1

Int. J Sup. Chain. Mgt Vol. 5, No. 1, March 2016

Factor Analysis and Methods of Supplier

Selection
Tak K. Mak#1, Fassil Nebebe*2
#
Department of Supply Chain and Business Technology Management, Concordia University
1455 De Maisonneuve West, Montreal, Canada, H3G 1M8
1
[email protected]
*
Department of Supply Chain and Business Technology Management, Concordia University
1455 De Maisonneuve West, Montreal, Canada, H3G 1M8
2
[email protected]

Abstract— We discuss in this paper the decision 1. Introduction

making in choosing the best alternative from some
available options based on possibly a large number of
selection criteria. This multi-criteria decision problem
Decision makers in many areas of business,
typically arises in supplier selection in supply chain administrative and social sciences are often charged
management. Recently, there has been an increasing with the responsibilities of selecting the best course
interest in the applications of dimensional reduction of action that will meet a number (and often large)
methods such as factor analysis to such decision of criteria. This is typically true in Supply Chain
processes. There are, however, a number of inherent management where supplier selection constitutes
issues and difficulties which have not been adequately one of the most important organization activities.
addressed in the literature. For instance, there may be
The selection of quality suppliers enables the firm
some criteria which load significantly on more than
to achieve an edge in a competitive business
one factor. More importantly, it is seen in this paper
that it is not always sensible to determine the
environment and is essential for the sustainability
importance of an identified factor according to its of a profitable business.
amount of shared common variance or explained In general, the process of selecting suppliers
variation. Similarly, attempts to routinely determine involves the conceptual approach of identifying the
the local relative weight (within a factor) of appropriate selection criteria against which the
importance of a criterion based on its factor loading potential suppliers are evaluated. In addition, the
or correlation with the factor may also lead to results relative importance of each identified criterion
in sharp contrast to those obtained from the
must be determined and is often reflected in a
experience of the practitioner. The present paper
certain numerical “weight”. Each potential supplier
gives a simple, practical and easily implemented
procedure to alleviate these difficulties. Although
is then evaluated by the decision maker using a
factor analysis is employed, it merely serves as a combined score based on the scores obtained on the
means of facilitating the direct rating of importance selected criteria weighted according to their relative
of each criterion and therefore does not experience importance. Recently, the use of dimensional
the same difficulties of the classical factor analysis reduction methods in supplier selection has
approach. Two examples are given to illustrate the garnered increasing attentions in the literature,
proposed method and illustrate some potential especially in categorizing the selection criteria and
problems of current approaches in the literature. The
weight assignments.
discussion in this paper will assist the practitioner in
While the use of dimensional reduction method is
applying appropriately factor analysis in the decision
procedure. potentially useful, its applications to supplier
selection are not without problems. For instance,
Keywords – AHP, criterion weights, Factor there may be some criteria which load significantly
analysis, multidimensional reduction, supplier on more than one factor so that categorizing the
selection criteria into mutually exclusive groups may not be
feasible. More importantly, as seen in Sections 3
and 5, it is not always sensible to determine the
______________________________________________________________
International Journal of Supply Chain Management importance of an identified factor according to its
IJSCM, ISSN: 2050-7399 (Online), 2051-3771 (Print) amount of shared common variance or explained
Copyright © ExcelingTech Pub, UK (http://excelingtech.co.uk/)
 2
Int. J Sup. Chain. Mgt Vol. 5, No. 1, March 2016

variation. Similarly, determining the local relative At a global level, industrial-wise determination of
weight (within a factor) of importance of a criterion the importance and usefulness of these criteria is in
based on its factor loading or correlation with the general based on surveys of a large number of
factor may also lead to results not necessarily in experts using typically dimensional reduction
line with those based on the experiences of the multivariate methods, such as factor analysis [18, 9,
practitioner or expert. These problems arise if the 10]. On the other hand, there is also an extensive
factor analysis is interpreted as a measurement literature on the methodologies for individual
model (of the underlying factors), as it will ignore managers on assessing suppliers based on a set of
the role of the specific factor (residual error) as identified criteria [3]. These methods include
measurement error when in fact the specific factor approaches that attempt to provide a simple,
may represent characteristics of the criterion that is manageable and systematic procedure for
also important for the decision making. In Section weighting the importance of the selection criteria.
3, we discussed the major issues in these methods These methods are summarized and reviewed in
and suggest a new approach that circumvents these [20]. [25] gives an account of recent research
difficulties. To demonstrate the uses of the activities in a wide variety of methodologies. One
proposed method in general multi-criteria problem, such methodology is the analytical hierarchical
we also discussed an example from recruitment process, which breaks down the weight
data to illustrate the application of the proposed assignments to two or more levels, with smaller
method. tasks involving fewer numbers of criteria at each
level. Recently, the uses of dimensional reduction
2. Literature Review methods, such as factor analysis to create a
hierarchical structure [18, 17, 21] have become
increasingly popular. Some researchers [18] use
Selecting the right supplier has direct influence on
further a second level of factor analysis treating the
operating cost as well as the quality of the product
constructs as “items” and the latent variable that
or services provided by the firm [6, 1, 13, 19]. [11]
explains the common variances as a final “score”
showed empirically the importance of supplier
for comparing suppliers.
selection in influencing business performance and
also identified the major impacting factors that 3. Methodology
contribute to such relationship. The work of [24]
and [12] also gave evidence the importance of Consider the general multi-criteria decision
supplier attitude and participation in building a problem involving the selection of an optimal
long term relationship that benefits both the solution from a number of alternatives based on a
supplier and the firm. Because of the importance of potentially large number of relevant criteria. The
supplier selection in supply chain management, determination of criteria for the selection process is
considerable effort has been expended in the assumed to have already been done based on the
research community to develop analytical methods views and judgement of a subject matter expert(s)
that could facilitate this multi-criteria decision and is not the focus of our attention. Instead, a
process [22]. Criteria for selecting suppliers were systematic approach for evaluating the available
discussed in [5, 15, 26, 16], among others. alternatives based on their scores on these criteria
Considerable emphasis is also placed in the will be proposed, harnessing both the power of
literature on categorizing selection criteria into quantitative analyses and past experience and
major factors or constructs [16, 23, 4, 8]. [27] uses judgement of the decision maker. The steps
multivariate methods to examine the relationships involved are:
among various selection criteria. Recently, there
have been considerable interests in extending the 1. Employing dimensional reduction method,
analysis of supplier selection to include additional typically exploratory factor analysis, to
criteria for supplier selection in green supply chain identify a number of constructs. Unlike
management [9, 10]. previous approaches (also see step 2 below)
that use several individual single factor
In general, selection of supplier is a multi-criteria models, multi-factors in an integrated model
decision problem and therefore appropriate (which allows a criterion to load on more than
weighting of the relative importance and relevance one factor) is permitted, broadening
of these criteria is central to the decision process.
 3
Int. J Sup. Chain. Mgt Vol. 5, No. 1, March 2016

substantially its scope of application and the four courses. Thus the emphasis is on the
improving its goodness of fit. selected student’s abilities in quantitative analysis
2. Factor rotation performed to obtain and mastering scientific concepts. The performance
interpretable factors which can later be used to score is purely a function of Factor 1 and Factor 2
check for consistence of importance values only. Conditional on the values of the two factors,
assigned to the selection criteria in step 3 the conditional distributions of the scores of the
below. four subjects carry no further relevant information.
3. With references to a conceptual “performance This is in line with the spirit of factor analysis as a
score function”, the decision maker assigns measurement model where the interest is in
important values to the selection criteria based measuring the underlying constructs and the
on the factors identified in steps 1 and 2, using residual (specific factor) in each of the equation in
a new, easily implemented procedure.
Table 1. Factor model of the correlations
We now elaborate the 3 steps for multi-criteria
decision and explain why some of the traditional Factor Factor Unique Commu-
Subject
methods have some potentially major issues that 1 2 -ness nalities
need to be carefully addressed. Many of the issues Math 0.00 0.95 0.0975 0.9025
involved can be illustrated using a simple, easily
understood example. The proposed method will Physics 0.88 0.40 0.0656 0.9344
then be applied to a supplier selection example
discussed in [14]. Consider the following Chemistry 0.85 0.20 0.2375 0.7625
hypothetical situation where scores on four subjects
- Mathematics, Physics, Chemistry and Biology,
are obtained for a sample of 100 students. Suppose Biology 0.50 0.10 0.7400 0.2600
that the following sample correlation matrix is
observed from the scores of the four subjects. Variation
1.75 1.11
explained

1.0000 0.3800 0.1900 0.0950

0.3800 1.0000 0.8280 0.4800


0.1900 0.8280 1.0000 0.4450
the factor analysis model is purely seen as a
0.0950 0.4800 0.4450 1.0000
measurement error that does not carry any further
information about the constructs studied. In this
An analysis of this correlation matrix results in a case, it is only necessary to assign relative
factor analysis model of two uncorrelated factors weighting to each of Factor 1 and Factor 2. In the
that can be used to reproduce the correlations second scenario, suppose the purpose is to award a
among the scores of the four subjects. The factor scholarship to the best-performing student. The
loadings, uniqueness and communalities of the performance score is directly a function of the
scores of the four subjects are given in Table 1. scores of the four subjects obtained by the student.
Factor 1 loads heavily on all of the scores of the In contrast to the previous scenario, weightings
four courses except Math and therefore could be should be assigned directly to the four subjects, and
labelled as “general ability in mastering scientific categorizing the subjects into factors is not
concepts’. On the other hand, Factor 2 loads necessary and may not be most appropriate. In
heavily on Math and could be seen as a student’s vendor selection, it is therefore generally not valid
“quantitative skill”. We now show that how the to measure the local importance (within a
evaluation, or the performance score, of each of the category/factor) of a criterion by measuring its
100 students should be determined based on the loading (correlation) with the factor [17] as this
factor model depends strongly on the purposes of would omit the useful information specific to this
the selection of the “best” student. In the first particular criterion. In addition, further difficulties
scenario, suppose that the purpose is to recruit the in using factor analysis for vendor selection or
most suitable student to be trained as an assistant general business decision problem may also arise
for a research project that requires quantitative when some criteria load on more than one factor as
statistical modeling of an environmental problem seen in the example in Section 4. Even if a factor
involving scientific concepts different from that of analysis approach is applicable, the importance of
the identified factors should not be based on their
 4
Int. J Sup. Chain. Mgt Vol. 5, No. 1, March 2016

correlations (loadings) with a higher level common Quartimax to obtain factors such that each  loads

1,...,a, let Sj be the collection of criteria  for

factor [18] as it may again ignore the useful on as few factors as possible. For the jth factor, j =
information in the specific factor leading to

factor  . The decision maker is asked to identify

conclusions far different from those based on the
which the highest factor loading occurs with the
practitioners’ judgement and experience, as seen in
the example in Section 5. The fact that identified
factors/ constructs that have little shared variances the most important and the least important criteria
only implies that they have characteristics distinct in each Sj. Thus there are 2a criteria selected. In the

importance (value of  ) to these 2a criteria. Instead

from other constructs, but they may still represent first stage of rating, the decision maker assigns
important decision criteria. In point of fact, the
identified constructs may be statistically of having these importance values summing to one,
independent (see Section 5) so that an analysis they are rescaled so that the highest importance is
based on common variances at the higher level equal to 1. Denote the set of ordered criteria
factor model may not even be possible. (ordered by their importance in descending order)
by C = {C1,...,C2a}. For each Sj, let C(j) be the
ordered subset{Cj1,... !" } of C where Cj1 and
Despite the shortcomings of the methods proposed

!" are respectively the most important and least

in the literature as pointed out above, dimensional
reduction method remains an interesting approach
that could greatly facilitate the criteria weighting important criteria in Sj. In the second stage of
assignments in a multi-criteria decision exercise. rating, each criterion in Sj, excluding those already
When properly applied, they can substantially in C, is compared sequentially with Cj1,... !"
reduce the burden of the decision maker in
starting with Cj1 to identify two neighbouring
criteria () and ()# such that the criterion
subjective weighting of a large number of criteria

 be the score on the ith criterion of an alternative

and yield results of greater internal consistence. Let

(vendor) and = (  , … ,  )′ where p is the

considered is either considered to be as important
() or less important than () but more
important than ()# . In the first case, the criterion
as
number of criteria. Traditionally, the final score

 , This amounts to () . In the second scenario, the

computed for each alternative or candidate is a considered will be given the same importance value

assuming that the performance function  =  

weighted average of the as that of

is linear in x, with the vector  = ( , … ,  )′

decision maker will assign an importance value that
() and ()# (or any
normalized so that the components of  sum to 1.
is the average of those of
intermediate value assigned by the decision maker).
The function p can in fact be rescaled in any These procedures will be illustrated with an
manner since the purpose is only to compare the application in the next section. As a by-product,
final scores of all the alternatives. We find it more this importance value assignment procedure also

largest  is equal to 1. The value  reflects the

convenient by scaling the function p so that the yields importance values for the identified factors.
To see this, observe that
importance or how much the ith criterion weights on
 =   =   (Λ + ) = $   + ′

$  = ($ , … , $ ) = ′Λ and $ reflects

the final performance score. Thus when we say a

one, we are merely saying that its  value is twice

certain criterion is twice as important as another

the impact or importance of the %&' factor in

where

that of the other. The decision maker contributes to

 . To facilitate this process,

determining the final performance score p. It is
the decision process by providing judgemental
noted that the present approach does not attempt to

suppose that the correlations among the  , can be

input values for the
assign importance weights based on the amount of
shared common variances, the potential problem of
explained by a factor model which has been pointed out above. Rather, the

= Λ + 
weights are assigned through the two stages of
rating based on the decision maker’s valuable past

where  = ( , … ,  )′ and a is the number of

experiences and knowledge.

factors. One can use a rotational method such as the

 5
Int. J Sup. Chain. Mgt Vol. 5, No. 1, March 2016

4. A human resource example X2: Appearance

X3: Academic ability X4: Likeability
To illustrate the methodology discussed in Section X5: Self-confidence X6: Lucidity
3, we analyze here the application discussed in X7: Honesty X8: Salesmanship
X9: Experience X10: Drive
[14]. [14] examined the data collected by the X11: Ambition X12: Grasp
Human Resources Department of a large firm X13: Potential X14: Keenness to join
responsible for selecting from 48 applicants a X15: Suitability
candidate to fill in a certain position. Part of the
selection process involves interviews of these The sample correlation matrix of these ratings is
candidates by the same panel of four executives. also reproduced in [2, P435]. The following steps
of analysis as suggested in Section 3 are performed.
The panel would rate each candidate on 15
attributes based on the candidate’s responses during Step1: Factor analysis using SPSS is conducted,
the interview. These 15 attributes are: yielding the results in Table 2:

X1: Form of an application letter

Table 2. Eigenvalue analysis.

To tal Var ian ce Explain ed

I nit ial Eigenv alues Ex traction Sum s of Squared Loadings Rotation Sum s of Squared Loadings
% of % of % of
Fac tor Total Variance Cumulative % T ot al Varia nce Cu mulativ e % Total Variance Cumulat iv e %
1 7. 504 50. 027 50.027 4.500 30.002 30. 002 5.542 36. 945 36.945
2 2. 061 13. 743 63.770 3.967 26.447 56. 449 2.477 16. 510 53.456
3 1. 468 9.785 73.554 1.615 10.764 67. 213 2.139 14. 263 67.719
4 1. 209 8.061 81.615 1.090 7. 267 74. 480 1.014 6.761 74.480
5 .741 4.943 86.558
6 .484 3.227 89.785
7 .344 2.294 92.079
8 .310 2.068 94.147
9 .260 1.731 95.878
10 .206 1.372 97.250
11 .151 1.006 98.256
12 9. 33E-02 .622 98.878
13 7. 63E-02 .509 99.386
14 5. 77E-02 .384 99.771
15 3. 44E-02 .229 100.00 0
Extraction Method: Maxim um Lik elihood.

Using Kaiser’s rule, four factors with eigenvalues factor. For instance, X14 loads on Factors 3 and 4.
greater than 1 are selected. The scree plot (Figure 1 Thus the present example does not meet the
below) also suggests the same thing. assumption in traditional approaches that each item
measures only a single factor. This assumption,
Scree Plot however, is not needed in our suggested method.
8 Factor 1 loads on X2, X5, X6, X8, X10 to X13
which [2, P434] labelled it as “extroverted
6 personality”. Factor 2 loads on X1, X9 and X15
and may be labelled as “suitability”. Factors 3 and
4 4 load on, respectively, X4, X7, X14 and X3, X14
and can be seen as what [2] called “agreeable
2
personality” and “academic” ability.
Eigenvalue

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Table 3. Final rotated factors of the hiring data
Component Number

Figure 1. Scree plot of hiring data

Step 2. The initial factor obtained in 1) is rotated to
obtain a more interpretable solution (Table 3). Note
that there are items that load on more than one
 6
Int. J Sup. Chain. Mgt Vol. 5, No. 1, March 2016

Rotated Factor Matrix a

The Decision maker feels that among candidates
with the same values of the factors, those with
Factor
higher salesmanship (higher value of the specific
1 2 3 4
X1 .126 .727 .113 -.114 factor corresponding to salesmanship) are most
X2 .454 .139 .243 .172 preferred compared with other items in S1. Thus X8
X3 7.10E-02 .126 -1.2E-03 .674 is identified as the most important item in S1.
X4 .227 .241 .829 -5.3E-02 Suppose that the most important and least
X5 .922 -9.9E-02 .146 -8.4E-02
X6 .842 .118 .289 5.40E-02
important items identified by the decision maker
X7 .248 -.228 .752 -1.7E-02 are as given in Table 4.
X8 .897 .236 7.52E-02 -6.7E-02
X9 9.16E-02 .767 -4.9E-02 .172 Table 4. Most and least important items in
X10 .762 .393 .183 -5.6E-02
each of S1 to S4.
X11 .898 .190 .110 -6.3E-02
X12 .782 .282 .364 .159 Set Most important Least important
X13 .724 .353 .449 .263
X14 .419 .393 .566 -.591 X8 X11
X15 .362 .766 5.03E-02 .136 S1
(salesmanship) (ambition)
Extraction Method: Maximum Likelihood.
Rotation Method: Varimax with Kaiser Normalization. X15 X1
a. Rotation converged in 5 iterations. S2
(suitability) (application letter)
Step 3: Importance value assignments. For each
X7 X4
item X1 to X15, we determine which of Sj it falls S3
(honesty) (likeability)
into by identifying the factor which it has the
highest loading with. For instance, X1 has the X3 X14
highest loading with Factor 2 and belongs therefore S4
(academic ability) (keenness to join)
in S2. Proceeding in this manner, we have the four
sets:
S1 = { X2, X5, X6, X8, X10, X11, X12, X13}
S2 = {X1, X9, X15}; S3 = {X4, X7}; S4 = {X3, Table 5 lists the eight items in Table 4 in
X14} descending orders along with the importance values
assigned subjectively.
Table 5. Assigned weights

item X15 X8 X7 X4 X3 X14 X11 X1

Importance
1 1 .25 .2 .17 .15 .1 .1
value

Consider now other items not in Tables 4 and 5. value of (.2 + .17)/2 = .185. The complete list of
Take for example, X2 (appearance) in S1. Then C(1) importance values are given in Table 6. Given the
= {X8, X7, X4, X3, X14, X11} and X2 is values of X1 to X15 for any given candidate, these
considered to be less important than X4 but more importance values can be used to compute the final
important than X3 and is assigned an importance score of that candidate.
Table 6. Assigned importance values of the 15 items
Item X1 X2 X3 X4 X5 X6 X7 X8
Importance
0.1 0.185 0.17 0.2 0.1 0.185 0.25 1.0
value
Item X9 X10 X11 X12 X13 X14 X15
Importance
0.8 0.185 0.1 0.125 0.1 0.15 1.0
value
 7
Int. J Sup. Chain. Mgt Vol. 5, No. 1, March 2016

$  = ′Λ, it is found that

$  = (2.26,1.96,0.78,0.27), or in terms of
Using the formula performed a factor analysis on the data and
successfully identified three statistically
independent factors which can be interpreted as F1:
relative weightings, 0.43, 0.37, 0.15, and 0.05 for
Product/service attributes, F2: Vendor attributes
respectively factors 1 to 4. Had the weightings been
and F3: Economic attributes. The eigenvalues
based on the common variances of the factors from
obtained are 3.76, 1.49 and 1.11 for respectively F1
Table 1, the relative weightings would have been
F2 and F3. Thus the relative weighting attached to
equal to 0.61, 0.17, 0.12, and 0.10.
each factor is traditionally calculated as
eigenvalue/(sum of eigenvalues), yielding weights

F3. For local relative weight * of the ith selection

5. A supplier selection example. of 0.591, 0.234, 0.175 for respectively F1 F2 and

its factor loading [17], it is calculated as + / ∑ +

[7] reported and analyzed data collected from a
criterion within the jth factor that is determined by
sample of subjects responding to a questionnaire
regarding their views about the importance of each

. = / * , where lij is the loading of the ith

of nine supplier selection criteria in selecting . The global relative weight is then found as

selection criterion for the jth factor and / is the

providers for part or the entire information system
for a company. These criteria were determined and
selected for study based on the authors experience,
relative weight of the jth factor. Results of the
practitioners’ views and review of the relevant
calculations based on Table V of [7] are given in

. and those from the study of [7]

literature. The importance as expressed by each
Table 7, along with the rankings of the criteria
respondent is given on a nine-point scale; for
details, See [7]. As part of the analysis, [7] based on the

Table 7. Rankings of criteria based on factor loadings and views of practitioners

Practitioners’ Global Local ranks

Selection criteria
global ranking ranks within factor
F1: Product/service attributes
Flexibility 1 4 (.115) 1
Ease of use 2 5 (.114) 2
Integration with existing
6 6 (.110) 3
applications
Integration with existing
7 7 (.103) 4
database
Efficiency 4 8 (.076) 5
Ease of installation 8 9 (.074) 6
F2: Vendor attributes
Vendor support 5 2 (.118) 1
Vendor viability/reliability 3 3 (.116) 2
F3: Economic attributes
Cost 9 1 (.175) 1

Two important points are in order. First, the three level of factor analysis model treating
identified factors are statistically independent so “appropriateness of supplier” as the single
that the relative importance of the three factors underlying factor (and factors identified in the low
cannot be determined by the amount of shared level model as the “variables” in the high level
common variances as in [18]. This is a good factor model). As explained in section 3, this is due
illustration of the potential problem of using high to the fact that the residual errors (specific factors)
 8
Int. J Sup. Chain. Mgt Vol. 2, No. 1, March 2013

also contain relevant information in addition to room for input from the decision maker based on
what is contained in the common factor. Second, it the person’s skill and past experience.

. ) by the entirely automated procedure

is seen that the rankings (based on the global
weights
based on factor analytic model are markedly References
different from those determined in a survey
incorporating the experiences and expertise of the [1] Banker, R.D. and Khosia, I.S., “Economics
practitioners interviewed. The criterion “cost”, of operations management: A research
perspective”, Journal of operations
while rated as the least important by the
management, 12, pp. 423 – 425, 1995.
practitioners, actually has the greatest weighting by [2] Berenson, M.L., Levine, D.M. and
the factor analysis approach. Even within the same Goldstein, M., Intermediate statistical
factor “Product/service attribute”, the criterion methods and applications, Prentice Hall:
“efficiency” is ranked third, but fifth by the factor Englewood Cliffs, 1983
analysis approach. Again, the factor analysis [3] Chen, Y.H. and Chao, R.J., “Supplier
approach only takes into account the amount of selection using consistent fuzzy preference
relations. Expert Systems with
shared variance component, but not information in
Applications”, 39, pp. 3233–3240, 2012
the specific factor (residual term) considered to be [4] Choi, T.Y. and Hartley, J.L., “ An
relevant in supplier selection by the practitioner. exploration of supplier selection practices
across the supply chain” Journal of
Operations Management, 14, pp. 333–343,
6. Conclusions 1996
[5] Dickson, G.W. , “An analysis of vendor
Factor analysis is a common approach for the selection systems and decisions”, Journal of
decision process in selecting suppliers. It is Purchasing, 2, pp. 5–17, 1966
[6] Flynn, B.B., Schroeder, R.G. and
particularly useful in creating a hierarchy of factors
Sakakibara, S., “A framework for quality
and subcriteria and thereby reducing the dimension management. Research and an associated
of pairwise comparisons. However, its uses are not measurement instrument”, Journal of
without restriction and its routine application is not operations management, 11, pp. 339 – 366,
recommended. An inherent difficulty that may arise 1994
in some applications stems from the possibility of [7] Gustin, C.M., Daugherty, P.J. and Ellinger,
cross loading with a criterion loads on more than A.E., “Supplier Selection Decisions In
Systems/Software Purchases”, International
one factor. The process of grouping or categorizing
Journal of Purchasing and Materials
the criteria requires that each criterion be included Management, 33, pp. 41-46, 1997
in one and only one factor. More importantly, [8] Hsu, C.C. Kannan, V.R., Leong, G.K. and
importance of the factors should not be based on Tan, K.C., “Supplier selection construct:
the associated explained variation. Furthermore, the instrument development and validation”,
weightings of the criteria should not be The International Journal of Logistics
proportional to their loadings as the residual term Management, 17, pp. 213 – 239, 2006
[9] Hsu, C.H., Wang, F.K. and, Tzeng, G.H.,
may also include important information for the
“The best vendor selection for conducting
decision process. We address in this paper these the recycled material based on a hybrid
issues and propose a simple, practical and easily MCDM model combining DANP with
implemented procedure that assigns importance VIKOR” Resources, Conservation and
value to each of the decision criteria directly, using Recycling, 66, pp. 95-111, 2012.
a factor analysis approach only as a means of [10] Hsu C.W., Kuo, T.C., Chen, S.H. and Hu,
A.H., “Using DEMATEL to develop a
facilitating the rating process. The input of the
carbon management model of supplier
practitioner is incorporated into the process which selection in green supply chain
is more preferable to automated procedures that management”, Journal of cleaner production
require little human intervention. The method , 56, pp. 164-172, 2013
alleviates some of the shortcomings of some [11] Kannan, V.R. and Tan, K.C., “Supplier
traditional approaches, requiring less rigid Selection and Assessment: Their Impact on
restriction for it to be operational. The proposed Business Performance”, Journal of supply
chain management, 38, pp. 11-21, 2002
procedure reduces the amount of pairwise
[12] Kannan, V.R. and Tan, K.C., “Attitudes of
comparisons of the criteria, but still leaves enough US and European managers to supplier
 9
Int. J Sup. Chain. Mgt Vol. 2, No. 1, March 2013

selection and assessment and implications art review”, Management Science Letters, 2,
for business performance”, Benchmarking: pp. 1465–1490, 2012
An international journal, 10, pp. 472-489, [26] Weber, C. A., Current, J. R. and Benton, W.
2003 C. “Vendor Selection Criteria and Methods”,
[13] Kannan, V.R. and Tan, K.C., “Just in time, European Journal of Operational Research,
total quality management, and supply chain 50, pp. 2-18, 1991
management: understanding their linkages [27] Wu, M.Y. and Weng, Y.C., “A study of
and impact on business performance”, supplier selection factors for high-tech
Omega, 33, 153–162, 2005 industries in the supply chain”, Total Quality
[14] Kendall, M., Multivariate analysis, London: Management & Business Excellence, 21, pp.
Charles Griffin, 1975
[15] Lehmann, D.R. and O'Shaughnessy, J.,
“Decision Criteria Used in Buying Different
Categories of Products”, Journal of
Purchasing and Materials Management, 18,
pp. 9-14, 1982
[16] Nydick, R. L. and Hill, R.P., “Using the
Analytic Hierarchy Process to Structure the
Supplier Selection Procedure”, International
Journal of Purchasing and Materials
Management, 28, pp. 31-33, 1992

[17] Petroni, A. and Braglia, M., “Vendor

selection using principal component
analysis”, Journal of supply chain
management, 5, pp. 65 – 69, 2000
[18] Punniyamoorthya, M., Mathiyalaganb, P.
and Parthibanc, P., “A strategic model using
structural equation modeling and fuzzy logic
in supplier selection”. Expert Systems with
Applications, 38, 458–474, 2011
[19] Shil, N.C., “Customized supplier selection
methodology: An application of multiple
regression analysis” Supply chain forum, 11,
pp. 38 – 69, 2010
[20] Singh, P.O. and Gupta, A. K., “Supplier
selection criteria and methods in supply
chain: A review” International scholarly and
scientific research and innovation, 7, pp.
1395 – 1401, 2013
[21] Sinrat, S. and Atthirawong, W., “Integrated
Factor Analysis and Fuzzy Analytic Network
Process (FANP) Model for Supplier
Selection Based on Supply Chain Risk
Factors” Research Journal of Business
Management, 9, pp.106-123, 2015
[22] Sonmez, M., A review and critique of
supplier selection process and practices.
Occasional Paper, 2006:1. Loughborough:
Business School, Loughborough, 2006
[23] Swift, C. O., “Preference for single sourcing
& supplier selection criteria”, Journal of
Business Research, 32, pp. 105–111, 1995
[24] Vonderembse, M. A. and Tracey, M., “The
impact of supplier selection criteria and
supplier involvement on manufacturing
performance”, Journal of Supply Chain
Management, 35, pp. 33-39. 1999
[25] Ware, N.R., Singh, S.P. and Banwet, D.K.,
“Supplier selection problem: A state-of-the-