Mathematical and Computer Modelling 46 (2007) 10411053

www.elsevier.com/locate/mcm

The analytic hierarchy process and human resource allocation: Half

the story
Thomas L. Saaty a, , Kirti Peniwati b , Jen S. Shang a
a University of Pittsburgh, Pittsburgh, PA 15260, USA
b PPM Graduate School of Management, Jakarta, Indonesia

Received 28 November 2006; accepted 14 March 2007

Abstract
The Analytic Hierarchy Process (AHP) provides a way to rank the alternatives of a problem by deriving priorities. A question
that occurs in practice is: what is the best combination of alternatives that has the largest sum of priorities and satisfies given
constraints? This leads one to consider the interface between the AHP and the combinatorial approach inherent in Linear
Programming (LP). The priorities of the alternatives often serve as coefficients of the objective function of an LP problem. The
constraints are determined from existing measurements, such as the range for the number of employees needed and the salaries
required for various jobs. Another way to use the AHP might be to determine the coefficients of the constraints. This paper
addresses the first half of the problem. Through various examples, we show how to apply the absolute measurement mode of the
AHP together with LP to optimize human resource allocation problems. For example, one can determine which positions to fill,
or which mix of candidates to hire. We also give an example of how to allocate resources to maximize the returns to a corporation
of its training programs. Finally, we show that the combined AHP and LP model is capable of solving hiring problems involving
synergy, such as when two persons with different complementary skills work as a team.
c 2007 Elsevier Ltd. All rights reserved.

Keywords: AHP; Linear Programming; Resource allocation

1. Introduction
One often wonders why more people in organizations do not rush today to use a formal decision-making approach
to make their complex decisions. A strange thing about people is that they value money and other valuable resources
over their own loosely defined and not well-organized subjective value systems. In addition, money is something they
have learned to count and do other operations on such as using it for barter, making exchanges or putting in and
taking out of a bank. They feel sophisticated as they do it, and they feel comfortable with what they are doing. People
seem know more about the value of money than about their inner selves. Because they exert effort to earn money,
they feel compelled to distribute and use their money in a somewhat rational way. To them, the effort spent to allocate
money to projects and other things needs some justification. This is probably why prioritization in resource allocation
Corresponding author.

E-mail addresses: saaty@katz.pitt.edu (T.L. Saaty), shang@katz.pitt.edu (J.S. Shang).

c 2007 Elsevier Ltd. All rights reserved.
0895-7177/$ - see front matter
doi:10.1016/j.mcm.2007.03.010

1042

T.L. Saaty et al. / Mathematical and Computer Modelling 46 (2007) 10411053

is popular, particularly in business and government. Another area in resource allocation is to optimize the return on
investments made in hiring people. Because of the considerable subjectivity involved in hiring people, it is helpful to
have prioritization techniques to deal with intangibles and make the process more objective, Saaty [1], and Saaty et al.
[2].
This is an expository article about the uses of multi-criteria prioritization in resource allocation, and in particular
the use of absolute measurement in the optimal assignment of human resources. This paper places special emphasis on
the measurement of intangible criteria and on their incorporation into the allocation process by combining the Analytic
Hierarchy Process (AHP) and Linear Programming (LP) to rate and derive the best combination of people assigned
to jobs. Use of the AHP enables us to address the issue of synergy between people that affects their qualifications and
whether they are selected singly or as a group. In a comprehensive review of Human Resource Planning, James A.
Craft [3] speaks of Human Resource Planning as a process of moving an organization to its desired position, with the
right kind of people in the right job at the right time to maximize value creating activity to help employers effectively
meet human resource requirements. To do this, one needs to consider an organizations goals and objectives along
with its various resource constraints.
Much of the literature on Human Resource Planning deals with future thinking by placing emphasis on forecasting
needs [48]. The problem is one of how to implement the plan to satisfy the needs. One key implementation is that of
staffing the organization through selection and allocation when forecast needs have been established. How can this be
done in an optimal manner with consistency in achieving the organizational goals? The AHP offers new possibilities
for tackling this problem, because it is a multi-criteria methodology that deals with both tangibles and intangibles in an
integrated and comprehensive manner within a hierarchic structure that relates people to the criteria and to the goals of
the organization. As a mathematical procedure, it uses both available quantitative data and the judgments of decision
makers and experts to arrive at an overall optimum answer. Through the prioritization process of the AHP, one is able
to determine the relative contribution of each alternative to the goal of the organization, whether that alternative is
a job position or a particular individual with his or her qualifications. However, when a combination of positions or
of individuals that best satisfies the goal subject to resource constraints is required, one needs to use an optimization
procedure like LP. Our purpose here is to illustrate the use of absolute measurement in the AHP along with LP to
obtain optimum human resource allocation. LP is widely used in the literature and is more of a technical subject. The
problem is one of how to use the measurement derived from the AHP to formulate an LP model which is then solved
using a standard software program like LINDO or EXCELs Solver routine. In the following two sections, we present
practical examples to illustrate the proposed approach.
2. Solving human resource requirement problems
2.1. Human resource recruiting at BDS [9]
Biological Detection Systems (BDS), a small biotech start-up firm located north of Pittsburgh, Pennsylvania, is
attempting to position itself for future growth by expanding its employee base. It has identified areas in Marketing,
Manufacturing, and Research and Development which need increased manpower. BDS has $520,000 to invest in new
employees. Therefore, it wants to select those applicants which will provide the most benefits to the organization. For
proprietary reasons, we do not provide a detailed description of the companys operation. The companys customer
base was expanded from 75 users in its first year to over 500 through the third quarter of its third year, and BDS
needed a way to prioritize the job positions to be filled as the rapid pace of growth continued through the next two
years. The positions to be filled are listed and briefly described in Table 1.
The object of this example is to determine the optimal number of positions required to be filled by the company, and
to select the most qualified applicants for each position, who together would contribute the most to the achievement
of the following organizational objectives:

to increase its marketing effort in the clinical market areas;

to satisfy future demand for its products;
to better insure product quality;
to develop new products;
to manage/supervise employees.

T.L. Saaty et al. / Mathematical and Computer Modelling 46 (2007) 10411053

1043

Table 1
Personnel requirements
Position

Variable Salary & benefits

in $ (000)

Personnel
required

Department

Description

V.P. of Marketing

y1

120

Marketing

y2

50

Marketing

Customer service
representative
Shipping clerk
Lab technician

y3

40

13

Marketing

y4
y5

40
40

01
25

Marketing
R&D

Chemists
Biologists
Quality specialists

y6
y7
y8

50
45
70

01
12
01

Manufacturing
Manufacturing
Manufacturing

Operations supervisor

y9

90

Manufacturing

Take BDS products into use in the clinical

or real world applications
Maintain customer database, organize
mailing pieces and advertisements.
Take orders, answer phones, first line of
customer support.
Package and ship products to customers.
Carry out experiments as specified by the
lab supervisor.
Manufacture dyes.
Puts dyes on the biologicals.
Make sure products meet government
regulations. Calibrate equipment.
Schedule the production of reagents
according to customer needs. Oversee
operations.

Marketing Assistant

2.1.1. The evaluation hierarchy

When evaluating candidates, BDS is confronted with a large number of tradeoffs over a diverse set of criteria. All
the applicants for a single position will be rated and evaluated based on the same set of criteria for that position and
priorities derived for each individual. Fig. 1 shows the AHP hierarchy for BDS personnel selection problem. The goal
of the hierarchy is to allocate the most qualified candidates to the various departments. The second level represents
the five objectives of the company, followed by the three departments for which applicants will be hired in the third
level. Level four identifies the positions that need to be filled within each department. The criteria for evaluating the
candidates for each position are located in the fifth level. For example, the VP of Marketing will be evaluated based
on experience, contacts, synergy, education, and leadership. Experience will be measured in the number of years
of industry experience the person has; contacts is used to evaluate the number of contacts the candidate has in the
clinical health care industry; synergy attempts to determine how well the individual will fit within the organization;
education is used to evaluate the candidates formal education; and leadership attempts to measure a candidates
general leadership skills.
To implement the absolute measurement mode in AHP, each criterion is divided into several intensity ranges to
differentiate the qualifications of the candidates with respect to that criterion. These intensities should be located
in level six of Fig. 1, and a list of some of them can be found in Table 2. For example, the evaluation criteria for
VP Marketing have the following intensities: (i) experience is divided into years and quality and years includes
three intensities of high (corresponds to 15+ years of experience), medium (615 years), and low (5 years or less);
(ii) quality, contacts and leadership are also divided into high, medium, and low; (iii) synergy into extreme, high,
medium, and low with extreme rating awarded to individuals who are judged to be a perfect fit with BDS; and finally
(iv) education is divided into PhD, masters, bachelors, and secondary. The complete selection criteria and their
intensities for each job together with their relative importance are listed in Table 3. The priorities of the intensities are
derived by pairwise comparisons and idealized by dividing each by the largest so that the largest becomes one and the
others proportionately less.
2.1.2. Intensities, criteria weights, and the composite scores
The nine positions are prioritized in terms of their contribution to the organizational objectives; four under
marketing, four under manufacturing, and one under R&D. The priorities of each are multiplied by the relative number
of positions in its group to the total of 9 and the result is then normalized. The criteria for evaluation are then prioritized
with respect to each of the positions, and the criteria intensities (not shown in the hierarchy) are prioritized with respect
to their corresponding criterion. The applicants are then evaluated by assigning the appropriate intensity according to

1044

T.L. Saaty et al. / Mathematical and Computer Modelling 46 (2007) 10411053

Fig. 1. Selection hierarchy for BDS with local (global) priorities.

their qualifications. Since each applicant will be evaluated for a single position only, he or she will be rated according
to the criteria for that position. His or her rating for the criteria for the other positions will be zero.
Each applicant receives a score which is the total of the weights of the corresponding intensities (obtained by
multiplying the priority of each intensity by the global priority of its corresponding criterion) assigned to him or her
on each criterion. The higher the score obtained using the rating process, the more qualified the applicant is for the
job. Table 4 shows the ranks of the applicants according to their scores.
Each employees total value in Table 4 has been adjusted to account for the differences in the number of elements
in the departments. Note that there is only one position for the R&D Department and four positions for the Marketing
and Manufacturing Departments. To account for this, the lab technicians final ratings scores were multiplied by 1/9,
and for the other three positions in Marketing and Manufacturing, the final scores were multiplied by 4/9.
2.1.3. Manpower allocation for BDS
In the following, we present two different, but comparable linear programming models to help make the best human
resource allocation decisions for BDS.
Model 1: Optimizing the individual applicants
The coefficients of the objective function below are the priorities of the individuals given in Table 4. The decision
variables are (0, 1), subject to a salary constraint, and the upper and lower bound constraints on the number of people
and the salary constraint given in Table 1. The binary integer LP problem is formulated as follows.
Maximize:
.0684x1 + .0489x2 + .0440x3 + .0360x4 + .0196x5 + .0200x6 + .0049x7 + .0067x8 + .0053x9
+ .0084x10 + .0040x11 + .0031x12 + .0053x13 + .0193x14 + .0140x15 + .0117x16 + .0218x17

1045

T.L. Saaty et al. / Mathematical and Computer Modelling 46 (2007) 10411053

Table 2
Rating applicants for each position on the applicable criteria
Applicants

James Plowski (x1 )

Marsha Macon (x2 )
Nancy Billings (x3 )
Stanley Kent (x4 )

VP Marketing
Experience
Years
.0174

Contacts

Synergy

Quality
.0239

.0852

High
High
Low
Medium

High
Low
High
Medium

High
Medium
Medium
Low

Applicants

Melvin Harper (x14 )

Elizabeth Ott (x15 )
Heather Fink (x16 )
Synder Smith (x17 )
Symore Drone (x18 )
Raymond Garner (x19 )
Mary Cunningham (x20 )
Applicants

Sammy Morgan (x28 )

Prashant Gupta (x29 )

Education

Leadership

.0316

.0103

.0431

Medium
High
Low
Extreme

BS
PhD
MS/MBA
MS/MBA

Medium
Hgh
Medium
Medium

Lab Technician
Education
.0348

Experience
.0418

Technical
.1566

Dependability
.0770

High
Medium
High
Medium
Medium
Low
High

Medium
High
Medium
Low
Low
High
High

Medium
Medium
Medium
High
Low
Low
Medium

High
Medium
Low
Medium
High
Low
Medium

Quality specialists
Education
.0265

Experience
.0378

Technical
.0448

BS
Masters

Medium
Medium

Medium
High

+ .0120x18 + .0076x19 + .0167x20 + .0044x21 + .0058x22 + .0102x23 + .0044x24 + .0120x25

+ .0129x26 + .0164x27 + .0191x28 + .0369x29 + .0271x30 + .0436x31 .
Subject to the salary constraint:
120x1 + 120x2 + 120x3 + 120x4 + 50x5 + 50x6 + 40x7 + 40x8 + 40x9 + 40x10 + 40x11 + 40x12
+ 40x13 + 40x14 + 40x15 + 40x16 + 40x17 + 40x18 + 40x19 + 40x20 + 50x21 + 50x22 + 50x23 + 45x24
+ 45x25 + 45x26 + 45x27 + 70x28 + 70x29 + 90x30 + 90x31 520.
The constraints on the number of people in each position are:
x1 + x2 + x3 + x4 = 1
x5 + x6 = 1
1 x7 + + x11 3
0 x12 + x13 1
2 x14 + + x20 5
0 x21 + x22 + x23 1
1 x24 + + x27 2
0 x28 + x29 1
x30 + x31 = 1

Vice President of Marketing

Marketing Assistant
Customer Service Representative
Shipping Clerk
Lab Technician
Chemists
Biologists
Quality Specialists
Operations Supervisor

Model 1 was solved using the Solver optimization routine in Excel. The names of the employees that are to be
hired to maximize the goals of BDS are given in Table 5.
Model 2: Optimizing the positions
In the second approach, we use an objective function whose coefficients are the priorities of the nine positions
given in the fourth level of the hierarchy of Fig. 1. The positions are denoted as Y1 Y9 , whose variables are integers
representing the number of jobs. This model determines the optimal number of jobs, then selects the best applicants for
those positions. The previous model simply did the selection based on the rating of the applicants, taking the relative

1046

T.L. Saaty et al. / Mathematical and Computer Modelling 46 (2007) 10411053

Table 3
Ideal priorities of the intensities used for rating alternatives
Criteria by job (global priority)

Intensities and idealized priorities

VP of Marketing
Experience (.041)
Years (.0174)
Quality (.0239)
Contacts (.0852)
Synergy (.0316)
Education (.0103)
Leadership (.0431)

High (1) Medium (.55) Low (.101)

High (1) Medium (.382) Low (.145)
High (1) Medium (.315) Low (.099)
Extreme (1) High (.561) Medium (.219) Low (.092)
PHD (1) Masters (.912) BS (.647) Second (.288)
High (1) Medium (.315) Low (.099)

Marketing Assistant
Organization (.0305)
Computer (.0113)
Education (.0054)
Synergy (.0333)

High (1) Medium (.315) Low (.099)

High (1) Medium (.315) Low (.099)
Masters (.25) BS (1) Second (.501)
Extreme (1) High (.561) Medium (.219) Low (.092)

Customer Service Representative

Synergy (.0058)
Computer (.0064)
Communication (.0127)
Secretarial (.0023)
Education (.0012)

Extreme (1) High (.561) Medium (.219) Low (.092)

High (1) Medium (.315) Low (.099)
High (1) Medium (.315) Low (.099)
High (1) Medium (.315) Low (.099)
Masters (.381) BS (1) Second (.876)

Shipping Clerk
Dependability (.0048)
Organization (.0069)
Computer (.0008)

High (1) Medium (.315) Low (.099)

High (1) Medium (.315) Low (.099)
High (1) Medium (.315) Low (.099)

Lab Technician
Education (.0348)
Experience (.0418)
Technical (.1566)
Dependability (.0770)

High (1) Medium (.315) Low (.099)

High (1) Medium (.315) Low (.099)
High (1) Medium (.315) Low (.099)
High (1) Medium (.315) Low (.099)

Chemists
Education (.0139)
Experience (.0069)
Technical (.0069)

PHD (.19) Masters (.252) BS (1) Second (.171)

High (1) Medium (.315) Low (.099)
High (1) Medium (.315) Low (.099)

Biologists
Education (.0223)
Experience (.0111)
Technical (.0111)

PHD (.19) Masters (.252) BS (1) Second (.171)

High (1) Medium (.315) Low (.099)
High (1) Medium (.315) Low (.099)

Quality Specialists
Education (.0265)
Experience (.0378)
Technical (.0448)

PHD (.251) Masters (1) BS (.653) Second (.15)

High (1) Medium (.315) Low (.099)
High (1) Medium (.315) Low (.099)

Supervisor
Education (.0094)
Synergy (.0198)
Managerial (.1046)
Experience (.0418)

PHD (1) Masters (.389) BS (.189) Second (.062)

Extreme (1) High (.561) Medium (.219) Low (.092)
High (1) Medium (.315) Low (.099)
High (1) Medium (.315) Low (.099)

importance of the positions into consideration. The salary constraint and the upper and lower bounds on personnel
requirements are the same as before. When the exercise is completed, the applicants with the highest scores in their
category are chosen. The solution is obtained by linear programming as shown below.
Maximize:
.2822y1 + .1076y2 + .0382y3 + .0000y4 + .0689y5 + .0000y6 + .0556y7 + .1360y8 + .2187y9 .

T.L. Saaty et al. / Mathematical and Computer Modelling 46 (2007) 10411053

1047

Table 4
Applicants ranked by AHP ratings score
Name

Position

Score

James Plowski (x1 )

Marsha Macon (x2 )
Nancy Billings (x3 )
Hana Coors (x31 )
Prashant Gupta (x29 )
Stanley Kent (x4 )
Henry Presley (x30 )
Synder Smith (x17 )
Tom Storey (x6 )
Jane Capelli (x5 )
Melvin Harper (x14 )
Sammy Morgan (x28 )
Mary Cunningham (x20 )
Kim McDonald (x27 )
Elizabeth Ott (x15 )
Megan Rice (x26 )
Karen Kerns (x25 )
Symore Drone (x18 )
Heather Fink (x16 )
Beth Foster (x23 )
Tony Austin (x10 )
Raymond Garner (x19 )
Ed Flanders (x8 )
Willie Whiteshoe (x22 )
Jim Bostead (x9 )
Terri Moore (x13 )
John Jones (x7 )
Brian Buzelli (x21 )
Teddy Star (x24 )
Charles Young (x11 )
Biff Van Swan (x12 )

VP of Marketing
VP of Marketing
VP of Marketing
Operations Supervisor
Quality Specialist
VP of Marketing
Operations Supervisor
Lab Technician
Marketing Assistant
Marketing Assistant
Lab Technician
Quality Specialist
Lab Technician
Biologist
Lab Technician
Biologist
Biologist
Lab Technician
Lab Technician
Chemist
Customer Service Rep.
Lab Technician
Customer Service Rep.
Chemist
Shipping Clerk
Customer Service Rep.
Customer Service Rep.
Biologist
Chemist
Customer Service Rep.
Shipping Clerk

0.0684
0.0489
0.0440
0.0436
0.0369
0.0360
0.0271
0.0218
0.0200
0.0196
0.0193
0.0191
0.0167
0.0164
0.0140
0.0129
0.0120
0.0120
0.0117
0.0102
0.0084
0.0076
0.0067
0.0058
0.0053
0.0053
0.0049
0.0044
0.0044
0.0040
0.0031

Table 5
The optimal solution for Model 1
Name

Position

Salary

James Plowski (x1 )

Tom Storey (x6 )
Tony Austin (x10 )
Melvin Harper (x14 )
Synder Smith (x17 )
Kim McDonald (x27 )
Prashant Gupta (x29 )
Hana Coors (x31 )

VP Marketing
Marketing Assistant
Customer Service Rep.
Lab Technician
Lab Technician
Biologist
Quality Specialist
Operations Supervisor

$120,000
$50,000
$40,000
$40,000
$40,000
$45,000
$70,000
$90,000

Total salary

$495,000

The coefficients are obtained from the fourth level of the hierarchy (Fig. 1) adjusted by multiplying the Marketing
positions by 4/9, the R&D position by 1/9, the Manufacturing positions by 4/9, and then the results are normalized.
We have the salary constraint:
120y1 + 50y2 + 40y3 + 40y4 + 40y5 + 50y6 + 45y7 + 70y8 + 90y9 520.
We have the constraints for the number of people in each department:
y1 = 1,

y2 = 1,

1 y7 2,

1 y3 3,

0 y8 1,

0 y4 1,

y9 = 1,

2 y5 5,

the yi are integers,

0 y6 1

and i = 19.

1048

T.L. Saaty et al. / Mathematical and Computer Modelling 46 (2007) 10411053

Table 6
The optimal solution of the second model using linear programming
Position

Selection

Position salary

Total salary

VP Marketing (y1 )
Marketing Assistant (y2 )
Customer Service Rep. (y3 )
Shipping Clerk (y4 )
Lab Technician (y5 )
Chemists (y6 )
Biologists (y7 )
Quality Specialists (y8 )
Operations Supervisor (y9 )

1
1
1
0
2
0
1
1
1

$120,000
$50,000
$40,000
$40,000
$40,000
$50,000
$45,000
$70,000
$90,000

$120,000
$50,000
$40,000
$0
$80,000
$0
$45,000
$70,000
$90,000

Total salary

$495,000

The second approach produces the following LP solution. It is consistent with that of the first approach in terms
of the number of positions filled. The solutions are given in Table 6. Note that linear programming is needed to find
the optimum solution in this problem because BDS has allotted only $520,000 to spend on salary and benefits for
additional employees, but to fill all the positions would require $830,000.
2.2. CNPITC employee training program [10]
China National Publishing Industry Trading Corporation (CNPITC) is a diversified state-owned multi-billionChinese-Yuan trading company located in Beijing. The Department of Publications in charge of the most important
and the most profitable business activity of the company has clients in Taiwan, Hong Kong, Singapore, Japan, and other
Asian countries. However, the business has not been growing for the last few years because of internal competition,
saturated markets, termination of funding, and lack of flexibility in human resource hiring, firing, and compensating.
The department foresees a potential market in western countries, but it is not well known internationally, and only
a few of its employees have been exposed to western culture or speak other languages. Only twenty out of its 90
full-time employees are college graduates, and most of them are new employees.
The department decided to carry out extensive training programs. There are 7 alternative training programs that
can be considered for each trainee:
1. An overseas 11 month MBA program with an estimated cost of $40,000 per trainee.
2. An in-house training program, which costs little and would have a considerable positive impact, but that lacks
exposure to an international environment.
3. College program in Beijing, which costs only about $1500 per student, but the quality of its business program is
not as good as that of an overseas MBA program.
4. Recruiting specialists, which would cost about $10,000 per person.
5. Combine programs 2 and 3: in-house training and college program.
6. Combine programs 2 and 4: Recruiting specialists and providing them with in-house training.
7. Combine programs 3 and 4: Recruiting specialists and providing them with college education.
The question is, which programs to choose, and how many people for each program? The training programs are
evaluated using absolute measurement, based on their benefits (Fig. 2) and costs (Fig. 3), for the organization as a
whole. The results of the analysis are given in Table 7.
The benefits priorities in Table 7 were obtained from the hierarchical structure shown in Fig. 2, and the costs
priorities were obtained from Fig. 3. The constraints must satisfy the following assumptions:
a. The department cannot have more than 40 employees in training in the same year.
b. The total funds available are $180,000 per year. The overseas training program must not cost more than $160,000
per year. The cost for the in-house training program is assumed to be negligible.
c. In-house training programs must have more than 10 and less than 30 participants.
d. At least 2 but no more than 16 trainees may receive the college education. At least 2 people are chosen for each
basic program. At least two trainees are sent for overseas training, and no more than 10 specialists are hired.

1049

T.L. Saaty et al. / Mathematical and Computer Modelling 46 (2007) 10411053

Fig. 2. The benefits hierarchy.

Fig. 3. The costs hierarchy.

Table 7
Benefits/Costs (B/C) ratios are used as the coefficients of the objective function
Variables

Alternatives

$ Amount needed

Benefits score (B)

Costs (C)

B/C

x1
x2
x3
x4
x5
x6
x7

Overseas training (1)

In-house training (2)
College program (3)
Recruit specialist (4)
(2) + (3)
(2) + (4)
(3) + (4)

$40,000
$0
$1,500
$10,000
$1,500
$10,000
$11,500

0.402
0.385
0.224
0.278
0.446
0.635
0.299

0.674
0.144
0.273
0.156
0.283
0.308
0.305

0.60
2.67
0.82
1.78
1.58
2.06
0.98

1050

T.L. Saaty et al. / Mathematical and Computer Modelling 46 (2007) 10411053

e. A trainee in the overseas training program cannot participate in another program, and no trainee can take more than
two courses of training in the same year.
The sets of criteria in Fig. 2 are as follows:
The Criteria 1 elements include Culture, Productivity, International Contact, Domestic Contact, Quality, and
Knowledge.
The Criteria 2 elements include all those in Criteria 1 plus Personal Gains
The Criteria 3 elements includes all those in Criteria 1 plus contacts (both International and Domestic).
The LP formulation is as follows:
Maximize:
.60x1 + 2.67x2 + .82x3 + 1.78x4 + 1.58x5 + 2.06x6 + .98x7 .
Subject to:
40,000x1 + 1,500x3 + 10,000x4 + 1,500x5 + 10,000x6 + 11,500x7 180,000
x1 + x2 + x3 + x4 + x5 + x6 + x7 40
x1 2,

x1 4,

2 x4 + x6 + x7 10,

10 x2 + x5 + x6 30,

2 x3 + x5 + x7 16.

The solution is:

x1
x2
x3
x4
x5
x6
x7

Overseas training program

In-house training program
College program
Specialists recruited
(do not take any other program)
In-house and College programs
Specialists recruited and taking
in-house training program
Specialists recruited and taking
college program

2
30
0
6
0
0
2.

3. Solving problems with synergy conditions using AHP and LP combined

There are situations where interaction among cooperative individuals actually contributes more (or less) than the
sum of what several members do independently. A very simple and important human cooperation example is men and
women working as couples in raising a family. In business, teamwork of employees with different organizational and
technical skills is often required. In the political arena, a common reason why politicians cooperate is that it brings
synergy. For example, suppose that political candidates X and Y can each win 10 million votes. Now X and Y agree
to team up as president and vice-president, and the result is that together they win 25 million votes. Their synergy
would have produced five million more votes. Synergy arises when two persons with different complementary skills
cooperate. Positive human resource synergy takes the form of (1 + 1)  2.
In the following numerical examples, the coefficients of the objective function can be derived by the AHP to
include all the intangible (and tangible when desired) benefits, and sometimes the ratios of the benefits to the costs. In
the following examples, tangible costs in dollars are included as constraints in the optimization problem.
3.1. Negative synergy
Negative synergy occurs when the sum of the effects of the collaboration is less than that of the individual
components of the mix. It can be called the (1 + 1) < 2 effect. Individuals operating alone may each produce
two units of output. However, with negative synergy, the combination of efforts results in less output than would be
achieved if individuals work alone. Negative synergy can result from inefficient committees, groups lacking strategic
fit, and poorly coordinated joint efforts. In the following examples, Pi denotes the ith person or group, and xi denotes
the value as shown in Tables 8 and 9.

1051

T.L. Saaty et al. / Mathematical and Computer Modelling 46 (2007) 10411053

Table 8
An example of negative synergy

Benefits
Costs

P1
x1

P2
x2

P3
x3

P1 + P2
x4

Restriction

.5
3

.3
2

.2
3

.6
5

Normalized if desired
6

P1
x1

P2
x2

P3
x3

P1 + P2
x4

Restriction

.5
3

.3
2

.3
2

.9
5

Normalized if desired
6

Table 9
An example of positive synergy

Benefits
Costs

Here the problem in Table 8 may be formulated as follows:

Maximize: .5x1 + .3x2 + .2x3 + .6x4
subject to 3x1 + 2x2 + 3x3 + 5x4 6.
The mutually exclusivity constraints are:
x1 + x4 1,

x2 + x4 1.

The constraints to prevent double counting are:

x1 + x2 1
x1 , x2 , x3 , x4 {0, 1}.
We cannot have x1 and x2 together because the combined effect is replaced by x4 . Similarly, choosing x4 would
naturally exclude the choice of either x1 or x2 , since x4 includes both x1 and x2 . The maximum value of the objective
function is equal to 0.7, with x1 = 1 and x3 = 1.
3.2. Positive synergy
In human resource management, synergy requires tapping into the human potential at its deepest level. An
essential precondition is a deep commitment to cooperation and integration. Positive synergy is an expected feature of
cooperating groups, and is a negotiating point between the acquirer and the acquired entities. An example of positive
synergy is given in Table 9.
The solution to the problem is arrived at as follows.
Maximize .5x1 + .3x2 + .3x3 + .9x4
Subject to 3x1 + 2x2 + 2x3 + 5x4 6.
The mutual exclusivity constraints are:
x1 + x4 1,

x2 + x4 1.

The constraints to prevent double counting are:

x1 + x2 1
x1 , x2 , x3 , x4 {0, 1}.
The maximum value of the objective function is 0.9, with x4 = 1.

1052

T.L. Saaty et al. / Mathematical and Computer Modelling 46 (2007) 10411053

Table 10
An example of mixed synergy

Benefits
Cost 1
Cost 2

P1
x1

P2
x2

P3
x3

P4
x4

P5
x5

P1 + P2
x6

P2 + P3
x7

P1 + P2 + P3
x8

Restriction

.1
2
10

.2
4
3

.3
6
8

.4
8
7

.5
10
6

.5
6
19

.7
10
17

.8
12
27

Normalized if desired
25
35

3.3. Multiple synergy/restrictions

In the following example, there are positive and negative synergies, indicating that an organization may benefit
from some specific employees teamwork, or may be penalized by putting the wrong mix of workers in the same
group. An example of mixed synergy is given in Table 10.
In the following, x6 and x7 contain positive synergy, while x8 gives negative synergy. The problem can be solved
with the linear programming model shown below.
Maximize: .1x1 + .2x2 + .3x3 + .4x4 + .5x5 + .5x6 + .7x7 + .8x8
Subject to: 2x1 + 4x2 + 6x3 + 8x4 + 10x5 + 6x6 + 10x7 + 12x8 25
10x1 + 3x2 + 8x3 + 7x4 + 6x5 + 19x6 + 17x7 + 27x8 35.
The mutual exclusivity constraints are:
x1 + x6 + x8 1,

x2 + x6 + x7 + x8 1,

x3 + x7 + x8 1.

The constraints to prevent double counting are:

x1 + x2 1,
and xi {0, 1}

x2 + x3 1,

x1 + x2 + x3 1,

x6 + x8 1,

x7 + x8 1

for i = 17.

The maximum value of the objective function is equal to 1.4, with x4 = 1, x5 = 1, and x6 = 1.
In this section, we have shown how negative, positive, and multiple synergy problems can be solved with the
LP model. Coefficients of costs and/or benefits can be derived using AHP. The combined AHP and LP approach
discussed above can be easily modified to solve problems where people applying for a job are evaluated by including
such criteria as flexibility, growth, challenge, independence, and productivity.
4. Summary and conclusion
Linear programming is a useful optimization technique for solving allocation problems by taking tangible
measurements into consideration. Many real world problems, however, cannot be readily solved using traditional
LP, which requires measurable factors, because they often contain intangible variables that cannot be quantified.
Because AHP is a model for creating measures for intangibles, combining the two makes it possible to deal with all
optimization problems whether the factors are tangible or intangible. This subject is investigated in detail in [2].
In Section 2, we showed through examples how to apply the absolute measurement of the AHP together with LP to
determine which positions to fill and which candidates to hire to satisfy the salary and employee number requirements
and constraints for each position. Section 3 illustrated how the AHP enables one to include synergy in its prioritization
to determine the best combination of people to hire. In this paper, we have shown that the combined AHP and LP
model provides an effective tool for resolving human resource allocation problems with their many subjective and
objective resource constraints, and a formal optimization technique is the best way to derive maximal benefits for the
investment organizations.
References
[1] T.L. Saaty, Decision Making with Dependence and Feedback: the Analytic Hierarchy Process, RWS Publications, 4922 Ellsworth Avenue,
Pittsburgh PA 15213, 1996.

T.L. Saaty et al. / Mathematical and Computer Modelling 46 (2007) 10411053

1053

[2] T.L. Saaty, L.G. Vargas, K. Dellmann, The allocation of intangible resources: the analytic hierarchy process and linear programming, SocioEconomic Planning Sciences 37 (2003) 169184.
[3] J.A. Craft, Human resource planning: its roots and development in management thought, Working Paper, Katz Graduate School of Business,
University of Pittsburgh, Pittsburgh, 1995.
[4] C.D. Fisher, et al., Human Resource Management, Houghton Mifflin Company, Boston, 1993.
[5] R.L. Mathis, J.H. Jackson, Human Resource Management, West Publishing Corp., Minneapolis/St. Paul, 1991.
[6] G.T. Milovich, J.W. Boudreau, Human Resource Management, Irwin Publishers, 1991.
[7] R.A. Noe, et al., Human Resource Management: Gaining a Competitive Advantage, Austen Press, Burr Ridge, IL, 1994.
[8] R.S. Shuler, V.L. Huber, Personnel and Human Resource Management, West Publishing Corp., New York, 1993.
[9] C. Gabler, D. Prado, A manpower allocation model for biological detection systems, BQOM 2410 Class Project, University of Pittsburgh,
1994.
[10] Y. Peng, Y.-K. Huang, Choosing the best employee training programs, BQOM 2410 Class Project, University of Pittsburgh, 1994.