Journal of Cleaner Production 11 (2003) 397–409

www.cleanerproduction.net

A strategic decision framework for green supply chain

management
Joseph Sarkis ∗
Clark University, Graduate School of Management, 950 Main Street, Worcester, MA 01610, USA

Accepted 10 May 2002

Abstract

The number of organizations contemplating the integration of environmental practices into their strategic plans and daily oper-
ations is continuously increasing. Numerous initiatives have provided incentives for organizations to become more environmentally
benign. Some of these regulation driven programs are mandatory, but increasingly numerous voluntary environmental programs are
also introduced by organizations. Organizations view many of these environmental programs, which may include technological and
organizational development projects, as possible alternatives for gaining or maintaining a competitive advantage. One environmental
program area that continues to gain in importance is one that focuses on the external relationships among organizations. To help
evaluate alternatives that will effect this relationship we present a strategic decision framework that will aid managerial decision-
making. This decision framework is based on literature and practice in the area of environmentally conscious business practices.
The focus of this paper will be on the components and elements of green supply chain management and how they serve as a
foundation for the decision framework. We shall explore the applicability of a dynamic non-linear multiattribute decision model,
defined as the analytical network process, for decision making within the green supply chain. Issues facing the modeling approach
are also discussed.
 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Strategic decision making; Natural environment; Analytic hierarchy/network process; Supply chain management

1. Introduction with strong internal and external linkages. One approach

to model the dynamic nature of business and its relation-
Environmentally conscious business practices have ship to the natural environment into a decision frame-
been receiving increasing scrutiny from both researchers work is a technique that is capable of considering the
and practitioners. Interdisciplinary research has inte- multidimensional qualitative and strategic character-
grated the efforts of management, engineering, physical istics.
and social sciences to investigate the issues relevant to This paper identifies and structures the primary stra-
this topic. Similarly, multifunctional groups within tegic and operational elements for a framework that will
organizations and external stakeholders have a role in aid managers in evaluating green supply chain alterna-
decisions related to organizations and the natural tives. These alternatives may include such factors of who
environment. When organizational environmental to partner with, what type of technology to introduce, or
decisions are to be made they will necessarily be stra- what type of organizational practice to adopt. The
tegic and usually more complex for this reason. These decision to adopt one of these alternatives will be neces-
decisions will have internal and external implications for sary for an effective green supply chain, but will be
the management of an organization. Green supply chain dependent on a number of factors and elements. The
decisions are one of the latest issues facing organizations structure that is developed in this paper is a “network
hierarchy” that can be used to evaluate these alternatives.
The technique for analyzing the decision is based on the
∗
Corresponding author. Tel.: 508-793-7659; fax: 508-793-8822. analytical network process (ANP) or the systems-with-
E-mail address: [email protected] (J. Sarkis). feedback approach first introduced by Saaty [13]. The

0959-6526/02/$ - see front matter.  2002 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0959-6526(02)00062-8
398 J. Sarkis / Journal of Cleaner Production 11 (2003) 397–409

dynamic characteristics and complexity of this decision Xerox, and Digital Equipment Corporation have intro-
environment (which is true for most strategic decisions) duced some form of initiative for greening their supply
makes the ANP technique a suitable tool. Managerial chains including the integration of suppliers, distributors,
decision-making is supported through the application of and reclamation facilities [1,2,4,9]. With the increasing
this tool. Issues and possible extensions to the ANP acceptance of ISO 14001 environmental standards, there
approach identify some of its application limitations and is a greater role for supply chain management in organi-
its flexibility. zational environmental practice [15]. Realizing the sig-
The paper’s flow begins with an introduction into the nificance of these practical examples of organizations
issues relevant to green supply chains and their manage- involved in managing green supply chains, a tool that
ment. The paper then structures the various elements of will aid in managerial decision making in this complex
the green supply chain into a decision framework. An environment will be beneficial to management and
illustrative example is used to explore the application of decision makers. We shall describe some important
the ANP technique to this problem. A number of issues elements of a multidimensional decision environment
and future directions are summarized in the final sections that faces management. Included among the various
of this paper. elements are influences and relationships of the product
life cycle, operational life cycle, organizational perform-
ance measurements, and environmentally conscious
2. Green supply chain management business practices. These elements serve as the foun-
dation for a decision framework for prioritizing or sel-
Industrial Ecology has been gaining popularity among ecting systems by the organization that will aid in man-
the corporate and research communities [5]. Lowe [8] aging green supply chains.
defines industrial ecology as, “a systematic organizing
framework for the many facets of environmental man- 2.1. Product life cycle influence
agement. It views the industrial world as a natural sys-
tem—a part of the local ecosystems and the global bios- An organizational strategic factor that will influence
phere. Industrial ecology offers a fundamental the management of a supply chain is the product life
understanding of the value of modeling the industrial cycle positioning of the product(s) of an organization.
system on ecosystems to achieve sustainable environ- The typical product life cycle is comprised of four
mental performance.” phases; a product introduction phase that is characterized
An industrial ecology (ecosystem) has been defined to by investment in product research and development, a
exist on three levels [6]. These levels are characterized growth phase characterized by increasing production
by the amount of recycling or reuse of material that is capacity and logistics channels, a maturity phase, where
within the system (or the system’s “openness”). The first process and cost efficiencies are typically implemented,
level is a completely closed system with no material or and a decline phase where the focus is on product
energy leaving the system. The third level is a com- divestment.
pletely open system with little material or energy, once The product life cycle phase will necessarily impact
consumed, remaining within a system. The second level the greening of the supply chain. For example, in the
is characterized by some factor of energy and material introductory phases, the product is more greatly influ-
is reused within the system. The second level seems to enced by the design, and design for the environment
be the most applicable model for actual systems. It is issues will play a larger role at this stage. In the mature
within these industrial ecosystems models that green and decline stages of the product life cycle the improve-
supply chains will play a critical and practical role. ment of processes and having an efficient reverse logis-
Some partial industrial ecosystems are currently in tics system in place will impact the environmental prac-
operation. For example, a privately organized industrial tices of the organization. For a multiproduct analysis,
ecosystem model exists in Kalundborg, Denmark. The environmental management decisions become increas-
Kalundborg industrial ecosystem consists of a network ingly complex. But, within the product portfolio of the
of organizations composed of an electric power generat- company there should be differential environmental stra-
ing plant, an oil refinery, a biotechnology production tegies and development product life cycle foci which
plant, a plasterboard factory, a sulfuric acid producer, will be depend upon the products’ life cycle maturity.
cement producers, local agriculture and horticulture, and
district heating utilities [17]. Domestically, government 2.2. The operational life cycle
agencies have supported research and development of a
key element of industrial ecosystems called eco-indus- A more tactical set of organizational elements that will
trial parks where ecologically complementary organiza- influence how the supply chain is to be managed (either
tions are physically located in a regional area. internally or externally) can be described by the oper-
Private organizations such as Hewlett–Packard, IBM, ational life cycle (or value chain) of an organization. The
 J. Sarkis / Journal of Cleaner Production 11 (2003) 397–409 399

major elements of the operational life cycle will typically chain. Reverse logistics has also been studied from the
include procurement, production, distribution and perspective of returned and warranted items that may not
reverse logistics. We will also include packaging as an even have been used. This study of reverse logistics may
element within the operational life cycle. Packaging may be considered a subset of the environmental reverse
not be viewed as a typical stand alone operation but its logistics. Pohlen and Farris [12] in a study of the plastics
profound impact on the supply chain allows for its reverse logistics process have identified a number of
inclusion. stages within a reverse logistics channel. Included are:
The procurement or purchasing decisions will impact collection, separation, densification, transitional pro-
the green supply chain through the purchase of materials cessing, delivery, and integration. Thus, not only does
that are either recyclable or reusable, or have already there have to be a network for the reverse logistics col-
been recycled. The selection of vendors will also be an lection process, but a number of systems and processes
important decision at this stage. Vendors who have ISO may need to exist for the stages in the reverse logistics
14000 certification may be preferable since there is an channel. Depending on the organization, industry, and
expectation that the environmental risks associated with product type, the requirements may vary among the
these vendors is lessened (which is an analogous argu- stages.
ment to selecting vendors that are ISO 9000 certified Packaging has a strong relationship with other compo-
when reducing risk of poor quality purchases). nents of the operational life cycle. Packaging character-
Reduction of these risks improves the probability that istics such as size, shape and materials have an impact
these vendors will also be available for the long term. on distribution due to their affect on the transport charac-
Whether or not to outsource certain processes or compo- teristics of the good. Better packaging, along with
nents may also be a concern for the procurement depart- rearranged loading patterns, can reduce materials usage,
ment. increase space utilization in the warehouse and in the
Production processes can influence the greening of the trailer, and reduce the amount of handling required. Sys-
supply chain in numerous ways. Some of these impacts tems that encourage and adopt returnable packaging will
include: a process’ capability to use certain materials, require a strong customer supplier relationship as well
capabilities to integrate reusable or remanufactured as an effective reverse logistics channel. With JIT spe-
components into the system (which would require disas- cial kit packaging requirements will also be needed.
sembly capacities), and how well the processes are Efficiencies in packaging directly effect the environ-
designed for the prevention of waste. It is within this ment. In some countries, take-back legislation on pack-
function that much of the environmentally sound techol- aging has made the packaging operation and planning a
ogical and process innovations are most advanced (see critical environmental logistics consideration.
[3]; p. 14). This focus of environmental innovation could
be due to the fact that the production element of the 2.3. Environmentally influential organizational
operational cycle is the most internally focused for the practices
organization, allowing the organization to more directly
observe the benefits of any new technology or process There are a number of possible classifications for
that is introduced. environmentally conscious business practices. We focus
Distribution and transportation operations networks on five major practices or elements that will impact the
are also important operational characteristics that will waste generated by a supply chain. These practices
affect the green supply chain. A number of decisions (ordered on a most to least preferable environmental
including distribution outlet locations, mode of transpor- impact scale) include reduction (reduce), reuse, remanu-
tation to be used, control systems, and just-in-time poli- facture, recycle, and disposal alternatives.
cies, will not only influence the forward logistics net- Reduction is viewed as an in-process, relatively proac-
work, but also the reverse logistics network. Distribution tive, measure that can be taken by organizations. Typical
is also the operation that is most closely tied to the programs that may aid in this process include total qual-
characteristics and requirements of the customer. Thus, ity management and JIT programs that seek to minimize
customer involvement in distribution systems design and waste. Introduction of alternative processes and
development will more likely provide an effective and materials may be used to reduce more hazardous
efficient distribution network. For, example linking materials. Another example of reduction would be to
location decisions to those of vendors and customers will design the product and process to take into consideration
improve JIT systems. environmental factors (also defined as design for the
The reverse logistics operation is probably the least environment).
developed and studied of the operational functions. The End-of-pipe practices include the remaining four
definition of reverse logistics from an environmental per- elements. Reuse, remanufacture and recycle practices are
spective focuses primarily on the return of recyclable or similar, but only vary in degree of reuse of the material.
reusable products and materials into the forward supply Reuse typically keeps the original physical structure of
400 J. Sarkis / Journal of Cleaner Production 11 (2003) 397–409

the material with little substitution. Remanufacturing these performance measures is that they are not static.
requires some disassembly and replacement of parts or They tend to change over time and will be greatly influ-
components around a core. Recycling can take on new enced by the product life cycle. That is, in the introduc-
physical and chemical characteristics of the product. tion phases, flexibility and time may be more important
Each of these practices may require varying organiza- than cost. Whereas cost efficiencies tend to gain impor-
tional processes and technology. For example, disas- tance in more mature environments. These dynamical
sembly technology would probably be more preferable characteristics are incorporated into the decision frame-
for a remanufacturing practice than for a reuse practice. work.
Reuse may require more cleaning type systems than
recycling requires. Clearly, the choice of which practice
is best for an organization will depend on the organiza- 2.5. Green supply chain alternatives
tion and product characteristics. The issue of disposal
systems and vendors may have the most long-term sig-
nificance, where disposal of many materials, if not prop- Up to this point we have identified a few alternatives
erly treated, may return to haunt an organization in its available to organizations for improving the environ-
future. mental performance of their supply chains. These alter-
A summary of the possible relationships between natives may include technological, process, or organiza-
operational life cycle and environmentally conscious tional characteristics. For example one such alternative
organizational practices are shown in a flow diagram in might be an organizational goal to improve the total
Fig. 1. This diagram of the cycles is typical for a single quality environmental management (TQEM) [10] within
organization. A chain of these figures can be developed and between organizations. Similar to total quality man-
that show the relationships among a number of organiza- agement, TQEM is a pervasive program which should
tions. Indeed, feedback arrows shown in the figure may
include suppliers and customers. ISO 14000 certification
represent a number of organizations that are involved in
may also be a goal for the organization and its suppliers.
a reverse logistics process.
This alternative is based on maintaining documentation
and building an information network. Some organiza-
2.4. Organizational performance requirements
tions who have already gone through ISO 9000 certifi-
To complete this decision framework specific organi- cation may find this alternative easy to implement with
zational performance requirements are included. The cat- little additional cost and effort, and thus may be prefer-
egorization of elements for these requirements for our able to other alternatives. Other alternatives may be
example includes cost, quality, time and flexibility. information systems such as electronic data interchange
These generic strategic performance requirements, which may be justified for other reasons, but can be
which may not be environmentally based, are necessary evaluated from a greening perspective. These three
to help identify how well various alternatives can per- examples are only a few, emerging technologies, models,
form on these factors. They are necessary because the and processes that have yet to be developed can be
alternative that is selected should not only best support evaluated using the proposed decision framework as
the green supply chain, but also make business sense. well. A good discussion of various systems, require-
The use of these organizational performance measures ments and alternatives that can aid the development of
have been supported by a number of strategic thinkers green supply chains can be found in [15,19]. The
(see [7] and [18], for example). One characteristic of decision framework is now presented.

Fig. 1. Functional model of an organizational supply chain with environmentally influential practices (adapted from Sarkis [14]).
 J. Sarkis / Journal of Cleaner Production 11 (2003) 397–409 401

3. The decision framework operational life cycle element will be dependent on what
stage of the product life cycle is being considered. In
The decision framework is represented by an “analyti- addition, the importance of each product life cycle with
cal network hierarchy”, which varies from a standard respect to a given operational element will also be
decision hierarchical decision structure as defined by the determined (e.g. the early stages of the product life cycle
standard analytical hierarchy process. The variation will have more of an influence on the procurement oper-
occurs primarily because two way and ‘looped’ relation- ations than the decline stages). This model will also
ships are allowed among the various levels. These levels assume that there is internal feedback and dependencies
may also be defined as clusters. These relationships rep- among each of the operational life cycle elements. For
resent multiple dependencies and interdependencies example, some organizations will require procurement
among the elements within the clusters. Fig. 2 shows a capabilities if they plan on producing products more so
“high level” description of the analytical hierarchy net- than if they were focused on distribution as their major
work, which does not detail the components within business. These interdependencies within a cluster will
each cluster. have to be determined for each organization and will be
The objective or goal of the organization, which impacted by organizational structure.
appears on the right hand side of Fig. 2, is to develop The various environmental practices may also play
improved green supply chains. This objective will be distinct roles within the operational life cycle. An
influenced by the various clusters that have been pre- organization, due to its product’s characteristics, may
viously described in this paper. An example set of want to focus its efforts on reduction in the production
relationships among the clusters is shown. These operation rather than reuse. The effect would be to give a
relationships may vary due to assumptions made by the relatively larger importance valuation on reduction than
decision-maker, and the level of complexity that they reuse for the production element. Similarly, for the pack-
wish to model. The arrows represent the relationships aging portion, the reuse capability may be more
among the clusters. For example, the performance meas- important than disposal, and so on.
ures and their relative priority or importance (which will A detailed representation of all the clusters, the
be determined using the ANP technique described elements within the clusters, and their relationships are
below) will be dependent upon the stage of the product shown in Fig. 3. The terms within the parentheses below
life cycle. Another set of relationships exists between the each component in Fig. 3 are used to denote each of
organizational performance measures and the operational the elements within the supermatrix, which is discussed
life cycle elements. The relative importance of different below. The letters in parentheses near each arc rep-
performance measures may be allowed to vary among resents a sub-matrix that will represent the relationships
the operational life cycle elements. For example, quality within the supermatrix.
may be more important than the time performance meas-
ure for the packaging function. Within the network, the
relative impact (importance) of each alternative will be 4. Evaluating the analytical network hierarchy
evaluated for both the performance measures and the
environmental practices. The evaluation methodology will be composed of two
The most complicated set of relationships in this phases. The first phase will focus on the development of
model exist for the operational life cycle cluster. There pairwise comparisons for each of the dependency
is a two-way dependency between the operational and relationships to determine their relative importance
product life cycles. The relative importance of each weights. These weights will be used as an input to the

Fig. 2. High level graphical representation of clusters and influence relationships for decision framework for managing and improving the green
supply chain.
402 J. Sarkis / Journal of Cleaner Production 11 (2003) 397–409

Fig. 3. Graphical representation of relationships for the green supply chain evaluation framework.

systems-with-feedback supermatrix to help determine is “how much more important is procurement than pro-
the network influences from among the various relation- duction operations in the introductory phases of a pro-
ships diagrammed in Figs. 2 and 3. The supermatrix duct’s life cycle?” In this case, since we have 5 depen-
evaluation, the second phase, will encompass three steps, dent level elements within this cluster, 10 pairwise
the formation of the supermatrix, the normalization of comparison questions need to be answered for a com-
the supermatrix (making it “column stochastic”) and plete set of comparisons. An illustrative pairwise com-
convergence to a solution. The converged supermatrix parison matrix that shows the relative importances of the
will provide us with the relative priorities for each of operational life cycle elements within the introductory
the alternatives considered within the decision frame- phase of a product life cycle is shown in Table 1.
work. To maintain brevity exposition of the calculations The valuation scales used in the example are those
will only describe examples of the major steps in the recommended by Saaty [13], where 1 is equal impor-
two stage process. In addition, some example questions tance, 3 is moderate importance, 5 is strong importance,
that would be used to elicit pairwise valuations are 7 is very strong or demonstrated importance, and 9 is
also presented. extreme importance. Even numbered values will fall in
between importance levels. Reciprocal values (e.g. 1/3,
4.1. Pairwise comparison evaluations 1/5, etc.) mean less importance, strongly less impor-
tance, etc. Only the upper triangle of the matrix needs
The fundamental decision maker inputs required for to be completed. The lower triangle of the pairwise com-
the ANP technique are the pairwise comparisons of the parison matrix is composed of reciprocal values. For
elements within each cluster, from which pairwise com- example, in Table 1, we see that procurement is moder-
parison matrices are formed. These pairwise comparison ately more important than distribution during the intro-
matrices and their valuation elicitation are similar to ductory phase of a product life cycle. Thus, the value
those that are used for the AHP approach. For details on for a13=3, whereas a31=1/3 (or distribution is moderately
the use of AHP and its various calculations see [13]. A less important than procurement during the introductory
pairwise comparison matrix is required when the relative phase of the product life cycle). Once all the pairwise
importance of lower level elements are to be determined comparisons are complete, the relative importance
for thier ‘controlling’ element. For example, to deter- weight for each component is determined (these results
mine the relative importance of the operational life cycle are shown in the final column of Table 1). Given that A
elements (the lower level elements) to the introduction is the pairwise comparison matrix, the weights can be
phase of the product life cycle elements (the controlling determined by expression (1).
element) a number of pairwise comparison questions
will be asked of the decision-maker. Once such question Aw ⫽ lmaxw, (1)
 J. Sarkis / Journal of Cleaner Production 11 (2003) 397–409 403

Table 1
Pairwise comparison matrix for operational life cycle elements relative importances during introduction phase of the product life cycle

Introduction phase Procurement Production Distribution Reverse logistics Packaging Importance wts.

Procurement 1.000 6.000 3.000 8.000 2.000 0.448

Production 0.167 1.000 0.500 2.000 0.200 0.073
Distribution 0.333 2.000 1.000 4.000 0.500 0.151
Reverse logistics 0.125 0.500 0.250 1.000 0.167 0.043
Packaging 0.500 5.000 2.000 6.000 1.000 0.285

where λmax is the largest eigenvalue of A and w is the

relative importance weights or priority vector (actually
the eigenvectors for the principal eigenvalue λmax). The
values were calculated using MATLAB software. Pair-
wise comparisons leave room for possible intransitive
preference relationships and inconsistencies. A consist-
ency index (C.I.) and consistency ratio (C.R.) also need
to be calculated. The consistency index for a pairwise
comparison matrix is determined by:
lmax⫺n
C.I. ⫽ (2)
n⫺1
where n is the number of components that are evaluated
in the pairwise comparison matrix. The C.R. is calcu-
lated by taking the C.I. and dividing by a random incon-
sistency (R.I.) value. A random inconsistency table exists
in most AHP and ANP reference books, see [13]. For a
pairwise comparison matrix to be consistent, C.R.⬍0.10. Fig. 4. General submatrix notation for supermatrix.
In Table 1 we show the values for λmax, C.I., and C.R.
We see that this is a relatively consistent set of weights. mination of the relative impact of each operational life
The priority vector shows that for this organization and cycle phase on each of the four product life cycle stages.
industry, procurement (0.448), followed by packaging Four priority vectors will be required to complete C. We
(0.285), seem to be the functions that are deemed most have already shown one set of priority weights for C this
important for the early stages of a product’s life cycle. vector begins in the second column and sixth row of the
All these values can be easily calculated using AHP initial supermatrix, which appears in bold lettering in
software such as Expert Choice. Table 2.
The priority vectors for each pairwise comparison
matrix will be needed to complete the various superma- 4.3. The solution procedure
trix submatrices. We will need a total of 39 priority vec-
tors to complete our supermatrix. This requirement The supermatrix M is a reducible matrix with a mul-
means that 39 pairwise comparison matrices must be tiple root, as defined by Saaty [13]. To solve for the
completed. The pairwise comparison matrix results used values of the alternatives, Saaty recommends that the
below were all tested for achieved the consistency goals. values of M be column stochastic. That is, the sums of
the columns should be normalized to equal a value of
4.2. Supermatrix formation 1. One method of making M column stochastic is by
determining the relative importances of clusters and mul-
The supermatrix (M) is formed from a number of sub- tiplying their relevant matrix elements by their relative
matrices that are used to model Figs. 2 and 3 in matrix importance score. In this case we just assumed that all
notation. The supermatrix and its general submatrices are clusters were of equal importance. Two adjustments will
shown in Fig. 4. There will be 9 sub-matrices (A, B, C, need to be completed for the supermatrix to be translated
D, E, F, G, H and J) that will be formed using the pri- into a column stochastic matrix. The first adjustment
ority vectors. An additional identity sub-matrix (I) is influences the operational life cycle and performance
added for the alternatives cluster for computational measure clusters and their impact on the product life
requirements. cycle elements. Since there are two clusters, each rep-
The formation of sub-matrix C will require the deter- resentative submatrix, in this case submatrices C and E,
 404

Table 2
Initial supermatrix (M) for selection of alternatives for improvement of green supply chain illustration

GOAL Product life cycle Operational life cycle Performance measures Environmental practices Alternatives
GSC1 PLC1 PLC2 PLC3 PLC4 OLC1 OLC2 OLC3 OLC4 OLC5 PERF1 PERF2 PERF3 PERF4 ENV1 ENV2 ENV3 ENV4 ENV5 ALTA ALTB ALTC

GOAL GSC1 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Product life cycle PLC1 0.000 0.000 0.000 0.000 0.000 0.373 0.060 0.176 0.053 0.079 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PLC2 0.000 0.000 0.000 0.000 0.000 0.108 0.245 0.540 0.123 0.255 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PLC3 0.000 0.000 0.000 0.000 0.000 0.095 0.568 0.235 0.354 0.588 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PLC4 0.000 0.000 0.000 0.000 0.000 0.424 0.127 0.050 0.470 0.079 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Operational phases OLC1 0.189 0.448 0.104 0.075 0.159 0.000 0.287 0.065 0.532 0.102 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
OLC2 0.052 0.073 0.248 0.521 0.052 0.408 0.000 0.242 0.076 0.107 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
OLC3 0.186 0.151 0.463 0.077 0.074 0.059 0.113 0.000 0.291 0.204 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
OLC4 0.394 0.043 0.070 0.186 0.562 0.147 0.547 0.242 0.000 0.586 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
OLC5 0.179 0.285 0.116 0.140 0.153 0.386 0.053 0.451 0.101 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Performance criteria PERF1 0.000 0.538 0.181 0.089 0.141 0.131 0.101 0.537 0.067 0.130 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PERF2 0.000 0.105 0.069 0.285 0.141 0.275 0.538 0.210 0.067 0.328 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PERF3 0.000 0.054 0.087 0.575 0.263 0.545 0.284 0.210 0.522 0.502 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PERF4 0.000 0.302 0.662 0.051 0.455 0.050 0.077 0.042 0.343 0.041 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Environmental Practices ENV1 0.000 0.000 0.000 0.000 0.000 0.046 0.508 0.195 0.043 0.153 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
ENV2 0.000 0.000 0.000 0.000 0.000 0.494 0.096 0.048 0.537 0.340 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
ENV3 0.000 0.000 0.000 0.000 0.000 0.249 0.244 0.088 0.188 0.034 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
ENV4 0.000 0.000 0.000 0.000 0.000 0.164 0.092 0.090 0.186 0.399 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
J. Sarkis / Journal of Cleaner Production 11 (2003) 397–409

ENV5 0.000 0.000 0.000 0.000 0.000 0.046 0.06 0.580 0.046 0.074 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Alternatives ALTA 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.075 0.103 0.637 0.200 0.117 0.200 0.158 0.095 0.333 1.000 0.000 0.000
ALTB 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.333 0.216 0.105 0.600 0.268 0.600 0.082 0.250 0.334 0.000 1.000 0.000
ALTC 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.592 0.682 0.258 0.200 0.614 0.200 0.761 0.655 0.333 0.000 0.000 1.000
 J. Sarkis / Journal of Cleaner Production 11 (2003) 397–409 405

are multiplied by 0.5. The second adjustment will be for as well as the mathematical approach to solve the prob-
the four clusters that influence the operational life cycle lem. Using additive techniques to determine the solution
represented by submatrices A, D, F and G. Assuming, requires less computational effort and may be slightly
once again, that each cluster equally impacts the oper- more intuitive. Yet, more accuracy and consistency gen-
ational life cycle, we multiply submatrices A, D, F and erally occurs with the supermatrix solution procedure.
G, by 0.25. The adjusted column stochastic supermatrix The analysis of the supermatrix may not be as intuitive
(Ms) is shown in Table 3. to the decision maker (who usually care about under-
The final step in the process is to obtain a priority standing multiple criteria model capabilities, see [11]) as
ranking for each of the alternatives. We will determine the additive approach.
this ranking by calculating the influence of each of the There are some managerial aides that the supermatrix
alternatives on the objective of improving the green sup- does provide through its summarized structure. One of
ply chain. Saaty states that a simple hierarchy and the these aides is a summary of the various linkages and
additive solution approach is appropriate if strong depen- relationships. This summary allows managers to deter-
dencies among the criteria do not exist. But, in this case mine what patterns might exist among the various
the dependencies are considered to be strong. In relationships. For example, a management team is able
addition, Schenkerman [16] has shown that the superma- to look at submatrices C and E to see how the priorities
trix approach is capable of reducing the occurrence of of the organization are shifting over a product’s life
rank reversal, thus providing more accurate portrayals of cycle. The shifting of priorities can be monitored and
decision-maker preferences. Saaty recommends a simple evaluated by observing this supermatrix. This obser-
solution technique to solve this problem by raising the vation of the managerial significance of the supermatrix
supermatrix Ms to a large power until convergence has implications for sensitivity analysis. Yet, sensitivity
occurs. In this illustration we only needed to raise the analysis with the ANP approach is still relatively com-
supermatrix to a power of 16 (M16 s ) before convergence plex when compared to AHP. The computational and
occurred within the fourth decimal place (i.e. 10⫺4). The data requirements of ANP still make “what–if” analysis
converged supermatrix is shown in Table 4. The relative geometrically more cumbersome than AHP. There is a
influences of the alternatives on the objective of improv- need for investigating both sensitivity and what–if analy-
ing the environmental performance of the supply chain sis for ANP to determine both its strengths and limi-
are shown in the “Goal” column. The results show that tations in this area. This investigation is left for future
alternative B has a higher priority score than alternative study.
C which is better than the current situation (with scores An interorganizational application of this decision
of 0.373, 0.353 and 0.276). framework will have to incorporate the perceptions of a
number of stakeholders. Not only will there be diverse
preferences and perceptions within an organization, but
5. Discussion also those of other organizations. Aggregation tech-
niques for AHP (e.g. geometric averaging, consensus
The ANP approach, in practical application, requires scores) can currently be used for each submatrix column
significant decision maker input. Its application needs to in the supermatrix. Alterations to the supermatrix and an
be targeted to those areas where strategic decision mak- addition to the decision framework incorporating a
ing is required. For example, using this technique for “firm” control hierarchy, where relative organization
selecting a minor piece of equipment or a day-to-day influences or impact determinations are made can incor-
decision, may not be necessary. Its use should be limited, porate some of the diversity of opinion or preferences
especially if its use in the decision making process costs among the organizations.
more than the outcome of the decision. Yet, for the case The decision framework has only modeled internal
of greening the supply chain, the decision is strategic influences and relationships. A number of external fac-
and will broadly effect the operations of not just one, tors could be introduced into the model. For example,
but many organizations. Thus, the investment in making external factors such as potential for new environmental
a decision that would profoundly effect the operation of regulations or cooperation among competing supply
the supply chain clearly requires intensive and robust chains could also be integrated. The type of environmen-
managerial analysis. Considering critical factors and tal forces such as remediation, command and control,
their interdependencies is necessary for accomplishing or cooperative regulatory policies may also be modeled.
this goal. These models may be formed as control hierarchies or
One important consideration in the effectiveness and as part of network hierarchies for decision modeling pur-
efficiency of the decision framework begins at the mode- poses. A control hierarchy has the characteristic of being
ling stages. The model and the various dependencies will separate from a network interdependency model, where
determine the amount of effort required to arrive at a the results can be aggregated using an additive model.
solution. This effort includes input from decision-makers The network hierarchy would be integrated with the net-
 406

Table 3
Column stochastic supermatrix (Ms) for selection of alternatives for improvement of green supply chain illustration

GOAL Product life cycle Operational life cycle Performance measures Environmental practices Alternatives
GSC1 PLC1 PLC2 PLC3 PLC4 OLC1 OLC2 OLC3 OLC4 OLC5 PERF1 PERF2 PERF3 PERF4 ENV1 ENV2 ENV3 ENV4 ENV5 ALTA ALTB ALTC

GOAL GSC1 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Product life cycle PLC1 0.000 0.000 0.000 0.000 0.000 0.093 0.015 0.044 0.013 0.020 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PLC2 0.000 0.000 0.000 0.000 0.000 0.027 0.061 0.135 0.031 0.064 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PLC3 0.000 0.000 0.000 0.000 0.000 0.024 0.142 0.059 0.088 0.147 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PLC4 0.000 0.000 0.000 0.000 0.000 0.106 0.032 0.012 0.118 0.020 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Operational phases OLC1 0.189 0.224 0.052 0.038 0.079 0.000 0.072 0.016 0.133 0.026 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
OLC2 0.052 0.037 0.124 0.261 0.026 0.102 0.000 0.060 0.019 0.027 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
OLC3 0.186 0.076 0.231 0.038 0.037 0.015 0.028 0.000 0.073 0.051 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
OLC4 0.394 0.022 0.035 0.093 0.281 0.037 0.137 0.060 0.000 0.147 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
OLC5 0.179 0.143 0.058 0.070 0.076 0.097 0.013 0.113 0.025 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Performance criteria PERF1 0.000 0.269 0.091 0.044 0.070 0.033 0.025 0.134 0.017 0.032 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PERF2 0.000 0.053 0.035 0.142 0.070 0.069 0.135 0.053 0.017 0.082 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PERF3 0.000 0.027 0.044 0.288 0.131 0.136 0.071 0.053 0.131 0.126 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PERF4 0.000 0.151 0.331 0.026 0.228 0.012 0.019 0.011 0.086 0.010 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Environmental practices ENV1 0.000 0.000 0.000 0.000 0.000 0.012 0.127 0.049 0.011 0.038 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
ENV2 0.000 0.000 0.000 0.000 0.000 0.124 0.024 0.012 0.134 0.085 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
ENV3 0.000 0.000 0.000 0.000 0.000 0.062 0.061 0.022 0.047 0.008 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
ENV4 0.000 0.000 0.000 0.000 0.000 0.041 0.023 0.022 0.046 0.100 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
ENV5 0.000 0.000 0.000 0.000 0.000 0.012 0.015 0.145 0.011 0.019 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
J. Sarkis / Journal of Cleaner Production 11 (2003) 397–409

Alternatives ALTA 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.075 0.103 0.637 0.200 0.117 0.200 0.158 0.095 0.333 1.000 0.000 0.000
ALTB 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.333 0.216 0.105 0.600 0.268 0.600 0.082 0.250 0.334 0.000 1.000 0.000
ALTC 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.592 0.682 0.258 0.200 0.614 0.200 0.761 0.655 0.333 0.000 0.000 1.000
Table 4
Converged supermatrix (M16
s ) for selection of alternatives for improvement of green supply chain illustration. Alternative priorities are italicized

GOAL Product life cycle Operational life cycle Performance measures Environmental practices Alternatives
GSC1 PLC1 PLC2 PLC3 PLC4 OLC1 OLC2 OLC3 OLC4 OLC5 PERF1 PERF2 PERF3 PERF4 ENV1 ENV2 ENV3 ENV4 ENV5 ALTA ALTB ALTC

GOAL GSC1 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Product life cycle PLC1 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PLC2 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PLC3 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PLC4 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Operational phases OLC1 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
OLC2 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
OLC3 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
OLC4 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
OLC5 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Performance criteria PERF1 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PERF2 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PERF3 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PERF4 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
PERF5 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Environmental practices ENV1 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
ENV2 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
ENV3 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
ENV4 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
J. Sarkis / Journal of Cleaner Production 11 (2003) 397–409

Alternatives ALTA 0.276 0.210 0.235 0.344 0.285 0.277 0.249 0.248 0.290 0.279 0.077 0.103 0.657 0.200 0.121 0.200 0.158 0.095 0.333 1.000 0.000 0.000
ALTB 0.373 0.436 0.450 0.308 0.402 0.361 0.356 0.385 0.382 0.356 0.542 0.216 0.196 0.600 0.575 0.600 0.082 0.250 0.334 0.000 1.000 0.000
ALTC 0.353 0.357 0.317 0.349 0.312 0.366 0.396 0.368 0.328 0.367 0.382 0.682 0.147 0.200 0.304 0.200 0.761 0.655 0.333 0.000 0.000 1.000
407
408 J. Sarkis / Journal of Cleaner Production 11 (2003) 397–409

work model and the solution can be determined through in scope. The models and application of the ANP meth-
the supermatrix solution approach. odology, in the literature and practice, has been quite
A number of other decision factors and criteria can be limited due to its complex characteristics and need for
included in the model, yet the complexity of the decision further developments. The major disadvantage of the
environment will tend to increase. Increasing com- ANP approach is the large amount of decision-maker
plexity, even though more realistic, usually requires input required, even for rather simple networks. The
additional effort for preference elicitation from decision advantage of allowing managers and decision-makers
makers and more complex computations. The tradeoffs the flexibility to identify and incorporate major interde-
between amount of decision maker time and ‘realism’ pendencies among many factors and clusters in a
of the model need to be considered. The application of “dynamic” fashion, makes this technique a viable alter-
the ANP approach should not only be concerned with a native to AHP and other multiattribute approaches. Yet,
‘final’ solution to the problem, but it also should be with appropriate software tools, some of which are cur-
applied as a learning tool for decision makers and man- rently under development, this tool may have as much
agers to help understand the various linkages among the popularity and application as the AHP approach.
various components, clusters, and elements.
Rank reversal may still be an issue with the ANP
approach. Even though the possibilities of rank reversal References
are lessened with the supermatrix approach, total elimin-
ation of these problems needs to be studied. In addition, [1] Ashley S. Designing for the environment. Mechanical Engineer-
with the relative novelty of ANP practice and theory, a ing 1993;115(3):52–5.
number of theoretical and mathematical underpinnings [2] Bergstrom RY. An annotated essay: environmental affairs. Pro-
still necessitate development. duction 1993;105(4):36–41.
[3] Florida R. Lean and green: the move to environmentally con-
scious manufacturing. California Management Review
1996;39(1):80–105.
6. Summary and conclusion [4] Gillett J. Ensuring suppliers’ environmental performance, Pur-
chasing & Supply Management (1993), 28-30.
The issue of organizations incorporating the natural [5] Graedel TE, Allenby BR. Industrial Ecology. Englewood, NJ:
environment into strategic and operational decisions is Prentice Hall, 1995.
[6] Jelinski LW, Graedel TE, Laudise WD, McCall DW, Patel KN.
a reality that they will or have already encountered. The Industrial ecology: concepts and approaches. Proceedings of the
influences of the natural environment organizational National Academy of Sciences 1996;89:793–7.
decisions will not only effect the organization that makes [7] Kleindorfer PR, Partovi FY. Integrating manufacturing strategy
the decision, but its customers and suppliers, as well. and technology choice. European Journal of Operational Research
Incorporating various elements, functions and activities 1997;47:214–24.
[8] Lowe E. Industrial ecology—an organizing framework for
of supply chain management is one method to incorpor- environmental management. Total Quality Environmental Man-
ate some of the systemic organizational and inter-organi- agement 1990;3(1):73–85.
zational implications of environmentally influential poli- [9] Maxie E. Supplier performance and the environment. In: Inter-
cies. A number of business and environmental factors national Symposium on Electronics and the Environment, IEEE,
need to be integrated into this decision. One such San Francisco, CA. 1994.
[10] Oakley BT. Total quality product design—how to integrate
decision framework that considers these factors, whose environmental criteria into the production realization process.
goal is to improve the green supply chain, is introduced Total Quality Environmental Management 1993;2(3):309–21.
in this paper. The major elements and their relationships [11] Ozernoy VM. Choosing the “best” multiple criteria decision-
have a number of interdependencies. These elements making method. Infor 1992;30(2):159–71.
include product life cycle, operational life cycle, per- [12] Pohlen TL, Farris MT. Reverse logistics in plastics recycling.
International Journal of Physical Distribution & Logistics Man-
formance measures, and environmentally influential agement 1992;22(7):35–47.
organizational policy elements. The goal of the frame- [13] Saaty TL. The Analytic Network Process. Pittsburgh, PA: RWS
work is to help evaluate a number of alternatives Publications, 1996.
(projects, partnerships, systems or technologies, etc.) [14] Sarkis J. Manufacturing strategy and environmental conscious-
that impact these various factors. The decision frame- ness. Technovation 1995;15(2):79–97.
[15] Sarkis J. Supply chain management and environmentally con-
work is modeled and solved as an analytical network scious design and manufacturing. International Journal of
process (ANP). The formulation of the model and its Environmentally Conscious Design and Manufacturing
solution technique are described. Managerial and 1995;4(2):43–52.
decision science implications are also presented. [16] Schenkerman S. Avoiding rank reversal in AHP decision support
The ANP methodology is a robust multiattribute models. European Journal of Operational Research
1994;74(3):407–19.
decision making technique for analyzing the major [17] Schmidheiny S. Changing Course: A Global Business Perspective
issues facing green supply chains and environmentally on Development and the Environment. Cambridge, MA: MIT
conscious business practices, both of which are strategic Press, 1992.
 J. Sarkis / Journal of Cleaner Production 11 (2003) 397–409 409

[18] Wheelwright SC. Reflecting corporate strategy in manufacturing tems. International Journal of Physical Distribution & Logistics
decisions. Business Horizons 1978;20:57–66. Management 1995;25(2):20–38.
[19] Wu H-J, Dunn SC. Environmentally responsible logistics sys-