Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Applied Mathematical Modelling 36 (2012) 212–224

Contents lists available at ScienceDirect

Applied Mathematical Modelling


journal homepage: www.elsevier.com/locate/apm

A new benchmarking approach in Cold Chain


Amir Shabani a, Reza Farzipoor Saen b,⇑, Seyed Mohammad Reza Torabipour a
a
Faculty of Economic and Management, Islamic Azad University, Science and Research Branch, P.O. Box 14155-7451, Tehran, Iran
b
Department of Industrial Management, Faculty of Management and Accounting, Islamic Azad University, Karaj Branch, Karaj, P.O. Box 31485-313, Iran

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

Article history: The dark area of Supply Chain Management (SCM) is inadequate attentions to those prod-
Received 22 April 2010 ucts that have limitations such as shelf life, need to special equipments and facilities for
Received in revised form 18 May 2011 sales, storage and distribution and so on. For this reason the concept of Cold Chain Manage-
Accepted 22 May 2011
ment (CCM) was emerged. The main objective of this paper is to develop a linear pair
Available online 30 May 2011
model for selecting the best sales agents as a ‘‘Benchmark’’ in the presence of non-discre-
tionary factors and imprecise data under Free Disposability assumption.
Keywords:
Ó 2011 Elsevier Inc. All rights reserved.
Cold chain
Free disposal hull
Data envelopment analysis
Imprecise data
Non-discretionary factors
Benchmarking

1. Introduction

Economic vicissitudes, market requirements, competition severity, new technologies, change of costumer and organiza-
tion expects, variety of resources and other issues provide a special field for creating value in an economic which is based on
network. Profit allocation, sharing the values and finally modifying risks lead to develop a strategic concept in a chain of
value added activities [1].
According to Wong and Wong [2], Supply Chain Management (SCM) is a set of approaches utilized to efficiently integrate
suppliers, manufacturers, warehouses, and stores, so that merchandise is produced and distributed at the right quantities, to
the right locations, and at the right time, in order to minimize system wide costs while satisfying service level requirements.
One of the branches of SCM is Cold Chain Management (CCM). The purpose of the CCM is managing activities related to per-
ishable products like medicine, blood, dairy, meat, food, vegetables, mushrooms, flowers and fruit products, and so on which
must be distributed in a special time and kept in the particular environment condition [3–7]. Since the spoilage of the per-
ishable products leads to waste and toxicosis, so monitoring all stages of Cold Chain (CC) is necessary [8,3,5].
According to reports that have been published in several countries, neglecting to keep the perishable items in particular
environment and as well neglecting the available time for delivery to customers leads to undesirable results that in turn
wastes resources. For instance, total value 75,000,000,000 Yuan of vegetables and fruits are deteriorated in freight in China
every year. These 370,000,000 tons of vegetables and fruits which are discarded can fully feed 200,000,000 people [3]. In
India it is estimated that around 35–40% of the total production of fresh fruits and vegetables is wasted only because of inad-
equate monitoring, poor logistics, ineffective CC facilities at retail points and lack of other infrastructural supports [9]. The
high rate of wastes causes a significant motivation for authors to develop a mathematical model selecting and benchmarking
the best sales agents.

⇑ Corresponding author.
E-mail addresses: shabani_dp@yahoo.com (A. Shabani), farzipour@yahoo.com (R.F. Saen), torabipour.mohammad@gmail.com (S.M.R. Torabipour).

0307-904X/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.apm.2011.05.051
A. Shabani et al. / Applied Mathematical Modelling 36 (2012) 212–224 213

Benchmarking is a strong management tool for identifying the best level of organizational performance as a standard in
industries, and helps to the organization to improve continuously [10]. Each inefficient organization wants to determine the
best benchmark for benchmarking. The organizations that benchmark, intends to improve their weaknesses.
To determining the benchmarks, many scholars have studied evaluation of SCM. They emphasized on taking into account
a set of performance criteria [11,12].
One of the important success factors of the CC is sales agent [3]. The performance appraisal in the CC is a cumbersome
task, because it has many features that set them apart from the other models of SCM. Some of these features include shelf
life restrictions, products in different seasons, production facilities and equipments like refrigerated transportation require-
ments, refrigerated storage and depot, productions with long time process, traceability, quality and safety of product [13,14].
Optimal selection of benchmarks improves performance level of sales agent and the CCM simultaneously. Some of the cri-
teria for optimal selection are imprecise, for instance, capacity of fridge is usually indicated by upper and lower limits or
hygiene level of equipment in the CC cannot be mentioned by precise data. Also a criterion like distance of a sales agent from
distribution center is not controllable by management. According to the points mentioned above, selecting the best sales
agents is a Multiple Criteria Decision Making (MCDM) problem. So, benchmarking the best sales agents, requires a model
in which consider imprecise and uncontrollable criteria.
Data Envelopment Analysis (DEA) is a well known non-parametric technique for measuring relative efficiency based on
Farrell [15] original work. DEA was later popularized by Charnes, Cooper and Rhodes (CCR) in 1978 [16]. DEA examines the
amounts of inputs consumed and outputs produced by a number of homogeneous units. An entity that is an object to be
measured for efficiency is called a Decision-Making Unit or DMU. DEA has two groups of objectives, in one hand, the outputs
that must be maximized, and on the other hand, the inputs that must be minimized [17]. This mathematical programming
method does not require to any conversion of the variables into a unique unit, as well the DEA does not need to decision
maker ideas. Since, DEA can identify relatively efficient DMU(s) among a group of given DMUs, it is a strong tool for
benchmarking.
A few researches have been done on the CC in different aspects. Vorst [18], in order to develop food supply chain, recom-
mended a performance evaluation system with focused on bottlenecks. Aramyan et al. [13] presented the performance indi-
cators for CCM in which ranked qualitative and quantitative indicators such as customer satisfaction and return on
investment, and so on. Cadilhon et al. [19] offered a conceptual structure for the analysis of vegetable supply chains in Asian
countries. As well the function of wholesale markets was emphasized in their study.
To improve the performance of stages in the CC there is a need to search criteria by which the performance of products,
services, and processes can be appraised [20]. One of common criteria in the CC is profitability. Getting appropriate rate of
return on investment is desired by all companies that operate in the CC [21]. In addition, the capability of companies to pro-
vide adequate products to market to satisfy food needs is one of the important performance criteria [22,21]. Proper perfor-
mance criteria must depict customer preferences such as demands, quality, timeliness, and prices [23].
Bogataj et al. [24] have researched on the stability of perishable goods in farm stage and transportation chains. Maintain-
ing the perishable goods requires the efficient equipment with special features, proper operating modes, and appropriate
information system [25,26].
To the best of knowledge of authors, there is not any reference that discusses benchmarking sales agent in the context of
CCM in the presence of imprecise data and non-discretionary factors.
The objective of this paper is to propose a new output-oriented pair of non-discretionary factors-imprecise data under
free disposability assumption for benchmarking sales agents in the context of CCM. This paper has several contributions
as below:

 For the first time the benchmarking concept is utilized in the CCM. This paper suggests the benchmark as an actual unit,
not a hypothetical.
 The proposed model considers non-discretionary factors for benchmarking.
 The proposed model considers imprecise data for benchmarking.
 For the first time the imprecise data and non-discretionary factors are considered simultaneously with free disposability
assumption.
 An application of the methodology has been performed on a set of data retrieved from the information of 29 sales
agents.

This paper proceeds as follows. In Section 2, the literature of SCM benchmarking is illustrated. The proposed model and
numerical example are discussed in Sections 3 and 4, respectively. Finally Section 5 provides concluding remarks.

2. Literature review: SCM benchmarking

Supply chain (SC) is defined as a system that comprises four processes including plan, source, make and deliver [10].
Benchmarking of SC covers various aspects such as process, products, performances, and strategies. SC benchmarking can
be viewed as an integrated benchmarking concept, giving a holistic examination for the whole entity [10]. Table 1 summa-
rizes the recent benchmarking researches in the area of the SCM.
214 A. Shabani et al. / Applied Mathematical Modelling 36 (2012) 212–224

Table 1
A summary of studies in the area of SCM benchmarking.

Title of research Purpose


Manufacturing Best Practices In Malaysian Small and Medium To investigate the current level of best manufacturing practices in Malaysian
Enterprises (SMEs) [27] ISO9000 certified SMEs
Logistics simulation modeling across construction supply chains To present a logistical analysis of SC construction by assessing the impact of
[28] varying demand on the performance of builders’ merchants’ logistics
Benchmarking Supplier Risks Using Bayesian Networks [29] To provide a methodology for benchmarking supplier risks through the creation
of Bayesian networks. The networks are used to determine a supplier’s external,
operational, and network risk probability to assess its potential impact on the
buyer organization
Internal benchmarking for assessment of supply chain performance To propose a methodology for the internal benchmarking to reduce the
[30] variability in performance among SC of same focal firm
Benchmarking trust signals in supply chain alliances: moving In this paper a multi-faceted measure of trust is developed and used to
toward a robust measure of trust [31] benchmark the extent to which trust signals are used in alliance management
Identifying improvement areas when implementing green To discuss research gaps and the potential applications of analytic hierarchy
initiatives using a multitier AHP approach [32] process (AHP) in an internal benchmarking process used to identify
improvement areas when firms attempt to adopt green initiatives with a SC
perspective
Developing environmental supply chain performance measures To review literature and present a proposed research agenda to examine
[33] whether environmental, i.e. green performance measures, can be integrated
within an existing SC performance framework, explore what a meaningful
industry-recognized environmental measure should look like, and understand
the direct benefits of incorporating environmental measures within a SC
performance framework
Critical evaluation of paradigms for modeling integrated supply This paper tries to determine the most appropriate modelling paradigm for a SC
chains [34]
Benchmarking the viability of SCM for entrepreneurial business To benchmark the viability of collaborative SC business models for small
model design [35] businesses
Utilizing cash-to-cash to benchmark company performance [36] To show how the cash-to-cash (C2C) metric may be used to benchmark SC
performance
Benchmarking firms’ operational performance according to their The purpose of this paper is to identify the critical dimensions of usage of
use of internet-based interorganizational systems [37] Internet based interorganizational systems (IOISs) of the best performing firms
Designing a supply chain management academic curriculum using To utilize quality function deployment (QFD), benchmarking analyses and other
QFD and benchmarking [38] innovative quality tools for developing a new customer-centered
undergraduate curriculum in SCM
An approach to improving logistical performance with cross-unit To identify the best logistical practices and to implement them with minimal
benchmarking [39] organizational resistance in an environment with relatively similar business
units utilizing an internal benchmarking approach
Review of port performance approaches and a supply chain This paper proposes an integrative framework to port performance by
framework to port performance benchmarking [40] conceptualizing ports from a logistics and SCM approach
Development and benchmarking of an epoch time synchronization To present a new epoch time synchronization approach for distributed
method for distributed simulation [41] simulation federates
A benchmarking scheme for SC collaboration [42] To highlight that SC collaboration shifted the focus of benchmarking from a
single company level to an interorganizational level. Also, the authors
recommend an integrated benchmarking scheme for SC collaboration that
consists of enabling practices and collaborative performance system
Benchmarking SCM practice in New Zealand [43] This paper illustrates an empirical study of benchmarking on SC practices in
New Zealand companies
Measuring the success of collaboration across the virtual SC To develop a SC maturity model that reflects how companies progress in terms
through performance measurement systems and benchmarking of operational capability
[44]
Benchmarking a logistical operations based on causal model [45] To develop a causal model for identifying possible initiatives to bridge the
performance gap between a company and best-in-class performers

3. Proposed model

3.1. Free disposal hull

One of the remarkable ‘nonparametric’ models of the DEA family is the free disposal hull (FDH) model. The FDH as the
first formulated by Deprins et al. [46] is based on a fairly simple model. The FDH has received a noticeable amount of
A. Shabani et al. / Applied Mathematical Modelling 36 (2012) 212–224 215

researchers’ attention. Specifically, and contrary to other models of the DEA, convexity and/or proportionality assumptions
are not needed to compute the relative efficiency of DMUs. The basic motivation of the FDH is to ensure that the efficiency
evaluations are affected by only actually observed performances [47]. The production frontier of the FDH is ‘staircase’ shapes.
The efficiency evaluation is done by this kind of production frontier. In simple words, DMUo (the DMU under evaluation) is
compared just with an actual DMU, not hypothetical. For instance, Fig. 1 shows that the point Q is not allowed to be com-
pared because Q is a hypothetical point that derived from points B and C. Under the FDH assumption, the DMUp would like to
be compared just with B or C.
Formally,

PFDH ¼ fðx; yÞjx  xj ; y  yj ; x; y  0; j ¼ 1; . . . ; ng; ð1Þ

where xj P 0, yj P 0 are actually observed performances for j = 1, ... , n DMUs. In words, a point is a member of the production
possibility set if all of its input coordinates are at least as large as their corresponds in the vector of observed values Xj for any
j = 1, ... , n and if their output coordinates are no greater than their corresponds in the vectors yj of observed values for this
same j [48] .
The following mixed integer programming formulation is the basic output-oriented FDH model [47].

max ¼ h
X
m
s:t: kj xij  xio ;
i¼1

X
s
kj yrj  hyio ; ð2Þ
r¼1

X
m
kj ¼ 1;
i¼1

kj 2 f0; 1g; j ¼ 1; . . . ; n; 8i; r;

where X and Y contain the given input and output matrices and kj 2 f0; 1g means that the components of kj are constrained
P
to be bivalent. That is, they must all have values of zero or unity so that together with the condition m
i¼1 kj ¼ 1 one and only
one of the performances actually observed can be chosen.
The reason of choosing the output-oriented FDH model is to maximize expansion in all outputs keeping all inputs
unchanged.

3.2. Non-discretionary factors

In the decision making process, some of variables are uncontrollable. These variables in the DEA called non-discretionary
factors. The organization management does not have any control on this type of variables, at least in the short term. For
example, if DMUo is inefficient and one of variables takes non-discretionary role, to improve efficiency, the DMUo cannot
change the quantity of the non-discretionary factor [49]. Non-discretionary factors in the DEA models can take both input
and/or output roles.

Fig. 1. FDH production frontier.


216 A. Shabani et al. / Applied Mathematical Modelling 36 (2012) 212–224

3.3. Imprecise data

Cooper et al. [50], developed a special model that take into account imprecise data, and called Imprecise Data Envelop-
ment Analysis (IDEA). The concept of imprecise data is referred to interval data, ordinal data and ratio interval data. The idea
of IDEA for the additive model was developed by Lee et al. [51]. They also presented a reason that the Cooper et al. [50], pro-
cedure makes the DEA model so complicated. Wang et al. [52], developed a new pair of interval DEA models for handling the
imprecise data. Their model is able in deal with the interval data, the ordinal preference information, fuzzy data and their
mixture. As compared to the IDEA model developed by Cooper et al. [50], their interval DEA models are much easier to
understand, and also more convenient to use. In addition, compared with the interval DEA models developed by Despotis
and Smirlis [53], their interval DEA models utilize a fixed and united production frontier as a target for measuring the effi-
ciencies of all DMUs. Hence, their models are more rational and more reliable. Besides, they treated more logical with ordinal
preference information than the way of Zhu [54]. Jahanshahloo et al. [55], proposed a model for dealing with the imprecise
data on the FDH model. But their model is based on the procedure of Cooper et al. [50] and the same criticism of Cooper et al.
[50] applies to their studies.
However, all of the above-mentioned references do not consider non-discretionary factors and imprecise data under the
assumption of FDH, simultaneously.
Now suppose that there are n DMUs to be evaluated. For each DMU consider m inputs to produce s outputs. Specially,
DMUj, j = 1, . . . , n consumes the quantity of Xj = {xij} of inputs, i = 1, . . . , m and produces the quantity of Yj = {yrj} of outputs
r = 1, . . . , s. The basic logic of the IDEA is that all the inputs and outputs cannot be exact, because of the presence of imprecise
data. They are only known to lie within the upper and lower bounds represented by the intervals ½xLij ; xUij  and ½yLrj ; yUrj , where
xLij  0; yLrj  0.
To solve the problem of imprecise data, a pair of output-oriented model is developed with regard to interval efficiency
evaluation of Wang et al. [52], method.
Let
Ps
ur yrj
hj ¼ Pr¼1
m ; j ¼ 1; . . . ; n ð3Þ
i¼1 v i xij
be the efficiency of DMUj. The concept of interval data is as below.

Ps h i hP Ps i"Ps #
s Ps
r¼1 ur yLrj ; yUrj L
r¼1 ur yrj ;
U
r¼1 ur yrj ur yLrj r¼1 ur yrj
U
hj ¼ Pm ¼ hP Pm i ¼ Pr¼1
m ; P m ; j ¼ 1; . . . ; n: ð4Þ
i¼1 v i ½xij ; xij  i¼1 v i xij i¼1 v i xij
L U m L U
i¼1 v i xLij ; i¼1 v i xU
ij

Clearly hj gets an interval value that can be shown by ½hLj ; hUj .


Let
"Ps Ps #
L U
r¼1 ur yrj r¼1 ur yrj
hj ¼ ½hLj ; hUj  ¼ Pm ; Pm # ð0; 1; j ¼ 1; . . . ; n: ð5Þ
U
i¼1 i xij v L
i¼1 i xij v
So
Ps L
r¼1 ur yrj
hLj ¼ Pm >0 ð6Þ
v U
i¼1 i xij

and
Ps U
r¼1 ur yrj
hUj ¼ Pm  1: ð7Þ
v L
i¼1 i xij

Two pairs of the following fractional programming can calculate the lower and upper efficiency bounds of the DMUo.
Ps
ur yLro
max hLjo ¼ Pr¼1
m ; j ¼ 1; . . . ; n
i¼1 v i xio
U

Ps U
r¼1 ur yrj
s:t: hUj ¼ m  1; j ¼ 1; . . . ; n; ð8Þ
P
v i xLij
i¼1

ur ; v i  0; 8r; i;
A. Shabani et al. / Applied Mathematical Modelling 36 (2012) 212–224 217

Ps
ur yUro
max hUjo ¼ Pr¼1
m ; j ¼ 1; . . . ; n
i¼1 v i xio
L

Ps U
r¼1 ur yrj ð9Þ
s:t: hUj ¼ Pm  1; j ¼ 1; . . . ; n;
i¼1 v i xij
L

ur ; v i  0; 8r; i:
It is obvious that the above models are non-linear and can be linearized into the output-oriented models by Charnes and
Cooper [16] transformation as below.

X
m
min hLjo ¼ v i xUio
i¼1

X
s
s:t: ur yLro ¼ 1;
r¼1 ð10Þ
X
s X
m
ur yUrj  v i xLij  0; j ¼ 1; . . . ; n;
r¼1 i¼1

ur ; v i  0; 8r; i:

X
m
min hUjo ¼ v i xLio
i¼1

X
s
s:t: ur yUro ¼ 1;
r¼1 ð11Þ
X
s X
m
ur yUrj  v i xLij  0; j ¼ 1; . . . ; n;
r¼1 i¼1

ur ; v i  0; 8r; i;
where xLij
and xUij demonstrate the lower and upper bounds of ith input, respectively. yLrj and yUrj demonstrate the lower and
upper bounds of rth output, respectively. Also hLio and hUio stand for the best lower and upper possible relative efficiency of
DMUo. In addition, ur and vi show the weights of rth output and ith input, respectively.
Note that both the best and the worst production activities of the DMUo are considered in Model (11), i.e. one in the con-
straint set and the other in the objective function.
At this juncture, the new models are proposed. Dual program of the Models (10) and (11) are mentioned below:

max hLjo ¼ h
X
s
s:t kj yUrj  hyLro ; r ¼ 1; . . . ; s;
r¼1

X
m ð12Þ
kj xLij  xUio ; i ¼ 1; . . . ; m;
i¼1

h free;
kj  0; j ¼ 1; . . . ; n;

max hUjo ¼ h
X
s
s:t: kj yUrj  hyUro ; r ¼ 1; . . . ; s;
r¼1

X
m ð13Þ
kj xLij  xLio ; i ¼ 1; . . . ; m;
i¼1

h free;
kj  0; j ¼ 1; . . . ; n:
The Model (13) assigns the production frontier for all the DMUs and the Model (12) utilizes the production frontier as a
benchmark to evaluate the lower bound efficiency of each DMU.
218 A. Shabani et al. / Applied Mathematical Modelling 36 (2012) 212–224

Now, suppose that the input and output variables may be dichotomized into the subsets of discretionary (D) and non-
discretionary (N) variables. Thus, the Models (12) and (13) change into the following form.

max hLjo ¼ h
X s
s:t: kj yUrj  hyLro r 2 D; r ¼ 1; . . . ; s;
r¼1
X
s
kj yUrj  yLro r 2 N; r ¼ 1; . . . ; s;
r¼1
ð14Þ
Xm
kj xLij  xUio ; i ¼ 1; . . . ; m;
i¼1

h free;
kj  0; j ¼ 1; . . . ; n;

max hUjo ¼ h
X
s
s:t: kj yUrj  hyUro r 2 D; r ¼ 1; . . . ; s;
r¼1
X s
kj yUrj  yUro r 2 N; r ¼ 1; . . . ; s;
r¼1
ð15Þ
Xm
kj xLij  xLio ; i ¼ 1; . . . ; m;
i¼1

h free;
kj  0; j ¼ 1; . . . ; n:
P
According to the concept of the FDH, the only difference between the CCR and the FDH is two constraints, i.e. m i¼1 kj ¼ 1
and kj 2 f0; 1g by the FDH model [47]. Therefore, by adding these two constraints in the Models (14) and (15), there will be

max hLjo ¼ h
X
n
s:t: kj yUrj  hyLro r 2 D; r ¼ 1; . . . ; s;
j¼1
X
n
kj yUrj  yLro r 2 N; r ¼ 1; . . . ; s;
j¼1
X
n ð16Þ
kj xLij  xUio ; i ¼ 1; . . . ; m;
j¼1
X
n
kj ¼ 1; j ¼ 1; . . . ; n;
j¼1

kj 2 f0; 1g; j ¼ 1; . . . ; n;
h free;

max hUjo ¼ h
X n
s:t: kj yUrj  hyUro r 2 D; r ¼ 1; . . . ; s;
j¼1
X
n
kj yUrj  yUro r 2 N; r ¼ 1; . . . ; s;
j¼1
X
n ð17Þ
kj xLij  xLio ; i ¼ 1; . . . ; m;
j¼1
X
n
kj ¼ 1; j ¼ 1; . . . ; n;
j¼1

kj 2 f0; 1g; j ¼ 1; . . . ; n;
h free:
Consequently, the Models (16) and (17) take into account non-discretionary factors and imprecise data, simultaneously.
A. Shabani et al. / Applied Mathematical Modelling 36 (2012) 212–224 219

To convert Models (16) and (17) to linear programming problem, the binary assumption of kj , j = 1, . . . , n were removed
from the Models (16) and (17). Note that the radial expansion variable h is replaced by a well defined set of individual expan-
sion factors hj for each DMUj.
X
n
max hLjo ¼ hj ;
j¼1

s:t: kj yUrj  hj yLro r 2 D; r ¼ 1; . . . ; s; j ¼ 1; . . . ; n;


kj yUrj  kj yLro r 2 N; r ¼ 1; . . . ; s; j ¼ 1; . . . ; n;
kj xLij  kj xUio ; i ¼ 1; . . . ; m; j ¼ 1; . . . ; n; ð18Þ
X n
kj ¼ 1; j ¼ 1; . . . ; n;
j¼1

hj ¼ free; j ¼ 1; . . . ;
kj  0; j ¼ 1; . . . ; n;

X
n
max hUjo ¼ hj
j¼1

s:t: kj yUrj  hj yUro r 2 D; r ¼ 1; . . . ; s; j ¼ 1; . . . ; n;


kj yUrj  kj yUro r 2 N; r ¼ 1; . . . ; s; j ¼ 1; . . . ; n;
kj xLij  kj xLio ; i ¼ 1; . . . ; m; j ¼ 1; . . . ; n; ð19Þ
X
n
kj ¼ 1; j ¼ 1; . . . ; n;
j¼1

hj ¼ free; j ¼ 1; . . . ; n;
kj  0; j ¼ 1; . . . ; n:

Theorem 1. The Models (18) and (19) are equivalent to the Models (16) and (17).

Proof. For proof, please see Agrell and Tind [17]. h

Definition 1. A DMU, DMUo, is said to be DEA efficient if its best possible upper bound efficiency hU
o ¼ 1 otherwise, it is said
to be DEA inefficient hU
o > 1.
The Models (18) and (19) are called Non-discretionary-Imprecise Linear Free Disposal Hull (NILFDH).

3.4. The transformation of ordinal preference data

To use the pair of NILFDH models, a procedure for obtaining the interval data from the ordinal preferences data is
presented.
Consider a condition in which some of data for DMUs may be ordinal preference data. There are three kinds of ordinal
preference information [52].

3.4.1. The strong ordinal information

yrj > yro and xij > xio : ð20Þ

It is not different if Eq. (20) mentioned as follows:


yrj > vr yro and xij > gi xio : ð21Þ

Note that vr > 1 and gi > 1 are the degrees of preferences of output and input, respectively. For instance, the procedure of
changing the ordinal preference data of input xij, j = 1, . . . , n is mentioned. The other outputs and inputs which have ordinal
data can be converted into the interval data by the same method. Consider xi1 > xi2   > xin as the strong ordinal preference
data, then 1  ^ xij  gi ^
xij ; ^ xi;jþ1 ðj ¼ 1; . . . ; n  1Þ and ^ xij  ri is the ordinal relationship after scale transformation. Where ri is
a small positive number reflecting the ratio of the possible minimum of {xij|j = 1, . . . , n} to its possible maximum. The values
of the upper and lower bounds for each ^ xij are obtained by
^xij 2 ½ri gnj 1j
i ; gi ; j ¼ 1; . . . ; n and ri  g1n
i : ð22Þ
220 A. Shabani et al. / Applied Mathematical Modelling 36 (2012) 212–224

3.4.2. The weak ordinal data

yrj  yro and xij  xio : ð23Þ

Suppose that xi1 P xi2 P    P xin is as the weak ordinal preference data. Then 1  xi1  xi2      xin  ri is the ordinal
relationship after scale transformation. Where ri is also a small positive number reflecting the ratio of the possible minimum
of {xij|j = 1, . . . , n} to its possible maximum. For each ^
xij the lower and upper values are obtained as follows:
^xij 2 ½ri ; 1; j ¼ 1; . . . ; n: ð24Þ

3.4.3. The indifference relationship

yrj ¼ yro and xij ¼ xio : ð25Þ

Obtaining the values of indifference relationships ordinal data is completely same as the weak ordinal data method.
In summary, discussion in Section 3 can be presented as below.

Step 1. Define inputs and outputs.


Step 2. Collect the related data of inputs and outputs.
Step 3. Transform the ordinal data to the interval data via Parts (22) or (24).
Step 4. Compute efficiencies by Models (18) and (19).

Fig. 2 represents these steps graphically.

4. Numerical example

Iran Dairy Industries CO. (IDIC) was established in 1954 for producing dairy products. Nowadays after over 56 years, IDIC
is lied in top 100 Iranian companies rank. Also, IDIC has become a most major and pioneer producer and exporter of different
kinds of dairy products, in Iran. IDIC is formed by 50 different companies. A number of them like Commercial Service Co. and
Pegah Protein Co. are engaged in the stage of pre-production process. These companies are established with the aim of in-

Fig. 2. The computations process.


A. Shabani et al. / Applied Mathematical Modelling 36 (2012) 212–224 221

crease in number of new and expansion of the existing husbandries and consequently increase in production of raw milk
with higher quality. Seventeen other companies are responsible in production and process of raw milk and other dairy prod-
ucts. Other 27 companies are responsible in sales and export. These companies are the largest chain distribution of dairy
products in the country and also are responsible for the market survey. In order to meet the need of the consumers of dif-
ferent ages, and considering the different tastes, IDIC is working on more than 300 different products. The major products of
the company are milk, yoghurt, cheese, cream, yoghurt drink, butter, curd, and ice cream, milk & whey powder, mineral
water, and fruit juice.
One of the main distribution centers of IDIC is located in near Tehran, Iran’s capital city. The inventories of this distribu-
tion center are sent every day to all provinces. There are 29 main sales agents which are selected for efficiency evaluation.
Here sales agents are considered as DMUs. The information in terms of inputs and outputs of sales agents have been repre-
sented in Table 2. Note that the data of Table 2 were changed by IDIC due to data security. The annual waste and distance of
each sales agent from the distribution center are two inputs of problem. The annual sale, hygiene level, and capacity of fridge
are considered as outputs, which are respectively the cardinal, ordinal preferences and the interval data. Note that the fridge
capacity is a non-discretionary factor, because management does not have any control on it, at least in short term. Since the
hygiene level is expressed as the ordinal preferences information, it requires to be transformed into the interval information.
Suppose that the preference intensity parameter v2 = 1.12 and the ratio parameter r2 = 0.001. As Section 3.4, Table 3 pro-
vides data transformation results. Table 4 presents the interval efficiency values for each DMU derived by Models (18) and
(19). For instance, consider the DMU4 (Bandar Abbas). The values [0.131, 0.452] denote [hL, hU] for the DMU4, i.e. the lower
bound is 0.131 and the upper bound is 0.452.
Note that the output-orientated DEA models are adopted in this paper and according to the Definition 1, the optimal val-
ues of the Models (18) and (19) will be always greater than or equal to unity. However, both lower and upper bounds of
1
efficiency shown in Table 4 have been provided by means of efficiencyvalue .
Benchmarks columns display the lower and upper bounds benchmarks for inefficient DMUs. Consider again the DMU4.
The lower bound benchmark for DMU4 is DMU11, i.e. if DMU4 desires to be efficient, it must benchmark DMU11. So DMU4
must increase its annual sales from 3169255 to 24115288, similar to annual sales of DMU11. Also, DMU25 is selected as
the upper bound benchmark for DMU4. The hygiene level of DMU4 is 9; therefore, DMU4 must improve this criterion similar
to DMU25, i.e. 2.
Please note that since the main advantage of the FDH model is to compare the DMUo with an actual observation, the
benchmarks are attainable for all inefficient DMUs.

Table 2
The data related to inputs and outputs of 29 sales agents.

No. DMUs Inputs Outputs


Annual waste (US dollar) Distance (km) Annual sale (US dollar) Hygiene levela Fridge capacity (kg)
1 Ahvaz 157,556 881 11,483,099 7 [2200, 2650]
2 Arak 79,142 288 5,418,166 29 [890, 1250]
3 Ardebil 48,352 588 2,946,111 25 [386, 495]
4 Bandar Abbas 141,731 1501 3,169,255 9 [700, 850]
5 Birjand 198,407 1313 10,425,373 26 [797, 1043]
6 Bojnord 71,443 1367 8,853,938 28 [1275, 1425]
7 Boshehr 333,062 1215 10,275,867 24 [700, 795]
8 Gorgan 133,899 381 8,645,678 19 [690, 850]
9 Hamedan 86,102 336 6,343,774 10 [800, 995]
10 Ilam 380,475 710 836,4318 22 [578, 811]
11 Isfahan 85,425 414 24,115,288 6 [978, 1203]
12 Kerman 299,119 1064 6,799,502 12 [1171, 1389]
13 Kermanshah 293,920 525 3,479,782 11 [775, 1012]
14 Kohoram Abad 99,674 1001 5,003,216 14 [1500, 1826]
15 Mashad 167,908 924 18,770,358 4 [1221, 1458]
16 Orumiyeh 90,155 946 3,356,169 13 [541, 799]
17 Qazvin 64,843 150 1,598,596 5 [889, 1113]
18 Qom 56,973 132 1,959,727 8 [594, 712]
19 Rasht 217,401 323 4,251,662 15 [680, 860]
20 Sanandaj 144,787 512 4,927,831 20 [695, 832]
21 Sari 116,064 250 6,873,149 16 [1500, 1948]
22 Semnan 237,121 228 3,173,005 21 [723, 980]
23 Shahrekord 71,125 521 2,510,237 23 [905, 1025]
24 Shiraz 113,534 895 19,992,124 1 [650, 683]
25 Tabriz 97,560 624 14,357,914 2 [985, 1102]
26 Yasuj 201,150 738 4,260,214 18 [751, 990]
27 Yazd 72,618 677 5,080,142 3 [700, 960]
28 Zahedan 29,0870 1605 8,430,431 27 [730, 760]
29 Zanjan 99,344 330 4,838,680 17 [700, 745]
a
Ordinal scale from 1 = best to 29 = worst with the supposed preference intensity parameter v2 = 1.12 and the ratio parameter r2 = 0.001.
222 A. Shabani et al. / Applied Mathematical Modelling 36 (2012) 212–224

Table 3
Interval estimate for the 29 sales agents after transformation of the ordinal preference data.

No. DMUs Hygiene level


1 Ahvaz [0.012, 0.506]
2 Arak [0.001, 0.041]
3 Ardebil [0.0015, 0.065]
4 Bandar Abbas [0.009, 0.403]
5 Birjand [0.0014, 0.058]
6 Bojnord [0.0011, 0.046]
7 Boshehr [0.0017, 0.073]
8 Gorgan [0.0031, 0.130]
9 Hamedan [0.008, 0.360]
10 Ilam [0.0022, 0.092]
11 Isfahan [0.013, 0.567]
12 Kerman [0.0068, 0.287]
13 Kermanshah [0.007, 0.321]
14 Kohoram Abad [0.005, 0.229]
15 Mashad [0.017, 0.711]
16 Orumiyeh [0.0061, 0.256]
17 Qazvin [0.015, 0.635]
18 Qom [0.010, 0.452]
19 Rasht [0.0048, 0.204]
20 Sanandaj [0.0027, 0.116]
21 Sari [0.0043, 0.182]
22 Semnan [0.0024, 0.103]
23 Shahrekord [0.0019, 0.082]
24 Shiraz [0.023, 1]
25 Tabriz [0.021, 0.892]
26 Yasuj [0.0034, 0.145]
27 Yazd [0.019, 0.797]
28 Zahedan [0.0012, 0.052]
29 Zanjan [0.0038, 0.163]

Table 4
The efficiencies interval and benchmarks for the 29 sales agents.

No. DMUs Efficiency interval Lower bound benchmarks Upper bound benchmarks
1 Ahvaz [1,1]
2 Arak [1,1]
3 Ardebil [1,1]
4 Bandar Abbas [0.131, 0.452] 11 25
5 Birjand [0.432320, 0.432339] 11 11
6 Bojnord [1, 1]
7 Boshehr [0.426112, 0.426114] 11 11
8 Gorgan [1, 1]
9 Hamedan [1, 1]
10 Ilam [0.346848, 0.346981] 11 11
11 Isfahan [1, 1]
12 Kerman [0.281, 0.403] 11 1
13 Kermanshah [0.144, 0.567] 11 15
14 Kohoram Abad [1, 1]
15 Mashad [1, 1]
16 Orumiyeh [0.139, 0.452] 11 11
17 Qazvin [1, 1]
18 Qom [1, 1]
19 Rasht [0.618, 1] 21
20 Sanandaj [0.2043, 0.2046] 11 11
21 Sari [1, 1]
22 Semnan [1, 1]
23 Shahrekord [1, 1]
24 Shiraz [0.829, 1] 11
25 Tabriz [0.595, 1] 11
26 Yasuj [0.176, 0.256] 11 11
27 Yazd [1, 1]
28 Zahedan [0.341, 0.359] 11 11
29 Zanjan [0.893, 1] 2
A. Shabani et al. / Applied Mathematical Modelling 36 (2012) 212–224 223

5. Concluding remarks

The sales agent is a main factor in the CCM [21]. Ignoring this factor leads to wastes for those companies that operate in
perishable product markets. The efficient sales agent can be as a benchmark for other sales agents. Benchmarking is a man-
agement tools to implement the best practices found in similar industries or even in different industries in order to improve
the performance of an organization [35]. Therefore, the benchmark sales agent can be a standard which stimulates other
sales agents to improve their performance.
In this study a pair model was proposed for the performance evaluation of sales agents in the CCM. First, the concept of
CCM and its limitations were introduced. Next, the free disposability assumption was incorporated into the imprecise data
and non-discretionary factors. Through the proposed model the decision makers are enabled to consider many qualitative
and quantitative criteria. Finally, to obtain a global optimal solution, the binary model was changed into the linear model.
The problem considered in this research is at the initial stage of investigation and further studies can be done based on
the results of this paper. Some of them are as below:

 Combination of the FDH model in the presence of weights restrictions is an interesting topic for future research. The sen-
sitivity of the FDH model is high and model cannot accept all experts’ opinions, hence the study on the Assurance Region
(AR) seems to be worthwhile.
 There is possibility to occur some ties among efficient DMUs. Developing a model for ranking the efficient DMUs will be
another research topic.
 In this study, the proposed model has been applied to benchmarking the best sales agents in the context of Cold Chain.
However, the same models could be applied, with minor modifications, to other problems related to benchmarking.

Acknowledgements

The authors wish to thank an anonymous reviewer for valuable suggestions and comments.

References

[1] A. Shabani, S.M.R. Torabipour, R. Farzipour Saen, Container selection in the presence of partial dual-role factors, Int. J. Phys. Distrib. Logistics Manage.,
in press.
[2] W.P. Wong, K.Y. Wong, Supply chain performance measurement system using DEA modeling, Ind. Manage. Data Syst. 107 (2007) 361–381.
[3] W. Lan, Z.Z. Ya, A research on related questions of Chinese food cold chain development, in: International Conference on Management of e-Commerce
and e-Government, Jiangxi, 2008, pp. 18–21.
[4] R. Montanari, Cold chain tracking: a managerial perspective, Trends Food Sci. Technol. 19 (2008) 425–431.
[5] H. Rediers, M. Claes, L. Peeters, K.A. Willems, Evaluation of the cold chain of fresh-cut endive from farmer to plate, Postharvest Biol. Technol. 51 (2009)
257–262.
[6] Y. Samant, H. Lanjewar, D. Parker, L. Block, G.S. Tomar, B. Stein, Evaluation of the Cold-chain for Oral Polio Vaccine in a Rural District of India, Public
Health Reports 122, 2007, pp. 112–121.
[7] Z.Y. Xia, Study of countermeasures relating to developing food cold chain logistics, Logistics Sci. Technol. 16 (2007) 21–23.
[8] J.C. Kuo, M. Chen, Developing an advanced multi-temperature joint distribution system for the food cold chain, Food Control 21 (2010) 559–566.
[9] N. Viswanadham, Can India be the Food Basket for The World? Working Paper, Series ISBHydrabad. <http://www.isb.edu/faculty/
Working_Papers_pdfs/Can_India_be_the_Food_Basket_for_the_World.pdf> (accessed 17.05.11).
[10] W.P. Wong, K.Y. Wong, A review on benchmarking of supply chain performance measures, Benchmark. Int. J. 15 (2008) 25–51.
[11] A. Gunasekaran, C. Patel, R.E. McGaughey, A framework for supply chain performance measurement, Int. J. Prod. Econ. 87 (2004) 333–347.
[12] A. Otto, H. Kotzab, Does supply chain management really pay? Six perspectives to measure the performance of managing a supply chain, Eur. J. Oper.
Res. 144 (2003) 306–320.
[13] L.H. Aramyan, A.G.J.M.O. Lansink, J.G.A.J. Vandervorst, O. Van Kooten, Performance measurement in agri-food supply chains: a case study, Supply Chain
Manage. Int. J. 12 (2007) 304–315.
[14] E. Mangina, I.P. Vlachos, The changing role of information technology in food and beverage logistics management: beverage network optimization
using intelligent agent technology, J. Food Eng. 70 (2005) 403–420.
[15] M.J. Farrell, The measurement of productive efficiency, J. Roy. Stat. Soc. 120 (1957) 253–290.
[16] A. Charnes, W.W. Cooper, E. Rhodes, Measuring the efficiency of decision making units, Eur. J. Oper. Res. 2 (1978) 429–444.
[17] P.J. Agrell, J. Tind, A dual approach to non-convex frontier models, J. Prod. Anal. 16 (2001) 129–147.
[18] J.G.A.J. Vorst, Performance Measurement in Agri-food Supply Chain Networks. An Overview, Quantifying the Agri-food Supply Chain, vol. 15, Springer,
Dordrecht, 2005. pp. 13–24.
[19] J.J. Cadilhon, A.P. Fearne, P. Moustier, N.D. Poole, Modeling vegetable marketing systems in South East Asia: phenomenological insights from Vietnam,
Supply Chain Manage.: Int. J. 8 (2003) 427–441.
[20] F.T.S. Chan, H.K. Chan, H.C.W. Lau, R.W.L. Ip, An AHP approach in benchmarking logistics performance of the postal industry, Benchmark. Int. J. 13
(2006) 636–661.
[21] C.A.D. Silva, S.M.D. Filho, Guidelines for Rapid Appraisals of Agri-Food Chain Performance in Developing Countries, Agricultural Management,
Marketing and Finance Occasional Paper, FAO, Rome, 2007.
[22] L. Manning, R. Baines, S. Chadd, Benchmarking the poultry meat supply chain, Benchmark. Int. J. 15 (2008) 148–165.
[23] A. Fearne, S. Barrow, D. Schulenberg, Implanting the benefits of buyer–supplier collaboration in the soft fruit sector, Supply Chain Manage. Int. J. 11
(2006) 3–5.
[24] M. Bogataj, L. Bogataj, R. Vodopivec, Stability of perishable goods in cold logistic chains, Int. J. Prod. Econ. 93 (2005) 345–356.
[25] K. Amjadi, Integrating food hygiene into quantity food production systems, Nutr. Food Sci. 35 (2005) 169–183.
[26] L. Manning, R.N. Baines, S.A. Chadd, Quality assurance models in the food supply chain, Brit. Food J. 108 (2006) 91–104.
[27] A. Anuar, R.M. Yusuff, Manufacturing best practices in Malaysian small and medium enterprises (SMEs), Benchmark. Int. J. 18 (2011) 1–17.
224 A. Shabani et al. / Applied Mathematical Modelling 36 (2012) 212–224

[28] C. Vidalakis, J.E. Tookey, J. Sommerville, Logistics simulation modelling across construction supply chains, Construct. Innov. Informa. Process Manage.
11 (2011) 212–228.
[29] A. Lockamy, Benchmarking supplier risks using Bayesian networks, Benchmark. Int. J. 18 (2011) 1–22.
[30] G. Soni, R. Kodali, Internal benchmarking for assessment of supply chain performance, Benchmark. Int. J. 17 (2010) 44–76.
[31] S.L. Jones, S.E. Fawcett, A.M. Fawcett, C. Wallin, Benchmarking trust signals in supply chain alliances: moving toward a robust measure of trust,
Benchmark. Int. J. 17 (2010) 705–727.
[32] R. Sarmiento, A. Thomas, Identifying improvement areas when implementing green initiatives using a multitier AHP approach, Benchmark. Int. J. 17
(2010) 452–463.
[33] S. Shaw, D.B. Grant, J. Mangan, Developing environmental supply chain performance measures, Benchmark. Int. J. 17 (2010) 320–339.
[34] K.H. Van Dam, A. Adhitya, R. Srinivasan, Z. Lukszo, Critical evaluation of paradigms for modelling integrated supply chains, Comput. Chem. Eng. 33
(2009) 1711–1726.
[35] S.E. Fawcett, C. Allred, G.M. Magnan, J. Ogden, Benchmarking the viability of SCM for entrepreneurial business model design, Benchmark. Int. J. 16
(2009) 5–29.
[36] W.S. Randall, M.T. Farris, Utilizing cash-to-cash to benchmark company performance, Benchmark. Int. J. 16 (2009) 449–461.
[37] P. Hadaya, Benchmarking firms’ operational performance according to their use of internet-based interorganizational systems, Benchmark. Int. J. 16
(2009) 621–639.
[38] M.E. Gonzalez, G. Quesada, K. Gourdin, M. Hartley, Designing a supply chain management academic curriculum using QFD and benchmarking, Quality
Assurance Educ. 16 (2008) 36–60.
[39] P. Niemi, J. Huiskonen, An approach to improving logistical performance with cross-unit benchmarking, Benchmark. Int. J. 15 (2008) 618–629.
[40] K. Bichou, Review of Port Performance Approaches and a Supply Chain Framework to Port Performance Benchmarking Research in Transportation
Economics, Chapter 24, 17 (2006) 567–598.
[41] A. Rathore, B. Balaraman, X. Zhao, J. Venkateswaran, Y. Son, R.A. Wysk, Development and benchmarking of an epoch time synchronization method for
distributed simulation, J. Manufact. Syst. 24 (2005) 69–78.
[42] T.M. Simatupang, R. Sridharan, A benchmarking scheme for supply chain collaboration, Benchmark. Int. J. 11 (2004) 9–30.
[43] C. Basnet, J. Corner, J. Wiense, K. Tan, Benchmarking supply chain management practice in New Zealand, Supply Chain Manage. Int. J. 8 (2003) 57–64.
[44] W.T. Polese, Measuring the success of collaboration across the virtual supply chain through performance measurement systems and benchmarking, in:
The Supply Chain World Conference and Exposition, New Orleans, LA, 2002.
[45] R. Van Landeghem, K. Persoons, Benchmarking of logistical operations based on a causal model, Int. J. Oper. Prod. Manage. 21 (2001) 254–267.
[46] D. Deprins, L. Simar, H. Tulkens, Measuring labor efficiency in post offices, in: M. Marchand, P. Pestieau, H. Tulkens (Eds.), The Performance of Public
Enterprises: Concepts and Measurement, Elsevier, Amsterdam, 1984, pp. 243–267.
[47] S.C. Ray, Data Envelopment Analysis: Theory and Techniques for Economics and Operations Research, Cambridge University Press, New York, 2004. pp.
134–158.
[48] W.W. Cooper, L.M. Seiford, K. Tone, Data Envelopment Analysis: A Comprehensive Text with Models, Applications, References and DEA-Solver
Software, second ed., Springer, New York, 2007.
[49] R. Farzipoor Saen, Technology selection in the presence of imprecise data, weight restrictions, and nondiscretionary factors, Int. J. Adv. Manufact.
Technol. 41 (2009) 827–838.
[50] W.W. Cooper, K.S. Park, G. Yu, An illustrative application of IDEA (imprecise data envelopment analysis) to a Korean mobile telecommunication
company, Oper. Res. 49 (2001) 807–820.
[51] Y.K. Lee, K.S. Park, S.H. Kim, Identification of inefficiencies in an additive model based IDEA (imprecise data envelopment analysis), J. Comput. Oper.
Res. 29 (2002) 1661–1676.
[52] Y.M. Wang, R. Greatbanks, J.B. Yang, Interval efficiency assessment using data envelopment analysis, Fuzzy Sets Syst. 153 (2005) 347–370.
[53] D. Despotis, Y. Smirlis, Data envelopment analysis with imprecise data, Eur. J. Oper. Res. 140 (2002) 24–36.
[54] J. Zhu, Imprecise data envelopment analysis (IDEA): a review and improvement with an application, Eur. J. Oper. Res. 144 (2003) 513–529.
[55] G.R. Jahanshahloo, R.K. Matin, A.H. Vencheh, On FDH efficiency analysis with interval data, Appl. Math. Comput. 159 (2004) 47–55.

You might also like