Measurement 52 (2014) 135–144

Contents lists available at ScienceDirect

Measurement
journal homepage: www.elsevier.com/locate/measurement

The most cost efficient automotive vendor with price

uncertainty: A new DEA approach
Mehdi Toloo a,⇑, Tijen Ertay b
a
Faculty of Economics, Technical University of Ostrava, Ostrava, Czech Republic
b
Faculty of Management, Istanbul Technical University, Istanbul, Turkey

a r t i c l e i n f o a b s t r a c t

Article history: Vendor’s performance evaluation is an important subject which has strategic implications
Received 30 December 2013 for managing an efficient company. However, there are many important criteria for pros-
Received in revised form 20 February 2014 pering company. These criteria may contradict together. In other words, while a criterion
Accepted 4 March 2014
is improved, the other may worsen. Indeed, similar to manufacturing manager in global
Available online 14 March 2014
market, purchasing manager who has significant practical implications deals with this
issue. The vendor selection problem (VSP) is obviously affected by the complexity and
Keywords:
uncertainty due to the lack of information associated with related business environment
Data envelopment analysis
Price uncertainty
of countries in a global market. On the other hand, in the automotive industry which plays
Cost efficiency an important role in the worldwide market, these decisions will be exacerbated by increas-
Vendor selection ing the outsourcing and opportunities. There are varieties of techniques, from simple
Automotive industry weighted scoring methods to complex mathematical programming, for handling VSP.
In this study, we propose a new cost efficiency data envelopment analysis (CE–DEA)
approach with price uncertainty for finding the most cost efficient unit. Potential uses
are then illustrated with an application to automotive industry involving 73 vendors in
Turkey.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction needs to be carefully considered. A review of vendor

selection criteria can be found in Weber et al. [32].
Over the last two decades, purchasing from the complex This paper considers the assessment of cost efficiency
fabricated components with very high value-added con- (CE) using data envelopment analysis (DEA) model in the
tent and services to the high technological products indi- strategic vendor selection problem (VSP). The CE can be
cates that many companies tend to emphasis on strategic interpreted as the ability of each decision making unit
partnership and alliances. Nowadays, vendor selection (DMU) or vendor with multiple inputs and multiple out-
and evaluation not only are a key element in industrial puts to produce current outputs at minimal cost. Talluri
buying process but also are one of the major activities of and Narasimhan [20] proposed a maximin productivity
professional industry. Selecting the most efficient vendor, based approach that derives vendor performance variabil-
as a strategic partner, is a multiple criteria problem in ity measures, which are then utilized in a nonparametric
which both decreasing and increasing of each criterion statistical technique to identify vendor groups for effective
selection. Talluri et al. [21] suggested a chance-constraint
data envelopment analysis (CCDEA) approach for
⇑ Corresponding author. Address: VŠB-TU Ostrava, Faculty of Economics, vendor evaluation problem in the presence of multiple
Sokolská tř. 33, 701 21 Ostrava, Czech Republic. Tel.: +420 59732 2009. performance measures that are uncertain. The CCDEA
E-mail addresses: [email protected], [email protected] (M. Toloo). model originated by Land et al. [17] is a well-established

http://dx.doi.org/10.1016/j.measurement.2014.03.002
0263-2241/Ó 2014 Elsevier Ltd. All rights reserved.
136 M. Toloo, T. Ertay / Measurement 52 (2014) 135–144

approach and provides an innovative and simple way to method for ranking the discovered rules from data min-
incorporate variability in input and output measures ing. Toloo and Nalchigar [30] suggested a DEA approach
through the decision-making process. Wu and Olson [34] for supplier selection in presence of both cardinal and
compared stochastic DEA with a multiple-criteria model ordinal data. Asosheh et al. [3] combined the approach
in VSP, reporting simulation experiments varying the of Toloo and Nalchigar [30] with balanced scorecard for
degree of uncertainty involved in model parameters. The ranking information system projects. Toloo [24] expressed
stochastic dominance version of DEA can be applied if some drawbacks of previous studies and considered a
the uncertain data is normally distributed. Otherwise, it new MIP–DEA model to obtain the best BCC-efficient unit.
can be said that simulation modeling approach is more Toloo [27] formulated a model for finding the best
suitable. When the data is presented with uncertainty, efficient unit without explicit inputs and applied it to find
stochastic DEA provides a good tool for performing the the best professional tennis player. An epsilon-free
efficiency analysis by handling both inefficiency and approach for finding the most efficient unit is formulated
stochastic error. Next, Wu [33] developed an approach to in Toloo [25]. The suggested models exclude the non-
measure international supplier performance by consider- Archimedean infinitesimal epsilon, and consequently are
ing risk and uncertainty associated with supplier perfor- simpler, more reliable, more stable, more succinct and
mance on multiple measures in multiple categorical more practical than previous studies.
suppliers. His proposed model is an extension of both the Two pessimistic and optimistic approaches are consid-
classical stochastic DEA model and the bilateral systematic ered in Camanho and Dyson [7] for dealing with the CE
DEA model in Cooper et al. [12]. with price uncertainty. Based on this study, we develop
Azadi and Farzipoor Saen [4] proposed a new CCDEA two integrated mixed integer linear programming (MIP)
model with undesirable outputs for dealing with supplier models to find the most cost efficient DMU with price
section problem. Azadi et al. [6] and Azadi and Farzipoor uncertainty in pessimistic and optimistic situations.
Saen [5] continued this line of research. In Azadi et al. [6] The paper proceeds as follows. Section 2 considers
nondiscretionary factors are considered whereas Azadi backgrounds about DEA models used for estimating the
and Farzipoor Saen [5] developed a chance-constrained technical and cost efficiency based on Farrell concept.
free replicability hull model. Furthermore, Momeni and Section 3 develops our new approaches for finding the
Farzipoor Saen [18], proposed a new Russell CCDEA ap- most cost efficient DMU with price uncertainty. In
proach to assist the decision-makers for finding the most Section 4, potential uses are illustrated with applications
appropriate third-party reverse logistics providers in the to an automotive company located in Turkey. Finally, some
presence of multiple performance measures that are conclusions are drawn.
uncertain.
CE evaluates the ability of a DMU to produce the current 2. Preliminaries: data envelopment analysis
outputs at minimal cost. In this paper, we propose a new
approach for finding the most cost efficient DMU with DEA is a non-parametric method that utilizes linear
input prices uncertainty. The originality of this paper is programming (LP) techniques to empirically obtain the
that it is the first application of CE–DEA model in VSP. best production (efficient) frontier and evaluates the
The applicability of the suggested approach is illustrated efficiencies of a set of similar organizations. In DEA
by a real data set in an automotive company located in models, efficiency is measured as the weighted sum of
Turkey. outputs divided by the weighted sum of inputs. It can
From an economic point of view, the managers have be said that DEA generates a summary measure of
an interest in identifying the most cost efficient DMU. performance while considering multiple inputs and
Hitherto, there have been several attempts in DEA litera- multiple outputs measures. DEA is a non-parametric
ture for determining the most efficient DMU. Afterwards, technique in that the specific form of production func-
in order to increase the discriminating power of DEA, tion is not required to be initially known or assumed.
Ertay and Ruan [13] proposed an evaluation procedure Farrell [16] initially introduced a non-parametric
that aims to analyze ‘‘cross-efficiency’’ by the fact that approach to measure productive efficiency and then
true efficient candidate and false positive candidate are started from the observed input–output coefficients of a
discriminated. In order to discriminate between relatively set of firms, as a standard for measuring the efficiency
efficient DMUs, cross-efficiency has been used effectively of the firms, instead of estimating the conventional
to surmount the problems associated with simple effi- production functions. However, this study was limited
ciency scores. Ertay et al. [14] offered a minimax method to single input and single output. Charnes et al. [8]
consists of a parameter that should be selected on a trial presented a ratio definition of efficiency formulated as
and error method to reach the most efficient DMU. Amin a fractional form. This formulation has been expanded
and Toloo [1] formulated a new integrated DEA model for from single input and single output in classical ratio
finding the best CCR-efficient, however Toloo and definition to multiple inputs and multiple outputs.
Nalchigar [28] extended it to variable returns to scale Moreover, the following well-known CCR model was
(VRS) situation. Based on this methodology, Toloo et al. formulated, using the Charnes and Cooper [10]
[31] and Toloo and Nalchigar [29] stated a new DEA transformation:
 M. Toloo, T. Ertay / Measurement 52 (2014) 135–144 137

X
s
where v ai ðv bi Þ is the weight for input ia(ib) of DMUo;
max ur yro
pmin
ia
ðpmin
ib
Þ is minimum bound estimated for the price of
r¼1

s:t: input ia(ib) of DMUo and pmax

ia
ðpmax
ib
Þ is the maximum bound
Xm estimated for the price of input ia(ib) of DMUo.
v i xio ¼ 1 This model is formulated by imposing 2  C m 2 input
i¼1 ð1Þ weight restrictions on the standard DEA model (1). The
X
s X
m
bounds of the CE measure are obtained from assessments
ur yrj  v i xij  0 j ¼ 1; 2; . . . ; n in the light of the most favorable price scenario and the
r¼1 i¼1
least favorable price scenario.
ur  e r ¼ 1; . . . ; s
Note that changing the objective function of the model
v i  e i ¼ 1; . . . ; m (2) from a maximization to a minimization leads to approx-
imately zero CE for all DMUs assessed. Hence it is not too
where yrj is the amount of rth output of DMUj; xij the
easy to extend a model for measuring the pessimistic CE
amount of ith input of DMUj; ur the weight of rth output;
with input price uncertainty. However, Camanho and
vi the weight of ith input; n the number of DMUs; m the
Dyson [7] tackled this issue and extended the following
number of inputs; s the number of outputs; o the index
model:
of under evaluation DMU and e the non-Archimedean
infinitesimal. X
s

Firstly, the epsilon was proposed in Charnes et al. [9] to min ur yro
r¼1
obtain the positive optimal weights. A number of attempts
were made for determining an assurance value for the s:t:
epsilon (for more details see [2]). Note that optimal solu- Xm

tions of model (1) for extreme efficient DMUs are often

v i xio ¼ 1
i¼1
highly degenerate and, consequently, have alternative
Xs X
m
optimal solutions for the weights. To deal with this issue, ur yrp  v i xip ¼ 0 ð3Þ
Cooper et al. [11] proposed a two-stage procedure which r¼1 i¼1
explores these alternatives in order to help make a choice Xs X
m

of optimal weights. Toloo et al. [26] obtained a closed form ur yrj  v i xij  0 j ¼ 1; 2; . . . ; n
r¼1 i¼1
of an initial basic feasible solution for model (1) and prac-
tically investigated that starting from this solution, which pmin
ia v a pmaxa a b
 i  imin 1i <i m
removes the role of artificial variables, decreases at least pmax
i b v i b p b
i
50% of the whole computations. ur  e r ¼ 1; . . . ; s
CE can be expressed as a measure of the potential cost
reduction reachable in the outputs which produced with where the index p represents the peer DMU underlying the
the current input prices at each DMU. In other words, given optimistic CE assessment of DMUo which is already ob-
the input prices, CE evaluates the ability of a DMU to pro- tained by model (2). Fining the set of peer DMUs in this
duce current outputs at minimal cost. There are some DEA model can be considered as a debatable topic.
models dealing with CE analysis: Fare et al. [15] attempted Notwithstanding these approaches, we propose an
to suggest a new approach, meanwhile Thompson et al. approach for finding the most cost efficient DMU with a
[23], Schaffnit et al. [19] and Taylor et al. [22] utilized common set of weights (CSW). These weights help us to
assurance region (AR) method in multiplier forms of DEA identify the most efficient DMU in an identical condition.
models to develop new DEA/AR models. The aim of these On the other hand, to find a single cost efficient DMU, there
studies was to propose an optimistic DEA model for mea- is no need to solve one optimization problem for each
suring the CE, whereas Camanho and Dyson [7] enhanced DMU, rank all cost efficient DMUs (using one of the ranking
these methods to account for different scenarios relating approaches), and finally determine a DMU with the highest
to input price information. rank score. More importantly, the proposed optimistic
The following model measures the optimistic CE with and pessimistic approaches in this study can be applied
input price uncertainty: independently. These aspects can be considered as the main
X
s advantages of the proposed method in this paper over the
max ur yro previous studies. In the next section, based on two opti-
r¼1
mistic and pessimistic scenarios, we develop an integrated
s:t: approach to find the most cost efficient DMU with price
Xm uncertainty.
v i xio ¼ 1
i¼1
ð2Þ 3. The most cost efficient unit with price uncertainty
Xs X
m
ur yrj  v i xij  0 j ¼ 1; 2; . . . ; n
r¼1 i¼1 We first develop an approach to find the most optimis-
pmin
ia v a pmaxa a b
tic cost efficient DMU with price uncertainty and then deal
 i  imin 1i <i m with the most pessimistic cost efficient DMU. Consider the
max
pb
i
v i b p b
i
following integrated DEA model that seeks to minimize the
ur  e r ¼ 1; . . . ; s sum of deviation of all DMUs from the efficiency:
138 M. Toloo, T. Ertay / Measurement 52 (2014) 135–144

X
n 
Let Eopt ¼ fj : dj ¼ 0g. If Eopt is singleton and k e Eopt,
min dj
then the model (5) can determine DMUk as the most cost
j¼1
efficient unit under an optimistic perspective. Otherwise,
s:t: we propose the following MIP model for finding the most
Xm
cost efficient DMU:
v i xij  1 j ¼ 1; 2; . . . ; n
X
n
i¼1
ð4Þ min dj
Xs X
m
j¼1
ur yrj  v i xij þ dj ¼ 0 j ¼ 1; 2; . . . ; n
r¼1 i¼1 s:t:
dj  0 j ¼ 1; 2; . . . ; n Xm
v i xij  1 j ¼ 1; 2; . . . ; n
ur  e r ¼ 1; . . . ; s i¼1
v i  e i ¼ 1; . . . ; m Xs X
m
ur yrj  v i xij þ dj ¼ 0 j ¼ 1; 2; . . . ; n
where v = (v1, . . ., vm) and u = (u1, . . ., ur) are the CSW for in- r¼1 i¼1
puts and outputs, respectively, dj is the deviation of DMUj v i pmin
a  v i pmaxb 0 a
a
1i <i m
b
ð7Þ
b i
i
from efficiency and thus this unit is efficient (and is a can-
a b
didate for being the most efficient) DMU if and only if v i pmax
a
i b  v p
i i
min
b0 a 1i <i m

dj ¼ 0. X
n

We formulate the following integrated optimistic CE hj ¼ n  1

j¼1
model to determine the most cost efficient unit candi-
date(s) with CSW under an optimistic perspective: dj  Mhj j ¼ 1; 2; . . . ; n
X
n hj  Ndj j ¼ 1; 2; . . . ; n
min dj
hj 2 f0; 1g j ¼ 1; 2; . . . ; n
j¼1

ur  e r ¼ 1; 2; . . . ; s
s:t:
Xm where M and N are large enough positive numbers. In this
v i xij  1 j ¼ 1; 2; . . . ; n model there exists only one nonzero auxiliary binary
i¼1 variable, hj, and the imposed constraints imply that just
X
s X
m
ð5Þ one of the deviation variable, dj, has a value of zero. Note
ur yrj  v i xij þ dj ¼ 0 j ¼ 1; 2; . . . ; n that the model (7) can be utilized independent of the
r¼1 i¼1 
a b
model (5). e can be measured by imposing some suitable
v i pmin
a
i b v i pmax
i
0
b a 1i <i m variables and constraints into the model (6). Obviously,
v i pi  v i pi  0
a
max
b
min
b a
a
1i <i m
b 0 < e  e .
In a similar vein, we first formulate the following
dj  0 j ¼ 1; 2; . . . ; n integrated LP model for finding the most cost efficient
ur  e r ¼ 1; 2; . . . ; s candidate(s) with CSW under a pessimistic perspective:
a
where v ia is the common weight for input i and also pmin
ia
X
n

and pmax are minimum and maximum bound estimated max dj

ia
a j¼1
for the price of input i .
s:t:
Definition 1. DMUk is optimistic cost efficient with CSW Xm

and price uncertainty if and only if dk ¼ 0. v i xij  1 j ¼ 1; 2; . . . ; n
i¼1
The following LP model can be utilized for attaining an
Xs X
m
ð8Þ
assurance value of epsilon in the model (5): ur yrj  v i xij þ dj ¼ 0 j ¼ 1; 2; . . . ; n
r¼1 i¼1
e ¼ max e a b
s:t:
v i pmin
a
i b  v i pmax
i b 60 a 16i <i 6m
a b
Xm v i pi  v i pi P 0
a
max
b
min
b a 16i <i 6m
v i xij  1 j ¼ 1; 2; . . . ; n
i¼1
dj P 0 j ¼ 1; 2; . . . ; n

Xs X
m ur P e r ¼ 1; 2; . . . ; s
ur yrj  v i xij  0 j ¼ 1; 2; . . . ; n ð6Þ
In contrast with the proposed approach in Camanho
r¼1 i¼1
a b and Dyson [7], the pessimistic model (8) is obtained by
v min
ia pib v 0 pmax
ib i
a 1i <i m changing the objective function of the optimistic model
a b
v i pi  v i pi  0
a
max
b
min
b a 1i <i m (2) from a minimization to a maximization.
ur  e  0 r ¼ 1; 2; . . . ; s
Definition 2. DMUk is pessimistic cost efficient with CSW

It is easy to prove that the optimal objective value of and price uncertainty if and only if dk ¼ 0.
pes 
model (6) is bounded and hence model (5) is feasible for Let E ¼ fj : dj ¼ 0g. As will be seen shortly, practically
e e (0, e]. Nevertheless, to have the maximum discrimina- Epes might be singleton. In this case, model (8) can select
tion among cost efficient DMUs, we let e = e⁄. the most cost efficient unit with CSW under a pessimistic
Table 1
The data set for 73 vendors.

Vendors Inputs Outputs

x1 x2 x3 x4 x5 x6 x7 y1 y2 y3 y4 y5
Vendor1 1 9 6 4 30 24 52 640 97.04 1,910,370 TL 280,500 TL 1,003,590 TL
Vendor2 1 7 7 4 20 28 24 277 90.57 1,059,404 TL 872,136 TL 1,918,123 TL
Vendor3 1 13 6 3 21 30 32 455 89.23 1,008,395 TL 1,203,813 TL 659,653 TL
Vendor4 1 14 16 12 53 36 102 1294 89.49 3,133,756 TL 1,838,444 TL 13,603,758 TL
Vendor5 1 7 7 9 30 30 48 617 90 223,217 TL 1,026,029 TL 318,097 TL
Vendor6 1 13 15 8 56 32 121 1237 88.29 2,835,982 TL 531,900 TL 1,952,176 TL
Vendor7 2 21 15 19 48 44 155 1832 95.69 2,662,772 TL 8,583,540 TL 2,035,702 TL
Vendor8 1 10 12 6 28 27 62 711 95.06 1,466,014 TL 3,144,548 TL 8,675,483 TL
Vendor9 1 8 8 10 49 36 95 1308 94.35 3,370,414 TL 3,695,887 TL 14,623,374 TL
Vendor10 1 8 13 6 41 40 146 1500 96.84 2,769,254 TL 8,092,843 TL 1,744,480 TL
Vendor11 3 13 13 9 62 38 95 1167 91.1 2,903,007 TL 1,251,241 TL 2,554,249 TL
Vendor12 1 14 4 4 24 25 57 623 93.4 1,054,126 TL 3,016,386 TL 641,379 TL
Vendor13 3 13 3 10 58 34 52 674 85.9 4,099,307 TL 1,972,924 TL 7,118,558 TL
Vendor14 1 3 6 5 13 24 38 361 88.1 569,169 TL 1,405,713 TL 347,915 TL
Vendor15 2 11 4 4 24 24 36 412 88 1,009,351 TL 1,144,857 TL 858,399 TL
Vendor16 1 8 5 7 22 30 66 839 90.82 1,661,328 TL 3,018,817 TL 1,085,838 TL
Vendor17 1 5 1 4 21 24 32 140 91.67 800,394 TL 448,270 TL 575,468 TL
Vendor18 4 23 11 19 82 72 152 1978 91.67 4,831,579 TL 8,602,663 TL 16,859,827 TL
Vendor19 1 15 8 11 56 80 262 4270 90 4,148,022 TL 10,805,356 TL 2,488,767 TL
Vendor20 1 31 6 14 63 56 157 2260 90.52 4,316,224 TL 3,581,663 TL 16,333,754 TL
Vendor21 2 16 2 7 60 49 164 2577 92.14 4,939,996 TL 8,084,737 TL 3,314,862 TL
Vendor22 2 7 14 6 32 26 51 652 86.48 1,201,713 TL 793,400 TL 842,618 TL
Vendor23 1 12 6 13 52 50 135 2052 94.32 3,632,682 TL 2,175,415 TL 1,549,335 TL
Vendor24 1 13 9 5 37 34 79 1013 94.41 2,932,165 TL 2,693,782 TL 2,343,993 TL
Vendor25 3 7 4 7 58 36 94 1352 92.06 8,308,527 TL 3,752,045 TL 4,318,782 TL
Vendor26 1 18 2 7 30 13 28 41 65 18,798 TL 579,456 TL 147,215 TL
Vendor27 2 14 5 4 42 24 48 481 90.94 2,281,765 TL 2,954,104 TL 1,746,409 TL
Vendor28 2 17 8 12 39 27 69 967 94.9 3,116,901 TL 3,656,699 TL 3,427,893 TL
Vendor29 2 12 7 10 39 32 53 646 93.28 1,761,756 TL 2,028,868 TL 875,419 TL
Vendor30 3 23 25 13 107 58 176 2437 91.54 7,921,469 TL 3,145,388 TL 10,576,052 TL
M. Toloo, T. Ertay / Measurement 52 (2014) 135–144

Vendor31 2 10 7 4 27 25 27 321 92.61 1,077,919 TL 2,242,762 TL 1,361,882 TL

Vendor32 3 3 3 7 30 20 42 448 86.88 406,804 TL 2,537,174 TL 788,317 TL
Vendor33 1 6 5 4 22 10 28 130 90.97 736,004 TL 804,568 TL 443,961 TL
Vendor34 6 28 8 9 49 24 85 1171 86.76 2,635,455 TL 2,943,323 TL 2,283,596 TL
Vendor35 1 11 8 16 59 35 96 1141 91.95 3,210,906 TL 5,135,574 TL 19,608,883 TL
Vendor36 2 19 6 9 55 28 74 937 90.59 3,762,919 TL 1,853,635 TL 6,506,520 TL
Vendor37 3 13 2 5 41 30 63 745 89.1 1,511,385 TL 5,470,865 TL 1,163,909 TL
Vendor38 1 13 3 4 18 12 21 37 88 290,768 TL 553,214 TL 63,096 TL
Vendor39 1 9 7 9 28 10 50 198 92 37,106 TL 670,374 TL 252,852 TL
Vendor40 1 7 4 3 23 24 32 385 93.84 1,310,642 TL 1,697,151 TL 1,183,943 TL
Vendor41 1 17 7 7 54 33 68 1130 84.6 3,542,367 TL 2,309,716 TL 2,587,842 TL
Vendor42 2 8 5 5 35 27 46 854 96.35 1,641,654 TL 4,130,970 TL 933,502 TL
Vendor43 3 14 3 10 57 40 107 1321 89.39 3,136,621 TL 4,161,952 TL 5,582,650 TL
Vendor44 2 9 2 3 14 24 54 519 86.48 890,330 TL 2,658,355 TL 555,670 TL
Vendor45 1 4 4 4 30 16 52 331 93.75 1,220,491 TL 1,049,447 TL 2,533,058 TL
Vendor46 1 9 4 7 51 14 45 338 86.33 1,523,372 TL 1,586,915 TL 1,805,498 TL
Vendor47 1 15 6 7 40 32 85 1256 89.92 3,183,149 TL 2,656,590 TL 9,671,409 TL

(continued on next page)

139
Table 1 (continued)
140

Vendors Inputs Outputs

x1 x2 x3 x4 x5 x6 x7 y1 y2 y3 y4 y5
Vendor48 1 17 5 12 63 47 96 1774 92.42 3,010,272 TL 4,652,952 TL 13,493,640 TL
Vendor49 1 16 6 14 50 62 168 3370 90.1 4,172,978 TL 9,059,555 TL 22,050,080 TL
Vendor50 1 14 6 14 53 58 145 2611 91.83 2,757,365 TL 4,277,783 TL 13,246,972 TL
Vendor51 2 19 5 5 35 44 167 1112 93.21 1,888,575 TL 1,713,480 TL 2,131,156 TL
Vendor52 1 22 6 20 61 68 184 2307 93.1 5,128,151 TL 7,607,844 TL 21,397,558 TL
Vendor53 2 21 5 16 57 41 145 1287 91.23 5,758,179 TL 1,055,287 TL 34,259,621 TL
Vendor54 1 11 6 6 32 48 148 1310 93.86 2,660,087 TL 2,911,758 TL 2,338,698 TL
Vendor55 1 15 5 20 68 58 189 3689 90.24 4,044,013 TL 3,650,729 TL 29,173,672 TL
Vendor56 1 15 6 13 47 61 340 5923 91.41 5,777,861 TL 1,753,252 TL 4,033,602 TL
Vendor57 2 13 6 9 60 25 64 831 91.76 1,651,355 TL 1,937,189 TL 1,055,978 TL
Vendor58 1 5 13 3 30 21 49 664 90.71 1,078,255 TL 2,679,027 TL 749,988 TL
Vendor59 4 12 5 6 42 38 86 1092 95.74 2,555,576 TL 4,128,704 TL 1,612,117 TL
Vendor60 1 10 4 3 38 24 42 460 95.21 2,650,715 TL 1,043,431 TL 1,378,041 TL
Vendor61 1 16 4 21 75 47 211 2193 88.93 5,576,965 TL 13,186,358 TL 7,256,506 TL
Vendor62 1 12 6 8 37 24 59 868 94.72 2,161,652 TL 2,738,578 TL 918,075 TL
Vendor63 1 9 3 10 45 19 48 621 89.53 2,071,493 TL 2,335,022 TL 9,708,144 TL
Vendor64 1 5 5 5 16 11 36 288 89.41 1,056,523 TL 424,215 TL 4,447,047 TL
Vendor65 2 14 5 6 44 20 78 793 96.69 2,086,320 TL 2,088,380 TL 1,450,859 TL
Vendor66 3 10 8 6 47 24 115 859 93.42 2,511,645 TL 4,371,967 TL 1,585,457 TL
Vendor67 2 12 2 3 25 15 46 350 91 1,165,949 TL 1,717,854 TL 1,679,116 TL
Vendor68 1 15 6 9 57 55 115 1245 91.32 3,371,469 TL 3,107,083 TL 3,908,841 TL
Vendor69 1 12 5 7 22 28 76 551 89.94 1,395,345 TL 1,995,610 TL 1,134,891 TL
Vendor70 1 12 5 8 42 50 88 590 92.23 2,458,194 TL 973,404 TL 4,115,389 TL
Vendor71 1 9 5 4 23 24 55 498 89.41 1,321,073 TL 1,727,175 TL 872,499 TL
Vendor72 1 7 4 5 22 30 72 630 96.25 1,212,790 TL 5,149,724 TL 891,921 TL
Vendor73 2 12 4 8 60 34 78 689 87.34 1,852,594 TL 1,316,134 TL 1,394,026 TL
Min price (TL) 1,000,000 300,000 300,000 100,000 90,000 250,000 300,000
Max price (TL) 2,800,0000 7,700,000 5,700,000 1,100,000 1,100,000 550,000 650,000
M. Toloo, T. Ertay / Measurement 52 (2014) 135–144
 M. Toloo, T. Ertay / Measurement 52 (2014) 135–144 141

perspective. Otherwise, we formulate the following MIP Input 4 (x4): total number of selling adviser.
model: Input 5 (x5): total number of employee of vendors.
X
n Input 6 (x6): number of the exhibition and presentation
max dj activities realized by vendors.
j¼1
Input 7 (x7): number of advertisement broadcasting by
s:t: multi-media.
Xm Output 1 (y1): number of the sold vehicles.
v i xij  1 j ¼ 1; 2; . . . ; n Output 2 (y2): average satisfaction value for all vendors
i¼1
determined by the whole customers.
Xs X
m
Output 3 (y3): service endorsement of vendors after
ur yrj  v i xij þ dj ¼ 0 j ¼ 1; 2; . . . ; n
r¼1 i¼1
selling (in Turkish Lira).
a b Output 4 (y4): credit endorsement loaned by means of
v i pmin
a
i b  v i pmax
i
0
b a 1i <i m ð9Þ
financial founding for customers (in TL).
a b
v i pi  v i pi  0
a
max
b
min
b a 1i <i m Output 5 (y5): vendor’s selling endorsement on the
X
n spare parts (in TL).
hj ¼ n  1
j¼1 The last two rows of Table 1 report the minimum and
dj  Mhj j ¼ 1; 2; . . . ; n maximum input prices which are used in this paper.
hj  Ndj j ¼ 1; 2; . . . ; n The objective of the formulated DEA models in this study
is to determine the most cost efficient vendor in both the
hj 2 f0; 1g j ¼ 1; 2; . . . ; n
most and the least favorable price. In order to illustrate the
ur  e r ¼ 1; 2; . . . ; s applicability of our formulated CE–DEA models, we apply
It is easy to verify that the model (9) gives us a single them to the data set in Table 1. Toward this end, we utilize
pessimistic cost efficient DMU. GAMS 24.2.1 (http://www.gams.com) package for the
In the next section, we illustrate the use of suggested software codes of these CE–DEA models (see Appendix A).
models via a real data set. By applying the model (6) we have e* = 0.001114.
Regarding this assurance value, the model (5) implies that
4. An application of the new proposed CE–DEA Model Eopt = {40, 49, 56} which means that Vendor40, Vendor49,
for real-world efficiency assessment of automotive and Vendor56 are the most cost efficient vendor candidates
company under an optimistic perspective. Finally, Vendor40 will
be determined as the most cost efficient vendor via the
In this study, a Turkish automotive company that man- model (7).
ufactures both passenger car and light commercial vehicle To find the most cost efficient unit under a pessimistic
has been considered as an application of the proposed point of view, we apply the model (8) which leads to
CE–DEA models. This company is a global player that has Epes = {56}. Fortunately, in this case Epes is singleton and
been manufacturing five brands in its factory which has a the model (8) can individually find Vendor56 as the most cost
privileged position in the highest rank by achieving efficient with a pessimistic point of view. Therefore, it is
‘‘Silver’’ production level within 170 group factory of unnecessary to utilize the MIP model (9) with this data set.
World Class Manufacturing (WCM). As the pioneer of the
Turkish automotive industry, company exported its prod- 5. Conclusions and further research
ucts to 80 countries by the year of 2011 and its received
share from the total automotive industry export was at This study investigated the most cost efficient DMU
the level of 22.8% with 180,698 units. Turkey’s leading with price uncertainty while only the maximal and mini-
automotive company continued to create long-term value mal bounds of input prices for all DMUs are at hand. To
for its investors in 2010. Hence, the company succeeded do this, we considered two different scenarios: the most
in generating strong financial results and maintaining high favorable price (optimistic perspective) and the least favor-
levels of customer and distributor satisfaction. Company able price (pessimistic perspective). For each scenario, we
manufactured 312,000 vehicles in 2010, or 28.5% of the formulated a basic LP model for finding the most cost effi-
automotive sector total to become Turkey’s largest auto- cient candidate(s) and a MIP model for determining the
mobile and light commercial vehicle manufacturer. More most cost efficient unit among these candidates. To illus-
than 85% of company sales in 2010 were models manufac- trate the applicability of the new approach, we utilized
tured at its plant. Among global brands that manufacture an automotive industry data involving 73 vendors in Tur-
in Turkey, once again company is the top performer in key. VSP plays an important role in the worldwide market
terms of the ratio of local production to total sales. and hence the results of this paper support the manager for
As it is exhibited in Table 1, this real data set contains his/her decisions.
seven inputs and five outputs: The idea in this paper can be extended for measuring
revenue or profit efficiency with price uncertainty, impre-
Input 1 (x1): number of branch office. cise data, negative data, nondiscretionary factors and
Input 2 (x2): total number of exhibiting vehicle (includ- stochastic data. Furthermore, these research topics can be
ing branch office). considered for finding the most efficient DMU with certain
Input 3 (x3): total number of test vehicle. input prices.
142 M. Toloo, T. Ertay / Measurement 52 (2014) 135–144

Acknowledgements Appendix A

The research was supported by the Czech Science A.1. The formulated models for finding the most cost efficient
Foundation (GACR project 14-31593S) and through vendor in GAMS
European Social Fund within the project CZ.1.07/2.3.00/
20.0296.
M. Toloo, T. Ertay / Measurement 52 (2014) 135–144 143
144 M. Toloo, T. Ertay / Measurement 52 (2014) 135–144

References [18] E. Momeni, R. Farzipoor Saen, Developing a new chance-constrained

data envelopment analysis in the presence of stochastic data, Int. J.
Business Excellence 5 (3) (2012) 169–194.
[1] G.R. Amin, M. Toloo, Finding the most efficient DMUs in DEA: an
[19] C. Schaffnit, D. Rosen, J.C. Paradi, Best practice analysis of bank
improved integrated model, Comput. Indust. Eng. 52 (2007) 71–77.
branches: an application of DEA in a large Canadian bank, Eur. J.
[2] G.R. Amin, M. Toloo, A polynomial-time algorithm for finding e in
Oper. Res. 98 (1997) 269–289.
DEA models, Comput. Oper. Res. 31 (2004) 803–805.
[20] S. Talluri, R. Narasimhan, Vendor evaluation with performance
[3] A. Asosheh, S. Nalchigar, M. Jamporazmey, Information technology
variability: a max-min approach, Eur. J. Oper. Res. 146 (2003) 543–552.
project evaluation: an integrated data envelopment analysis and
[21] S. Talluri, R. Narasimhan, A. Nair, Vendor performance wıth supply
balanced scorecard approach, Expert Syst. Appl. 37 (8) (2010) 5931–
risk: a chance – constrained DEA approach, Int. J. Prod. Econ. 100
5938.
(2006) 212–222.
[4] M. Azadi, R. Farzipoor Saen, Developing a new chance-constrained
[22] W.M. Taylor, R.G. Thompson, R.M. Thrall, P.S. Dharmapala, DEA/AR
DEA model for supplier selection in the presence of undesirable
efficiency and profitability of Mexican banks: a total income model,
outputs, Int. J. Oper. Res. 13 (2012) 44–66.
Eur. J. Oper. Res. 98 (1997) 346–363.
[5] M. Azadi, R. Farzipoor Saen, Developing a chance-constrained free
[23] R.G. Thompson, P.S. Dharmapala, D.B. Humphrey, W.M. Taylor, R.M.
replicability hull model for supplier selection, Int. J. Logis. Syst.
Thrall, Computing DEA/AR efficiency and profit ratio measures with
Manage. 12 (4) (2012) 375–394.
an illustrative bank application, Ann. Oper. Res. 68 (1996) 303–327.
[6] M. Azadi, R. Farzipoor Saen, M. Tavana, Supplier selection using
[24] M. Toloo, On finding the most BCC-efficient DMU: a new integrated
chance-constrained data envelopment analysis with
MIP–DEA model, Appl. Math. Model. 36 (2012) 5515–5520.
nondiscretionary factors and stochastic data, Int. J. Ind. Syst. Eng.
[25] M. Toloo, An epsilon-free approach for finding the most efficient unit
10 (2) (2012) 167–196.
in DEA, Appl. Math. Model. (2014). <http://dx.doi.org/10.1016/
[7] A.S. Camanho, R. Dyson, Cost efficiency measurement: DEA
j.apm.2013.11.028>.
application to bank branch assessment, Eur. J. Oper. Res. 161
[26] M. Toloo, A. Masoumzadeh, M. Barat, Finding an initial basic
(2005) 432–446.
feasible solution for DEA models with an application on bank
[8] A. Charnes, W.W. Cooper, E. Rhodes, Measuring the efficiency of
industry, Comput. Econ. (2014), http://dx.doi.org/10.1007/s10614-
decision making units, Eur. J. Oper. Res. 2 (1978) 429–464.
014-9423-1.
[9] A. Charnes, W.W. Cooper, E. Rhodes, Short communication;
[27] M. Toloo, The most efficient unit without explicit inputs: an
measuring the efficiency of decision making units, Eur. J. Oper. Res.
extended MILP-DEA model, Measurement 46 (2013) 3628–3634.
3 (1979) 339.
[28] M. Toloo, S. Nalchigar, A new integrated DEA model for finding most
[10] A. Charnes, W.W. Cooper, Programming with linear fractional
BCC-efficient DMU, Appl. Math. Model. 33 (2009) 597–604.
functionals, Naval Res. Logis. Q. 9 (1962) 67–88.
[29] M. Toloo, S. Nalchigar, On ranking discovered rules of data mining by
[11] W.W. Cooper, J. Ruiz, I. Sirvent, Choosing weights from alternative
data envelopment analysis: some models with wider applications.
optimal solutions of dual multiplier models in DEA, Eur. J. Oper. Res.
Chapter 23 of new fundamental technologies in data mining. InTech
180 (2007) 443–458.
publisher, 2011. <http://www.intechweb.org/books/show/title/
[12] W.W. Cooper, L.M. Seiford, K. Tone, Data Envelopment Analysis: A
new-fundamental-technologies-in-data-mining>.
Comprehensive Text with Models, Applications, References and
[30] M. Toloo, S. Nalchigar, A new DEA method for supplier selection in
DEA-Solver Software, second ed., Kluwer Academic Publishers, USA,
presence of both cardinal and ordinal data, Expert Syst. Appl. 38
2007.
(2011) 14726–14731.
[13] T. Ertay, D. Ruan, Data envelopment analysis based decision model
[31] M. Toloo, B. Sohrabi, S. Nalchigar, A new method for ranking
for optimal operator allocation in CMS, Eur. J. Oper. Res. 164 (2005)
discovered rules from data mining by DEA, Expert Syst. Appl. 36
800–810.
(2009) 8503–8508.
[14] T. Ertay, D. Ruan, U.R. Tuzkaya, Integrating data envelopment
[32] C.A. Weber, J.R. Current, W.C. Benton, Vendor selection criteria and
analysis and analytic hierarchy for the facility layout design in
methods, Eur. J. Oper. Res. 50 (1991) 2–18.
manufacturing systems, Inf. Sci. 176 (2006) 237–262.
[33] D.D. Wu, A systematic stochastic efficiency analysis model and
[15] R. Fare, S. Grosskopf, C.A.K. Lovell, The Measurement of Efficiency of
application to international supplier performance evaluation, Expert
Production, Kluwer Academic Publishers, 1985.
Syst. Appl. 37 (2010) 6257–6264.
[16] M.J. Farrell, The measurement of productive efficiency, J. Roy. Stat.
[34] D. Wu, D.L. Olson, A Comparison of stochastic dominance and
Soc. 120 (Part III) (1957) 253–291.
stochastic DEA for vendor evaluation, Int. J. Prod. Res. 46 (8) (2008)
[17] K.C. Land, C.A.K. Lovell, S. Thore, Change-constraint data
2313–2327.
envelopment analysis, Manag. Decis. Econ. 14 (1993) 541–554.