1

0.9

0.09

Total

0.72

Analytical Hierarchy Process


Break down a big problem into small
digestible bites.
Prioritise alternatives in decision making
Streamline human decision processes
Easy to use
Well accepted in industry by decision makers
Can be used for multiple decision problems
Very controversial

Analytical Hierarchy Process


Where many factors are to be evaluated
accurately, a more scientific approach is
needed.
Analytical Hierarchy Process* involves pairwise comparisons.
Start by laying out the overall hierarchy of the
decision.
Pair-wise comparison reveals factor weights
and factor evaluations.
Ensure that comparisons are consistent.
* Saaty TL, 1980, The Analytic Hierarchy Process.

Example: To Buy a Computer


Alternatives:
System 1
System 2
System 3

Selection Criteria:
Hardware
Software
Vendor

1. Decision Hierarchy

2. Pair-wise Comparison
Intensity of
Importance
1
3

5
7
9
2, 4, 6, 8
Reciprocals of
above nonzero

Definition

Explanation

Equally preferred

Two activities contribute equally to the


objective
Moderately Preferred
Experience and judgment slightly favour
one activity over another
Strongly preferred
Experience and judgment strongly or
essentially favour one activity over
another
Very strongly preferred
An activity is strongly favoured and its
dominance demonstrated in practice
Extremely preferred
The evidence favouring one activity over
another is of the highest possible order
of affirmation
Intermediate values between the two When compromise is needed
adjacent judgments
If activity i has one of the above nonzero numbers assigned to it when
compared with activity j, then j has the reciprocal value when compared with i.

2. Pair-wise Comparison

2. Pair-wise Comparison
Consider the following criteria:
Hardware

Software

1) Compare Hardware to

Vendor

Software

Which is more important?


Say Software Slightly
Hardware

Software

2. Pair-wise Comparison
2. Compare Hardware to

Vendor

Which is more important?


Say Hardware moderately.
Hardware

3. Compare Software to

Vendor

Vendor

Which is more important?


Say Software strongly.
Software

Vendor

3. Ranking Table
Make a matrix from above comparisons.
If judgement value on left from 1, enter actual value.
If judgement value on right from 1, enter reciprocal
value.
Its matrix
algebra time!

Hardware

Software

Hardware

Vendor

Software

Vendor

Hardware Software
Hardware
Software
Vendor

Vendor

1/3

7
1

3. Ranking Table
To fill in the lower triangular matrix, use the
reciprocals of the upper diagonal.
aij = 1/aji
Hardware Software

Vendor

Hardware

1/3

Software

1/5

1/7

Vendor

3. Ranking Table
Add columns together
Hardware Software

Vendor

Hardware

0.3333

Software

Vendor

0.20

0.1429

TOTAL

4.20

1.4762

13

3. Ranking Table
Normalise the relative weight.
Divide each cell by the sum and add up
Hardware

Software

Vendor

Hardware

1/(4.20)

0.3333/(1.4762)

5/13

Software

3/(4.20)

1/(1.4762)

7/13

Vendor

0.20/(4.20)

0.1429/(1.4762)

1/13

TOTAL

4.20

1.4761

13

Hardware

Software

Vendor

Hardware

0.2381

0.2258

0.3846

Software

0.7143

0.6774

0.5385

Vendor

0.0476

0.0968

0.0769

TOTAL

4. Ranking Criteria
Find normalised Eigen vector by averaging
across rows.

1
X
3

0.2381

0.2258

0.3846

0.7143

0.6774

0.5385

0.0476

0.0968

0.0769

0.2828

0.6434
0.0738

Hardware
Software
Vendor

5. Consistency
Have we been consistent in our pair-wise
comparisons?
Consistency ratio is calculated from 4 values:
Consistency vector (CV)
Lambda ()
Random index (RI) (read from a table)
Consistency index (CI) (n = no. of variables)

Consistency Ratio (CR)


If CR < 0.10 then our comparisons were
consistent

CI =
CR =

n
n1
CI
RI

5. Consistency
Weighted sum vector
Multiply factor evaluation number (ranking no)
with column of 1st pair-wise comparison matrix.
(0.2828)(1)

(0.6434)(0.3333)

(0.0738)(5)

(0.2828)(3)

(0.6434)(1)

(0.0738)(7)

(0.2828)(0.20)

(0.6434)(0.1429)

(0.0738)(1)

0.8663
=

2.0084
0.2223

5. Consistency
Consistency vector.
Divide the weighted sum vector by factor
evaluation values determined previously.
0.8663/0.2828

CV =

2.0084/0.6434
0.2223/0.0738

3.0632
=

3.1215
3.0122

Lambda is the average of CV


=

3.0656

5. Consistency
Consistency index (CI)
CI =

n
n1

CI =

3.0656 3
31

RI

CI =

0.0328

0.00

0.58

0.90

CI
RI

1.12

1.24

0.0328
0.58

1.32

1.41

Consistency Ratio (CR)


CR =
CR =
CR =

0.0566

6. System Selection
Now follow the same procedure to measure
each selection criterion to each alternative
system.

6. System Selection
Compare each system for Hardware using
pair-wise comparison.
Hardware System 1

System 2

System 3

System 1

0.5

System 2

0.25

0.14286

System 3

6. System Selection
Follow the same procedure as before to
rate the various alternatives against the
Hardware criterion.
NORMALISED
Hardware System 1

System 2

System 3

Average

CV

System 1

0.308

0.333

0.304

0.315

3.0019

System 2

0.077

0.083

0.087

0.082

3.000

System 3

0.615

0.583

0.609

0.602

3.004

CI =

0.0013

CR =

0.0022

6. System Selection
Now repeat the same for the Software.
Software

System 1

System 2

System 3

System 1

0.5

System 2

System 3

0.1667

0.125

NORMALISED
Software

System 1

System 2

System 3

System 1

0.316

0.308

0.400

0.341

3.0200

System 2

0.632

0.615

0.533

0.593

3.0315

System 3

0.053

0.077

0.067

0.065

3.0034

CI =

0.0089

Average

CR =

CV

0.0162

6. System Selection
And the same for Vendor
Vendor

System 1

System 2

System 3

System 1

0.25

System 2

System 3

0.3333

0.1667

NORMALISED
Vendor

System 1

System 2

System 3

System 1

0.188

0.176

0.300

0.221

3.0399

System 2

0.750

0.706

0.600

0.658

3.1094

System 3

0.063

0.188

0.100

0.093

0.0131

CI =

0.0270

Average

CR =

CV

0.0472

7. Make The Final Selection


Now put it all together.
ALTERNATIVES RATING
Criteria

Weights

System 1

System 2

System 3

Hardware

0.2828

0.315

0.082

0.602

Software

0.6434

0.341

0.593

0.065

Vendor

0.0735

0.221

0.658

0.093

0.3247

0.4531

0.2189

Final Rating

And the
winner is

More About AHP


It allows multi criteria decision making.
Becomes difficult to use when number of criteria or
alternatives is high, e.g. higher than 7.

Allows for qualitative as well as quantitative


decision evaluation.
Applicable for group decision making
environments.
Hidden assumptions like consistency.
Can be cumbersome.
Adding or removing a criterion calls for new evaluation.
If consistency does not fit, repeat the evaluation.

Using Excel with AHP


First our Hierarchy Tree.
We need to rank 5 different projects with 6
selection criteria.

We do this by pair-wise comparison of the


relative importance of each selection
criterion.
Then we repeat the process by evaluating
each alternative against each criterion.
This will lead us to an objective final
ranking.

Build a Hierarchy

Weight Selection Criteria

Match With Options

Final Selection

Fasten Your Seatbelts


To err is human

to really mess things up


takes a computer!

Opening Screen
Enter project names
here

Select the Criteria


Evaluation Spreadsheet

Enter selection criteria

Pair-wise comparison
So much to see, but so little to do
Carried
Pair-wise
over from the
work comparisons
area to the matrix

Normalised weights
carried over to FINAL
spreadsheet

The Work Area


Enter pair-wise
comparison numbers
in this area. All other
calculations are done
based on information
entered here.

Consistency Ratio is calculated


automatically. If this ratio is less than
0.1 it will turn green, else it will be red
to indicate that the pair-wise
comparisons are inconsistent and the
results therefore unreliable.

Pair-wise Comparison

Consistency Ratio will


signal green if less
than 0.10

Criteria are ranked

Final Criteria
from highest to lowest.
weights

Criteria Spreadsheets Are


Completed

Work area where pair-wise


comparison information is
entered.
Criterion that is being matched
with the various options

Consistency Ratio
must be green.

Other Spreadsheets Are


Completed

These 6 spreadsheets are


prepared for the evaluation of
each criterion with each possible
option and must all be completed
in the same way as this one.

Other calculations
are done as in
previous
spreadsheet

And The Final Outcome

Options are
Ranked
Highest
Consistency Ratio

Whats Next?
Apply the Analytical Hierarchy Process to
rank your shortlisted 5 projects.
Select the best three
Prepare a WOW presentation on your
favourites.
Develop a Business Plan with the
knowledge acquired.

References
Saaty, TL: The Analytic Hierarchy Process;
1980.
Render, B; Stair RM: Quantitative Analysis for
Management; 9th Ed; Allyn & Baker.
Triantaphyllou, E; Mann, SH: Using the
Analytic Hierarchy Process for Decision
Making in Engineering Applications: Some
Challenges; International Journal of Industrial
Engineering: Application and Practice; Vol. 2,
No. 1, pp 35-44; 1995

You might also like