This article appeared in a journal published by Elsevier.

The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
 Author's personal copy

Omega 37 (2009) 788 – 800

www.elsevier.com/locate/omega

Single or dual sourcing: decision-making in the presence of supply

chain disruption risks
Haisheng Yua , Amy Z. Zengb,∗ , Lindu Zhaoa
a Institute of Systems Engineering, Southeast University, Nanjing, Jiangsu 210096, PR China
b Department of Management, Worcester Polytechnic Institute, Worcester, MA 01609, USA

Received 20 March 2007; accepted 20 May 2008

Available online 29 May 2008
Processed by B. Lev

Abstract
The focus of this paper is placed on evaluating the impacts of supply disruption risks on the choice between the famous single
and dual sourcing methods in a two-stage supply chain with a non-stationary and price-sensitive demand. The expected profit
functions of the two sourcing modes in the presence of supply chain disruption risks are first obtained, and then compared so that
the critical values of the key factors affecting the final choice are identified. Finally, the sensitivity of the buyer’s expected profit
to various input factors is examined through numerical examples, which provide guidelines for how to use each sourcing method.
䉷 2008 Elsevier Ltd. All rights reserved.

Keywords: Supply chain disruption; Sourcing; Supply management; Risk management; Sensitivity analysis; Decision-making/process

1. Introduction in 1997 [1]; the Taiwan earthquake of September 1999

It has been observed that the relationships between
suppliers and their immediate buyers have evolved from
fragmented, scattered links to today’s integrated, in-
terdependent supply chain networks. Although such a
change has led to numerous benefits related to efficiency
and productivity, it may also result in severe problems,
one of which is the risk of supply chain disruptions that
has been witnessed by the entire world in the past few
years. For example, Toyota production line would have
been shut down for two weeks when their sole supplier
of the brake-fluid proportioning valves had a big fire
∗ Corresponding author. Tel.: +1 508 831 6117;
fax: +1 508 831 5720.
E-mail address: azeng@wpi.edu (A.Z. Zeng).
0305-0483/$ - see front matter 䉷 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.omega.2008.05.006
created a panic and huge losses for many electronic
firms that use Taiwanese manufacturers as suppliers
[2]. The March 2000 fire at the Philips microchip plant
in Albuquerque, NM, sent their major buyers, Nokia
and Ericsson, to chaos. Fortunately, Nokia learned of
the impending chip shortage in just three days and took
advantage of its multi-tiered supplier strategy to obtain
chips from other sources. Ericsson, however, could not
avoid a production shutdown because it was sourcing
only from that plant. As a result, the company suffered
$400 million in lost sales [3]. A devastating shortage
of flu vaccine in the fall of 2004 occurred in the US
after 46 million doses produced by Chiron, one of the
only two suppliers, were condemned of bacterial con-
tamination [4]. This shortage led to rationing in most
states and significant price increases in many places.
In summary, since the tragedies of 9-11 and many
 Author's personal copy
H. Yu et al. / Omega 37 (2009) 788 – 800
subsequent disastrous events, it is becoming increas-
ingly clear to the business world that risks, both pre-
dictable and unpredictable, exist in every link of a sup-
ply chain and that effective risk management should be
on top of every management’s agenda.
A quick review of the companies’ reactions to supply
disruptions mentioned above reveals that dual sourc-
ing or using multiple suppliers can be an effective tool
in dealing with unexpected supply breakdowns. Even
though the debate on single and multiple sourcing has
been around for a couple of decades, the objectives of
the research in this area are primarily focused on cost
reduction and service level improvement. The effort of
examining these two sourcing methods in the context
of supply chain risk management is still limited. In this
paper, we would like to study how to select between
single and dual sourcing modes in the presence of sup-
ply chain disruptions.
Scholarly works have also confirmed that supply
chains can be vulnerable. For example, Wong et al. [5]
have found that today’s supply chains are built to be
lean and efficient, but if they are unable to find alter-
natives quickly for unexpected disruptions, the chains
will be susceptible to system shocks and disruptions.
The studies by Christopher and Towill [6] and Tang
[7] suggest that as many firms implement various ini-
tiatives such as lean, agile, outsourcing, customized,
and global networks to gain cost advantage and market
share, their supply chains become more vulnerable at
the same time, because there tends to be very little
inventory in the system to “buffer” any interruptions in
supply. As a result, any disruption can have a dramatic
impact on the entire chain.
With many instances of supply chain failures ob-
served in the past, managers and decision-makers are
becoming aware of the impacts of supply chain disrup-
tions risks but are still struggling in finding ways for ef-
ficient supply chain disruption management. Although
there have been numerous discussions and studies over
the years that have examined business risks and disrup-
tions in the contexts of financial planning, new-product
development, and demand changes, quantitative studies
on risks associated with suppliers and the supply net-
works are still sparse.
This paper is focused on studying a two-stage supply
chain where a supply and a purchase take place. This
is an essential transaction that occurs in every supply
chain network. We consider such a two-stage chain that
the buying firm faces a non-stationary, price-sensitive
demand of a critical component for its final products and
that two suppliers (primary and secondary) are avail-
able. The two suppliers can be geographically different
789
and offer two wholesale prices. The supply disruption
risk is captured by a known probability. We then for-
mulate a set of expected profit functions (EPFs) with
the consideration of supply disruptions when the buying
firm uses single and dual sourcing strategies, respec-
tively. Based on these EPFs, we compare the perfor-
mance of these two sourcing methods and identify the
critical factors governing the choice between the two
sourcing alternatives. The results shed lights on how
to design an appropriate sourcing method when supply
disruption risk exists in a supply chain.
The remainder of the paper is organized as follows.
Section 2 contains a review of relevant literature from
two perspectives, one examines supply chain risk and
disruption, and the other analyzes single and dual sourc-
ing strategies. Section 3 presents and analyzes the EPFs
in the presence of supply chain disruption risks. A set
of numerical analysis of the EPFs and discussions of
the associated results are given in Section 4. Finally,
the managerial implications of the research results, the
limitations of the proposed model, and some possible
future research directions are reported in Section 5.
2. Literature review
There exists a large body of literature related to sourc-
ing strategies, risk management and decision-making
under uncertainty. Considering the relevance, we review
two streams of recent research efforts here: one is fo-
cused on supply chain risk and disruption management
and the other compares single and dual sourcing meth-
ods.
2.1. Supply chain risk and disruption management
Since the tragedy of 9-11, there has been a rapidly
increasing trend in research efforts that study the
supply chain risks and disruptions both qualitatively
and quantitatively. With such a heightened awareness,
Cranfield Management School [8] conducted a large-
scale research project on the global “supply chain
vulnerability”. The vulnerability is defined as “an ex-
posure to serious disturbance, arising from risks within
the supply chain as well as risks external to the sup-
ply chain”. In the study, all types of risks, whether it
is a supply disruption, demand uncertainty or the so-
called “internal risk”, are included. Another research
effort by Tang [7] combines the definitions developed
by others (e.g., Jüttner et al. [28] and Deloitte and
Touche, http://www.deloitte.com) and defines supply
chain risk management as “the management of supply
chain risk through coordination or collaboration among
 Author's personal copy
790 H. Yu et al. / Omega 37 (2009) 788 – 800
the supply chain partners so as to ensure profitabil-
ity and continuity”. Both Svensson [9] and Hallikas
and Vivolainen [10] have found that a small customer
structure change and a short lead time could result in
supply chain risk and disruption, and furthermore, the
relationship between the buyer and the supplier affects
those risks and disruptions.
As mentioned in the final report on “supply chain
vulnerability” obtained by Cranfield Management
School [8], supply chain risks and disruptions can be
caused by a number of sources, which may include: (1)
natural disasters; e.g. the Kobe earthquake, SARS, foot
and mouth disease, birds flu, and others; (2) terrorist
incidents, e.g. the attack on September 11, 2001; (3)
industrial or direct action; e.g. the fuel price protest in
September 2000 that rapidly affected almost every sup-
ply chain in the UK; (4) unexpected accidents; e.g. a fire
at a component supplier can have such a serious impact
on the OEMs that they are forced to shut down oper-
ations; (5) operational difficulties; e.g. if one supplier
experiences a production or supply related problem,
then every downstream organization will be affected.
To support the research effort dedicated to supply
chain risks, Tang [7] classifies supply chain risks into
two categories, namely operational risks and disrup-
tion risks. An operational risk refers to those inherent
uncertainties that inevitably exist in supply chains; for
example, an uncertain customer demand, an uncertain
supply, and an uncertain cost. A disruption risk is re-
ferred to as the major disruptions caused by natural and
man-made disasters such as earthquakes, floods, hurri-
canes, and terrorist attacks, or economic crises such as
currency fluctuations or employee strikes. In their study,
Chopra and Sodhi [11] discuss several supply risks that
a manager must account for when planning suitable mit-
igation strategies. Based on this research, Chopra et al.
[12] reported in a subsequent study that bundling oper-
ational risks and disruption risks may lead to erroneous
decisions, higher inventories, and a higher supply chain
cost than optimal.
A number of recent papers focus closely on supply
chain disruptions and discuss the measures that compa-
nies should use to design better supply chains, or study
the different ways that could help buying firms to mit-
igate the consequences of a supply disruption. In the
context of an uncertain environment, a study that was
conducted as early as in the 1960s proposed an opti-
mal ordering policy to deal with the uncertainties [30].
In particular, the study shows that the optimal ordering
policy takes on the form of “two order-up-to levels”—x
and y, where x < y. The optimal ordering policy can be
described as follows: if the inventory at the beginning of
a period (z) is less than x, then order (x.z) units by us-
ing the emergency mode and order (y.x) units by using
the regular mode; if x < z < y, then order (y.z) units
according to the regular mode; otherwise, order noth-
ing. When studying the risk of supply chain disruptions,
Vlachos and Tagaras [13] extended Fukuda’s model to
the case in which the emergency model is capacitated.
Kleindorfer and Saad [14] present a supply chain risk
framework called SAM, where “S” refers to specify-
ing sources of risk and vulnerabilities, “A” means as-
sessment, and “M” stands for mitigation. A very recent
study by Xiao and Qi [15] examines the effects of price
competition, cost and demand disruptions on the coor-
dination between one manufacturer and two competing
retailers.
In summary, existing research efforts differ from each
other with respect to the definitions and viewpoints
about supply chain risk and disruption, but they all unan-
imously consider supply chain disruptions as one im-
portant question for supply chain management, and the
most urgent task is to find effective strategies to mitigate
the effects of supply chain disruption. Since sourcing
is directly related to supply chain disruptions and can
be used to proactively cope with the disruption risks,
analyzing and designing an effective sourcing method
in the presence of supply chain disruptions is becom-
ing one of the focuses and hot topics in recent research
efforts.
2.2. Single versus dual sourcing
Successful supply chain management necessitates an
effective sourcing strategy to combat unreliable supply
and stochastic demand. We divide the most frequently
used approaches of sourcing into three types: (1) single
sourcing, (2) dual sourcing, and (3) multiple sourcing.
In the first type it is necessary to mention that single
sourcing differs from sole sourcing. Sole sourcing refers
to a buyer–supplier relationship where the supply base
contains only one supplier, whereas single sourcing is
when a buyer chooses a single supplier even though
other comparable suppliers exist in the supplier base
[31]. The second sourcing strategy, dual sourcing, indi-
cates that a buyer employs two suppliers, one of which
may dominate the other in terms of business share, price,
reliability, and others. In the last sourcing model, multi-
ple sourcing, a buyer does business with several suppli-
ers and plays one supplier against the other to enjoy the
best price advantage. In this paper, we focus our atten-
tion on how to select single sourcing and dual sourcing
methods in the presence of supply disruption risks.
 Author's personal copy
H. Yu et al. / Omega 37 (2009) 788 – 800
Single sourcing strategy strives for a strategic partner-
ship between a buyer and a supplier to foster a close col-
laboration and to optimize shared benefits. The tighter
coordination between the buyer and supplier is a pre-
requisite for a successful execution of the just-in-time
(JIT) initiative, which encourages the two partners to
streamline the supply chain process and encourage the
buyer–seller relationship to move toward a single sourc-
ing model. As single sourcing has become a priority
for many firms, many researchers have studied the ad-
vantages of this strategy. The general benefits of single
sourcing, as indicated in a survey study by Larson and
Kulchitsky [16], include higher quality at lower total
cost to the buyer and that suppliers are linked to higher
levels of buyer–supplier cooperation. However, the de-
pendence on a single source also exposes the buying
firms to a greater risk of supply chain interruption. For
example, Toyota’s brake valve crisis in 1997 exempli-
fies the possible occurrence of supply chain disruption
risks that are resulted from a single sourcing strategy in
a JIT system.
With an increasing awareness of the high risks asso-
ciated with single sourcing and companies’ expanded
efforts to mitigate risks or to build in contingency, mul-
tiple sourcing is gaining attention again; especially how
to determine the optimal supply size has resulted in a
great deal of research interest. For example, Berger et al.
[17] consider risks associated with a supplier network,
which include catastrophic super events that affect all
suppliers, as well as unique events that impact only one
single supplier, and then present a decision-tree based
model (which is referred to as BGZ model in the fol-
lowing discussion) to help determine the optimal num-
ber of suppliers needed for the buying firm. Following
the same decision framework, Ruiz-Torres and Farzad
[29] present an extension of the BGZ model by consid-
ering unequal failure probabilities for all the suppliers.
They have also compared their proposed models with
the BGZ model by conducting a sensitivity analysis in
order to better understand the effect of the input param-
eters on the optimal number of suppliers. In the other
effort, Berger and Zeng [18] study the optimal supply
size under a number of scenarios that are determined
by various financial loss functions, the operating cost
functions and the probabilities of all the suppliers being
down.
With respect to the selection decision between sin-
gle and dual sourcing methods, we see a number of
important research efforts. A research by Pochard [19]
examines how buying firms should prepare for disrup-
tions in their supply chain by using real options theories
to compare single sourcing with dual sourcing. Burke
791
et al. [20] indicate that single sourcing is a dominant
strategy only when supplier capacities are large relative
to the product demand and when the buying firm does
not obtain diversification benefits; but in all other cases,
dual sourcing is an optimal sourcing strategy. Using the
concept of switching costs in a principal-agent frame-
work, Wagner and Friedl [21] analyzed whether a firm
switches single sourcing to dual sourcing when there is
either symmetric or asymmetric information about the
alternative supplier’s cost structure. They have found
that the sourcing strategies depend on the buying firm’s
beliefs in the alternative supplier’s unit costs, switch-
ing costs, the price offered by the incumbent supplier,
and refinements of the price offered by the incumbent
supplier due to competitive reactions and economies of
scale.
The other research stream examines how to split or-
ders between different suppliers. In the case of dual
sourcing, Lau and Zhao [22] identify the optimal pro-
portion of split between two suppliers by minimizing
the sum of annual holding and ordering costs subject to
a maximum allowable stockout risk. Through numeri-
cal studies, they have found that the optimal proportion
of split varies with, among other factors, the difference
in the suppliers’ mean lead times. The issue of order
split between two suppliers is also studied by Kelle and
Miller [23] in the context that the objective of decision
is to minimize stockout risk. They not only provide an
exact formula for calculating the optimal order split rate,
but also show that large lead-time demand and lead-
time uncertainty usually favor dual sourcing. In a recent
study, Tomlin and Wang [24] examine the flexibility and
reliability provided by dual sourcing using the classic
newsvendor model. They have identified a number of
factors affecting the use of the second supplier, includ-
ing the resource costs and reliabilities, the firm’s down-
side risk tolerance, the number of products, the product
demand correlations and the spread in product contri-
bution margins In terms of supply chain performance,
Bichescu and Fry [25] rely on a numerical-analysis ap-
proach to examine how decision-making rights split be-
tween supply chain agents determine order quantity and
shipping frequency and then affect the supply chain per-
formance. Among the results presented in the study,
they have found that concentrating channel power with
the supplier can lead to supply chain profits that are
very close to a centralized scenario, but also results in
lower customer service levels.
In summary, from both the supplier’s and buyer’s
points of view, single and dual sourcing strategies pro-
vide advantages as well as disadvantages. The choice
between single sourcing and dual sourcing depends on
 Author's personal copy
792 H. Yu et al. / Omega 37 (2009) 788 – 800
the trade-offs of the two sourcing methods and nu-
merous environmental factors As such, a comparison
between these two sourcing models under various sit-
uations will continue attracting research interests and
efforts.
In this paper, we aim to identify the factors affecting
the decision on single or dual sourcing in a two-stage
supply chain with a non-stationary and price-sensitive
demand. In what follows, we will develop a set of EPFs
when a single supplier and two suppliers are used, re-
spectively in the presence of supply chain disruption
risks. By comparing and evaluating the EPFs, we iden-
tify the conditions under which one sourcing method
may outperform the other and obtain the critical factors
affecting the choice between the two sourcing alterna-
tives.
3. Models and analyses
Consider a sourcing decision problem faced by a
manufacturer (a buyer) and suppose that there are two
suppliers that have the capacity and technology to
provide a needed critical part for the manufacturer’s
final product. Supplier 1 is located outside the manufac-
turer’s country, and offers competitive price; however,
this supplier is prone to breakdowns or the supplied
material can experience substantial loss during transit
due to long lead time and distance (e.g., [26]). Supplier
2 is local, relatively reliable but more expensive. The
buying firm thus has two sourcing alternatives to select.
One is single sourcing, where supplier 1 is used as the
main and only source for the critical part. The other is
dual sourcing, where supplier 1 is the main supplier and
supplier 2 is secondary who is given a portion of the
demand, x, to produce during each supply cycle. Given
that a supplier breakdown is possible, how should the
manufacturer choose the sourcing method? This prob-
lem under study is typical in today’s supply chains
(e.g., [27]). As indicated by Sheffi [27], the impacts of
the new era will challenge supply chain managers to
adjust relations with suppliers and customers, and it is
important to look at how companies should organize to
meet those challenges efficiently.
3.1. Model assumptions
In this paper, we study a two-stage supply chain con-
sisting of one buyer and two suppliers, where the whole-
sale prices offered by the two suppliers are different, as
they may be located geographically far from each other
and differ in terms of price and reliability. This is a typ-
ical and commonly observed phenomenon in today’s
increased global sourcing environment where the for-
eign suppliers often offer competitive prices but lower
reliability or higher risk, while the local or domestic
vendors are more reliable but more expensive. In ad-
dition, considering that a supply disruption may result
in the change of market scale and demand, for exam-
ple, during the SARS period, the demand for respira-
tors has experienced a sudden surge, we assume that
the demand faced by the buyer is non-stationary and
that each supplier’s wholesale price changes with the
buyer’s purchase quantity.
We consider the situation where the main supplier is
completely down and unable to satisfy any of the buy-
ing firm’s demand when struck by unexpected events.
However, in a normal state, we assume that there is no
limitation on each supplier’s capacity. A limited supply
capacity is more realistic and will certainly complicate
the decision analysis. We plan to investigate the impact
of the supply capacity limitation on sourcing strategy
design in a separate paper. Additionally, we use a single
parameter, p, to capture the probability that the main
supplier may break down during each supply cycle.
The proposed model in this paper differs from the
existing ones in the following three perspectives. First,
we consider a situation where the demand is not only
correlated with the wholesale price, which is captured
by a sensitivity coefficient, but also dependent upon the
maximum market scale. Secondly, with the observation
that the demand for a particular item can be influenced
significantly by the occurrence of an unforeseen event
like in the SARS case, we treat the maximum market
scale in such a way that a supply disruption will create
a positive increase shift for the market scale. Finally,
our model helps identify the breakeven points and pro-
vides closed-form expressions that would clearly guide
a company to design a more profitable sourcing strategy.
3.2. The state variables
For the illustrative purpose, we use an exponential
function as follows to describe the buying firm’s demand
during a supply cycle:
D = M exp(−kC), (3.1)
In (3.1), the element, M, is the maximum market scale,
C is the unit wholesale price of a supplier, k (k > 0) is
a coefficient of price sensitivity, and D is the realized
demand for a given unit wholesale price of C. In a state
with supply chain disruptions, we assume that when
supply disruption occurs, the maximum market scale
faced by the buyer shifts from M to M + M.
 Author's personal copy

H. Yu et al. / Omega 37 (2009) 788 – 800 793

Table 1
The scenarios, state variables, and market demands .

Normal state (1 − p) Disrupted state ( p)

Prices Market demand Prices/loss Market demand

Single sourcing C s
m s M exp(−kC sm ) C
u (M +  M) exp(−kC sm )
Cm Cu
Dual sourcing d M exp[− 21 k(Cm
s + C d )]
m (M +  M) exp[− 21 k(Cbd + Cbn )]
Cm Cbd

Notation: p , the probability of disruptions faced by the main supplier during a supply cycle; S , the buyer’s unit sales price for the final product,
s , the main supplier’s unit wholesale price in single sourcing; C d , the main supplier’s unit wholesale price in dual sourcing; since
in $/unit; Cm m
the sales quantity of the main supplier in dual sourcing is smaller than that in single sourcing, it is reasonable to assume that Cm d > Cs ; C ,
m bn
d
the secondary supplier’s unit wholesale price in normal state. Let Cbn > Cm ; Cbd , the secondary supplier’s unit wholesale price in disrupted
state. Let Cbd > Cbn ; Cu , the buyer’s unit loss of the unsatisfied demand

The buying firm has two sourcing methods to choose which can be simplified as
from, one is single sourcing and the other is to use both
suppliers. The decision needs to be made by taking the dn = M exp{−0.5k(Cm d
+ Cbn )}
d d
supply disruption risks into consideration. To proceed, × {(S − Cm ) − x(Cbn − Cm )}. (3.3)
we present the decision scenarios, state variables, and
Note that in (3.3), x denotes the portion of the demand
the symbols in Table 1 ( a complete list of notation and
allocated to the secondary supplier during each supply
symbols used throughout the paper is provided in an
cycle.
Appendix). Note that there are two sets of subscripts
Similarly, we derive the buying firm’s profit functions
used to differentiate the parameters: the first set, (m, b),
during a cycle with supply chain disruptions. In the case
denotes the main supplier and the secondary supplier,
of single sourcing, when the main supplier is struck
respectively; and the other set, (n, d), refers to the two
down, the supply disruption occurs and the buying firm
states of nature: a normal state and a disrupted state,
experiences a complete loss. Consequently, the profit
respectively. Moreover, a set of superscripts, (s, d), is
function is
used to indicate the single and dual sourcing, respec-
tively. Note also that whenever two suppliers are used, sd = (−(M + M) exp(−kC sm ))Cu . (3.4)
we use the average of their wholesale prices to calculate
the market demand. When the dual sourcing method is used, that is, the
secondary supplier gets a portion, x, of the average de-
mand each cycle, the wholesale price charged by the
3.3. The EPFs secondary supplier changes with the occurrence of dis-
ruptions: for the first x portion of the demand, the price
During a normal state, if the buying firm uses the is previously acknowledged, which is Cbn ; for the rest
main supplier as the single source, then the realized (1 − x) portion, the price is increased to Cbd . There-
demand is calculated as D = M exp(−kC sm ), whereas fore, the buyer’s profit in the disrupted state with dual
the demand becomes D= M exp[− 21 k(Cm d +C )] when
bn sourcing, dd , becomes
the buying firm uses both suppliers. Thus, in a normal
state, the profit function of the buying firm with single dd = (1 − x)(S − Cbd )M exp{−0.5k(Cm
d
+ Cbn )}
sourcing, sn , is given by d
+ x(S − Cbn )M exp[−0.5k(Cm + Cbn )]
− M exp[−0.5k(Cbd + Cbn )]Cu ,
sn = (S − Cm
s
)M exp(−kC sm ). (3.2)
which can be simplified as
And the profit of the buyer when dual sourcing is used, dd = M exp[−0.5k(Cm d
+ Cbn )][(S − Cbd )
dn , can be calculated as + x(Cbd − Cbn )]
− {M exp[−0.5k(Cbd + Cbn )]}Cu . (3.5)
dn = (1 − x)(S − Cm
d d
)M exp{−0.5k(Cm + Cbn )}
Thus, the EPF of the buying firm with single sourcing
d
+ x(S − Cbn )M exp{−0.5k(Cm + Cbn )}, (i.e., only the main supplier is used) in the presence of
 Author's personal copy

794 H. Yu et al. / Omega 37 (2009) 788 – 800

supply chain disruptions, sm , is the weighted sum of will use the price, Cm s , as a base, and write all other

(3.2) and (3.4) by the disruption probability, p; that is, price values in the following format:
d s s
sm = (1 − p)sn + p · sd . (3.6) Cm = (1 + 1 )Cm ; Cbn = (1 + 2 )Cm ;
s
Cbd = (1 + 3 )Cm , (3.10)
And the associated expected profit function in the case
of dual sourcing for the buying firm is where 0 < 1 < 2 < 3 . Additionally, there are two
other price/cost elements, (S, Cu ), which are written in
s as a base as well:
d = (1 − p)dn + p · dd , (3.7) a similar fashion by using Cm
s s
where dn is given in (3.3) and dd is given in (3.5). An S = (1 + 1 )Cm , Cu = (1 + 2 )Cm . (3.11)
extreme case of (3.7) is given by x = 1, that is, the Notice that 1 > 3 > 2 > 1 , but 2 can be either pos-
secondary supplier (the local one) is used as the main itive or negative and there is no need to define its re-
and only source, which yields the following expression lationship with other parameters. Additionally, let the
of the EPF of the buying firm: market scale shift during supply disruptions, M, be
sb = (1 − p)[(S − Cbn )M exp(−kC bn )] given as M = M. Moreover, to simplify the expres-
sions to be obtained during subsequent analyses, we
+ p{[(S − Cbn )M exp(−kC bn )] − [(M + M) define the following terms:
× exp(−kC bd ) − M exp(−kC bn )]Cu }
0 = (MC sm ) exp(−kC sm ); (3.12)
= [(S − Cbn )M exp(−kC bn )] − p[(M + M)
e32 = exp(−(3 − 2 )kC sm ); (3.13)
× exp(−kC bd ) − M exp(−kC bn )]Cu (3.8)
e2 = exp(−2 kC sm ); (3.14)
This special case is also considered as a possible ver-
s
sion of the single sourcing method; hence, the function e12 = exp(−0.5k(1 + 2 )Cm ); (3.15)
given in (3.8) can be used as one of the benchmarks for s
subsequent analyses. Furthermore, it is straightforward e23 = exp(−0.5k(2 + 3 )Cm ). (3.16)
to see that the condition for the dual sourcing method to As a result, (3.6), (3.7) and (3.8) can be written in the
outperform the single sourcing counterpart, i.e., (3.7) is following expressions, respectively:
greater than (3.6), requires the following relationship,
after some algebra, to be true: sm = 0 [(1 − p)1 − p(1 + )(1 + 2 )]; (3.17)

⎨ e12 (1 − p)[(1 − 1 ) − x(2 − 1 )] ⎬

 −1
d − sd
p > 1 + ds . (3.9)
n − dn
In what follows, we will analytically examine the rela-
tionships among (3.6)–(3.8).
3.4. The analysis of the EPFs
The preceding section has obtained the EPFs for three
sourcing decision alternatives, namely single sourcing
with the foreign supplier, single sourcing with the local
supplier, and dual sourcing with both suppliers. Fac-
ing three alternatives, the buying firm will be interested
in knowing the conditions under which one alternative
would outperform the others. Thus, we compare the
EPFs to identify these conditions. To proceed, we write
a number of relevant terms in new expressions. First,
considering that a number of price values are used as
data inputs to the decision-making process and possess
the following relationship: Cbd > Cbn > Cm d > C s , we
m (3.20)
⎧ ⎫
d =  0 + p[e12 ((1 − 3 ) + x(3 − 2 ))
⎩ ⎭
−e23 (1 + 2 )]
and (3.18)
sb = 0 · e2 [(1 − 2 ) − p(e32 (1 + ) − 1)(1 + 2 )].
(3.19)
Thus, we can analytically compare the three sourc-
ing alternatives using (3.17)–(3.19) to obtain criti-
cal ranges of the disruption probability for optimal
selection.
First, we compare the single sourcing method with
the foreign supplier as the main source and the dual
sourcing method, i.e., the functions in (3.17) and (3.18)
are compared. We first write the right-hand-side of (3.9)
in the following terms by using (3.10)–(3.12), (3.15)
and (3.16):
dd − sd e12 [(1 − 3 ) + x(3 − 2 )] + (1 + 2 )(1 +  − e23 )
= .
sn − dn 1 − e12 [(1 − 1 ) + x(2 − 1 )]
 Author's personal copy
H. Yu et al. / Omega 37 (2009) 788 – 800
Then using the relationship in (3.9) to solve for p, we
arrive at the following result:
Proposition 1. When the probability of disruptions ( p)
satisfies the relationship
1 − e12 [(1 − 1 ) − x(2 − 1 )]
p > p1c = , compare the two sourcing alternatives: dual sourc-
1 − e12 (1 − x)(3 − 1 ) + (1 + 2 )(1 +  − e23 )
(3.21)
then d > sm , that is, the dual sourcing method out-
performs single sourcing with the foreign supplier as
the main source.
Proof. Substituting (3.20) into (3.9) and solving for p
yields the critical value given in (3.21). 
Next, we compare dual sourcing and single sourcing
with the local supplier (the secondary one), and arrive
at the following result for decision-making:
Proposition 2. When the probability of disruptions ( p)
meets the requirement
e12 [(1 − 1 ) − x(2 − 1 )] − e2 (1 − 2 )
p  p2c =
e12 (1 − x)(3 − 1 ) − e23 (1 + 2 ) − e2 (e32 (1 + ) − 1)(1 + 2 )
then sb > d , that is, using the local supplier (the sec-
ondary one) as the only source outperforms using both
suppliers.
Proof. Set the difference between (3.19) and (3.18) to
greater than or equal to zero, i.e., sb −d 0, and after
some algebra, the result in (3.22) follows. 
Combining the results in the above two propositions,
we see that two critical values of the disruption proba-
bility, ( p1c , p2c ), determine the situations under which the
following relationship is held: sb > d > sm ; hence,
it is not difficult to see that p1c < p2c . The thresholds of
the disruption probabilities will be used to determine
the most profitable sourcing method. We summarize the
guidelines for decisions in the following proposition.
Proposition 3. In choosing the most profitable sourcing
method, there are two critical values of the disruption
probabilities, ( p1c , p2c ), given by (3.21) and (3.22), re-
spectively, guiding the decisions in the following way:
If p < p1c , then choose the foreign supplier
⎧
⎪
⎪ (the main one) as the single source.
⎪
⎨ 4.2. The sensitivity of EPF to (p, x)
If p1c < p < p2c , then choose dual sourcing.
c To see the effects of x, we take a derivative of
⎩ If p > p2 , then choose the local supplier
⎪
⎪
⎪
(the secondary one) as the single source.
795
Proof. Using the results in Propositions 1 and 2 yields
the conclusions directly. 
4. A numerical analysis and implications
In this section, we rely on a numerical analysis to
ing and single sourcing with the foreign supplier (the
main one); specifically, we would like to provide a nu-
merical illustration as well as to obtain more insights
into the properties of the profit functions in (3.6) and
(3.7) and identify the factors affecting the choice be-
tween the two sourcing alternatives. In the subsequent
analyses, the base values summarized in Table 2 are
used.
4.1. An illustration of the choice of the most profitable
sourcing method
It is seen from (3.6) and (3.7) that the buying
firm’s expected profits are linearly correlated with the
, (3.22)
probability that the main supplier may experience a dis-
ruption during a supply cycle. The functions in (3.6)
and (3.7) are plotted against the disruption probability
ranging from 0 to 1 with the portion of demand being
fixed at x = 0.3 for the dual sourcing option. Addition-
ally, the profit function at the extreme situation where
the buyer gives 100% of demand to the secondary sup-
plier (x = 1) in (3.8) is also shown for the comparison
purpose. All these plots are displayed in Fig. 1.
Fig. 1 clearly shows the two critical values of the
disruption probability for decision-making, as stated
in Proposition 3; in particular, using (3.21) and (3.22)
yields that p1c = 0.08 and p2c = 0.25, respectively. With
the associated results in Proposition 3 or relying on Fig.
1, we see that if p < p1c =0.08, single sourcing from the
primary supplier (the international one) will be better;
if the disruption probability falls into the range (0.08,
0.25), then using dual sources outperforms the counter-
part; on the other hand, if the disruption probability is
larger than the higher critical value, i.e., p > p2c = 0.25,
then the buying firm should choose the secondary sup-
plier (the local one) as the main and only source.
(3.7) with respect to the variable, x, which yields the
 Author's personal copy

796 H. Yu et al. / Omega 37 (2009) 788 – 800

Table 2
Parameter base values .

Parameters s
Cm d 
Cm Cbn 2 Cbd 3 Cu 2 S 1 M M  x k
1

Values 330 340 360 400 200 420 10000 2000 0.3 0.002
Ratios 0.03 0.09 0.21 −0.39 0.27 0.2

x 105 Next we use numerical examples to show the sen-

6 sitivity of the buyer’s expected profit in the case of
dual sourcing to the two key input parameters, ( p, x).
5 Single sourcing (main supplier)
Holding all parameters fixed at the base values given in
Table 1 except for the portion of demand allotted to the
4 secondary supplier, x, we have calculated the buyer’s
Dual sourcing
expected profits against ( p, x), where p takes on six val-
Profit (Π)

Only local supplier (reference line)

3 ues ranging from 0.10 to 0.4 with an increment of 0.05
and a special case of p = 31 , and x takes values from
2
0.1 to 0.7 with an increment of 0.1. A total of 36 data
points are obtained and plotted in Fig. 2.
If we calculate the critical value of the disruption
1
probability given in (4.2), we see that p=(2 −1 )/(3 −
pc1 p2
c
1 ) = (0.09 − 0.03)/(0.21 − 0.03) = 31 . Referring to
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Fig. 2, we see that at p = 13 , the buyer’s profit remains
Disruption probability (p) unchanged even if the value of x varies. Furthermore,
when p < 13 , the buyer’s expected profit is negatively
Fig. 1. The expected profits under various sourcing alternatives. correlated with the demand portion given to the sec-
ondary supplier (x). On the other hand, when p > 31 , the
following partial differential equation:
buyer’s expected profit increases with x for each given
jd d d value of p. This suggests that when p exceeds the thresh-
= (Cm − Cbn )M exp[−0.5k(Cm + Cbn )]
jx old, the buyer would want to increase the portion of
d
+ p(Cbd − Cm d
)M exp[−0.5k(Cm + Cbn )]. demand allocated to the secondary supplier to increase
its expected profit.
(4.1)
Considering that Cmd < C < C , we see easily that
bn bd 4.3. The sensitivity of EPF to ( p, M)
when p > (Cbn − Cm d )/(C − C d ), then jd /jx > 0.
bd m
The probability range can be also written in the With the consideration that the market scale, M, can
price ratios given in (3.9). In summary, when the also affect the buyer’s EPF in (3.7), we have conducted
disruption probability, p, satisfies the following a set of four sensitivity analyses of the EPF with respect
relationship: to ( p, M), where p ranges from 0 to 1 and M changes
d from 10,000 to 20,000 (i.e.,  ranges from 0 to 1), and
Cbn − Cm  2 − 1
p> or p > , (4.2) the secondary supplier’s portion of demand, x, is chosen
d
Cbd − Cm 3 − 1 to be x = 0, 0.3, 0.6, 1 for each set. The results are
then the buying firm’s expected profit in the case of dual illustrated in Figs. 3a–d.
sourcing (given in (3.7)) increases as the portion of de- A careful examination of Fig. 3 yields two interesting
mand (x) allocated to the secondary supplier increases. indications. First, when x = 0 and p increases (shown in
Practically, this suggests that allocating a higher por- Fig. 3a), the only way to increase the EPF is to expand
tion of demand, x, to the secondary supplier will in- the market scale, M; however, once M reaches a certain
crease the profit of the buying firm. On the other hand, point, the EPF will start to decrease. Secondly, when
if p < (Cbn −Cm d )/(C −C d ), then jd /jx < 0, mean- x > 0 and p increases (shown in Figs. 3b–d), an increase
bd m
ing that the EPF is monotonically decreasing with re- of the market scale, M, will be required to increase
spect to x, and that increasing the secondary supplier’s EPF as well. Furthermore, as x increases, the market
allocation portion of the demand will lower the profit scale, M, must increase at a faster pace in order to en-
of the buying firm. sure an increase of EPF. This set of sensitivity analysis
 Author's personal copy

H. Yu et al. / Omega 37 (2009) 788 – 800 797

360

340
Π: The Expected Profit (in $1,000)
320

p = 0.1
300
p = 0.15

280

p = 0.3
260
p = 1/3
p = 0.35
240 p = 0.40

220

200
0.1 0.2 0.3 0.4 0.5 0.6
x: Portion of the demand allocated to the secondary supplier

Fig. 2. The Buyer’s expected profits with respect to ( p, x).

Fig. 3. The images of EPF with respect to ( p, M) in the four cases.

suggests that the buying firm would want to allocate a EPF, we have calculated the values of EPF against
higher portion of demand to the secondary supplier in (x, M) when p = 0.1, 13 , and 0.4, respectively, x
order to obtain higher EPF when the maximum market ranges from 0 to 1, and M changes from 0.1 to 0.4
scale and the disruption probability increase. (or  = 0.1, 0.2, 0.3, 0.4). The associated results are
reported in Table 3a–c.
4.4. The sensitivity of EPF to (x, M) Two observations can be made from Table 3. First,
To identify the possible impact of the market scale regardless of the magnitude of the disruption prob-
shift in the presence of disruptions on the buyer’s ability ( p), the buyer’s expected profit decreases as
 Author's personal copy

798 H. Yu et al. / Omega 37 (2009) 788 – 800

Table 3 that either single or dual sourcing can be effective de-

A sensitivity analysis of the EPF (in 1000) to (x, ) . pending on the magnitude of the disruption probabil-
 x ity. Additionally, we have conducted a set of sensitivity
analyses of the buyer’s expected profit function in the
0 0.1 0.2 0.3 0.4 1.0
case of dual sourcing, the results of which indicate how
(a) p = 0.1 the buyer’s profit would change with respect to the four
0.1 363.90 356.95 350.00 343.05 336.09 294.38 parameters, ( p, x, M, M) in each supply cycle.
0.2 354.55 347.60 340.65 333.69 326.74 285.03 This research provides closed-form solutions and crit-
0.3 345.20 338.24 331.29 324.34 317.39 275.67
0.4 335.84 328.89 321.94 314.99 308.03 266.32
ical values for helping managers and decision-makers
choose the most profitable sourcing strategies in the
(b) p = 31 presence of supply chain disruption risks. In particu-
0.1 286.05 286.05 286.05 286.05 286.05 286.05
0.2 254.87 254.87 254.87 254.87 254.87 254.87
lar, the results apply to the situations where the buyer’s
0.3 223.70 223.70 223.70 223.70 223.70 223.70 demand is sensitive to the supplier’s wholesale price,
0.4 192.52 192.52 192.52 192.52 192.52 192.52 the market shift during a supply disruption can be es-
(c) p = 0.4
timated, two suppliers are available but differ in price
0.1 −114.31 265.80 267.78 269.77 271.75 283.67 and reliability, and each supplier has no capacity limi-
0.2 −150.26 228.38 230.37 232.36 234.34 246.26 tation. The situations where a global supplier and a do-
0.3 −186.20 190.97 192.96 194.94 196.93 208.85 mestic vendor are used for purchasing commodities or
0.4 −222.15 153.56 155.54 157.53 159.51 171.43 non-strategic items are possible areas of applications of
the results presented in this paper.
It is necessary to point out a number of limitations
the market scale shift increases for a given value of of this research. First of all, the demand model is fairly
x. This is easy to understand because when the sec- simple and considers only price-sensitive demand. To
ondary supplier is given a fixed portion of demand consider more realistic demand patterns, we can add a
each cycle and the market scale expands, the buyer’s few time-related factors such as the product life cycle
loss or portion of unsatisfied demand grows. Second, time, the disruption duration, and supply cycle time to
when p is less than the critical value (i.e., 13 in this the model. Secondly, the supplier’s capacity is currently
case) as calculated in Section 4.1, for a fixed , the assumed to be infinite and that the main supplier’s ca-
buyer’s expected profit decreases as the value of x in- pacity is completely lost during disruptions, whereas in
creases, implying that the buyer should give less or no reality, supplier’s capacity is limited and may be par-
demand to the secondary supplier. In contrast, when tially lost in the occurrence of disasters. Thirdly, this
the disruption probability is larger than 31 , the buyer study focuses on only dual sourcing and single sourc-
would want to allocate more demand to the secondary ing. As an increased supply disruption risk calls for
source. more qualified suppliers, it becomes imperative to ex-
amine the sourcing decision in a multi-source situation.
5. Discussions and conclusions Finally, current analysis looks at only the buyer’s profit
but ignores the suppliers’ benefit and welfare; thus, it
This research examines the complexity of the sourc- would be more interesting to examine the sourcing deci-
ing decision in the presence of supply chain disrup- sions from both parties’ points of view. These represent
tions; in particular, the famous debate between single our future research directions.
sourcing and dual sourcing is revisited by taking supply
disruption risks into account. We study the sourcing
decision alternatives in the context that the demand is Acknowledgement
price-sensitive and the market scale increases when a
supply disruption occurs. The disruption risk is cap- This research is supported by the National Natu-
tured by a probability, the non-stationary demand is ral Science Foundation of China (706711021) for the
modeled with an exponential function of the wholesale project titled “Project Operation and Simulation of
price multiplied by the maximum market scale, and the Emergency Response Logistics Network in the Sys-
decision is analyzed based on EPFs. We have identi- tem of Anti-Bioterrorism”. We are also grateful to the
fied two critical values of the disruption probabilities, anonymous reviewers for their constructive comments,
which provide a guideline for choosing the most prof- which have improved the quality and presentation of
itable sourcing method for the buying firm. It is found this paper.
 Author's personal copy

H. Yu et al. / Omega 37 (2009) 788 – 800 799

Appendix A. List of notation and symbols used throughout the paper

D the realized demand for a given unit wholesale price

M the maximum market scale
M the amount of market scale shift when supply disruption occurs
 a proportion of M to represent the market scale shift, M = M
x the portion of the demand allocated to the secondary supplier each cycle, 0  x  1
k a coefficient of price sensitivity, k > 0
p the probability of disruptions faced by the main supplier during a supply cycle
p1c the first critical probability that breaks even the profit with the main supplier as the single
source and the profit with dual sourcing
p2c the second critical probability that breaks even the profit with the secondary supplier as
the single source and the profit with dual sourcing
S the buyer’s unit sales price for the final product, in $/unit
Cm s the main supplier’s unit wholesale price in single sourcing
Cm d the main supplier’s unit wholesale price in dual sourcing, Cm d > Cs
m
Cbn the secondary supplier’s unit wholesale price in normal state, Cbn > Cm d

Cbd the secondary supplier’s unit wholesale price in disrupted state, Cbd > Cbn
Cu the buyer’s unit loss of the unsatisfied demand
sn the buyer’s profit with single sourcing in a normal state
dn the buyer’s profit with dual sourcing in a normal state
sd the buyer’s profit with single sourcing in a disrupted state
dd the buyer’s profit with dual sourcing in a disruptedstate
sm the buyer’s expected profit with the main supplier as the single source
sb the buyer’s expected profit with the secondary supplier as the single source
d the buyer’s expected profit with dual sourcing
EPF expected profit function
(1 , 2 , 3 , 1 , 2 ) scale parameters used to write the price values of (Cm d , C , C , S, C ) in relation to C s
bn bd u m
(0 , e32 , e2 , e12 , e23 ) expressions in (3.12)–(3.16) used to simplify the functions of (sm , sb , d )

References [8] Cranfield Management School. Supply chain vulnerability. New

[1] Reitman V. To the rescue: Toyota’s fast rebound after fire at
supplier shows why it is tough. The Wall Street Journal; Eastern
Edition, 8 May 1997.
[2] Sheffi Y. The resilient enterprise. Cambridge, MA: MIT Press;
2005.
[3] Latour A. Trial by fire: a blaze in albuquerque sets off major
crisis for cell-phone giants, The Wall Street Journal; Eastern
Edition, January 29, 2001.
[4] Pearson H. US lacks back-up for flu vaccine shortfall. Nature:
International Weekly Journal of Science 2004;431(7010):726.
[5] Wong CY, McFarlane D, Zaharudin A, Agarwal V. The
intelligent product driven supply chain. In: Proceedings of IEEE
systems man and cybernetics, man and cybernetics, Tunisia,
2002.
[6] Christopher M, Towill D. Supply chain migration from lean and
functional to agile and customized. Supply Chain Management:
An International Journal 2000;5(4):206–13.
[7] Tang CS. Perspectives in supply chain risk management.
International Journal of Production Economics 2006;103(2):
451–88.
York (NY): Cranfield University Press; 2002.
[9] Svensson G. A conceptual framework for the analysis of
vulnerability in supply chains. International Journal of Physical
Distribution & Logistics Management 2000;30(9):731–49.
[10] Hallikas KP, Virolainen T. Risk management processes in the
network environment. In: Proceedings of 12th international
working seminar on production economics, 2002.
[11] Chopra S, Sodhi M. Managing risk to avoid supply-chain
breakdown. MIT Sloan Management Review 2004;46(1):
53–61.
[12] Chopra S, Reinhardt G, Mohan U. The importance of
decoupling recurrent and disruption risks in a supply chain,
2005. Working paper, Northwestern University.
[13] Vlachos D, Tagaras G. An inventory system with two
supply modes and capacity constraints. International Journal of
Production Economics 2001;72(1):41–58.
[14] Kleindorfer PR, Saad GH. Managing disruption risks in supply
chains. Production and Operations Management 2005;14(1):
53–68.
[15] Xiao T, Qi X. Price competition, cost and demand disruptions
and coordination of a supply chain with one manufacturer and
 Author's personal copy
800 H. Yu et al. / Omega 37 (2009) 788 – 800
two competing retailers. Omega: An International Journal of
Management Science 2008;36(5):741–53.
[16] Larson PD, Kulchisky JD. Single sourcing and supplier
certification: performance and relationship implications.
Industrial Marketing Management 1998;27(1):73–81.
[17] Berger PD, Gerstenfeld A, Zeng AZ. How many suppliers are
best? A decision-analysis approach. Omega: An International
Journal of Management Science 2004;32(1):9–15.
[18] Berger PD, Zeng AZ. Single versus multiple sourcing in the
presence of risks. Journal of the Operational Research Society
2006;57(3):250–61.
[19] Pochard S. Managing risks of supply-chain disruptions: dual
sourcing as a real option, Massachusetts Institute of Technology,
2003.
[20] Burke GJ, Carrillo JE, Vakharia AJ. Single versus multiple
supplier sourcing strategies. European Journal of Operational
Research 2007;182(1):95–112.
[21] Wagner SM, Gunther F. Supplier switching decisions. European
Journal of Operational Research 2007;183(2):700–17.
[22] Lau HS, Zhao L-G. Dual sourcing cost-optimization with
unrestricted lead-time distributions and order-split proportions.
IIE Transactions 1994;26(5):66–75.
[23] Kelle P, Miller PA. Stockout risk and order splitting. Interna-
tional Journal of Production Economics 2001;71(1–3):407–15.
[24] Tomlin B, Wang Y. On the value of mix flexibility and dual
sourcing in unreliable newsvendor networks. Manufacturing &
Service Operations Management 2005;7(1):37–57.
[25] Bichescu BC, Fry MJ. A numerical analysis of supply
chain performance under split decision rights. Omega, 2007,
doi:10.1016/j.omega.2007.04.001.
[26] Sounderpandian J, Prasad S, Madan M. Supplies from
developing countries: optimal order quantities under loss risks.
Omega: An International Journal of Management Science
2008;36(1):122–30.
[27] Sheffi Y. Supply chain management under the threat of
international terrorism. The International Journal of Logistics
Management 2001;12(2):1–11.
[28] Jüttner U, Peck H, Christopher M. Supply chain risk
management: outlining an agenda for future research. In:
Griffiths, Hewitt & Ireland, conference proceedings, 2002.
p. 443–50.
[29] Ruiz-Torres AJ, Farzad M. The optimal number of suppliers
considering the costs of individual supplier failures. Omega:
An International Journal of Management Science 2007;35(1):
104–15.
[30] Fukuda Y. Optimal policies for the inventory problem
with negotiable lead-time. Management Science 1964;10(4):
690–708.
[31] Newman RG. Single Sourcing: Short-term savings versus
long-term problems. Journal of Purchasing and Materials
Management 1989; 20–5.