You are on page 1of 30

International Journal of Production Research

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/tprs20

The impact of flexibility and redundancy on


improving supply chain resilience to disruptions

Masoud Kamalahmadi, Mansoor Shekarian & Mahour Mellat Parast

To cite this article: Masoud Kamalahmadi, Mansoor Shekarian & Mahour Mellat Parast (2021):
The impact of flexibility and redundancy on improving supply chain resilience to disruptions,
International Journal of Production Research, DOI: 10.1080/00207543.2021.1883759

To link to this article: https://doi.org/10.1080/00207543.2021.1883759

Published online: 20 Feb 2021.

Submit your article to this journal

Article views: 325

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=tprs20
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH
https://doi.org/10.1080/00207543.2021.1883759

The impact of flexibility and redundancy on improving supply chain resilience to


disruptions
Masoud Kamalahmadia , Mansoor Shekarianb and Mahour Mellat Parastc
a Miami Herbert Business School, University of Miami; b Lundquist College of Business, University of Oregon, Eugene; c Ira A. Fulton Schools of
Engineering, Arizona State University, Tempe

ABSTRACT ARTICLE HISTORY


This paper examines the relative impact on supply chain responsiveness of adding flexibility and Received 18 October 2019
redundancy. We seek to investigate the effectiveness of flexibility and redundancy in terms of min- Accepted 10 January 2021
imising expected supply chain cost and maximizing expected service delivery when a supply chain KEYWORDS
is exposed to supplier and environmental disruptions. Compared to a supply chain that uses neither Supply chain management;
strategy and is exposed to supplier and environmental disruptions, both practices reduce expected flexibility; supply network;
lost sales, reduce expected total cost, and increase the expected service level. Our results show that enterprise risk management;
the backup-suppliers practice is more effective than the flexible-suppliers practice, as measured operations management
by cost reduction and service-level improvement. In addition, we show that a hybrid practice cre-
ated by adding some level of flexibility to a backup-suppliers practice is more desirable in terms
of lowering pre-disruption investment, increasing the responsiveness of the supply chain, and low-
ering the expected total cost. Moreover, the result of the simulation shows that on average, the
proposed stochastic model could decrease the company’s total cost by 5.51%. It also shows that
planning and investing in proactive demand management could decrease the company’s total cost
by 7.87%. Our findings provide insights into the contingencies that support implementation of each
practice.

1. Introduction are challenged (Craighead et al. 2007; Parast and Shekar-


The past few decades have been notable for major ian 2019; Ruiz-Benítez, López, and Real 2018; Tom-
changes in supply chains due to an increasing level of lin 2006). Due to concerns about unexpected catas-
globalisation and a higher rate of innovation (Hosseini, trophic events, researchers and practitioners increased
Ivanov, and Dolgui 2019a; Kleindorfer and Saad 2005; their focus on supply chain risk management (SCRM)
Tang 2006a). Based on Xu et al. (2020), 1069 events of (Ghadge, Dani, and Kalawsky 2012; Sodhi, Son, and Tang
supply chain disruption happened in the first half of 2012; Tang and Musa 2011; Zhu, Krikke, and Caniëls
2018, which was the highest rate in three years. Catas- 2017). SCRM is concerned with identifying sources of
trophic events such as the recent COVID-19 pandemic risks and vulnerabilities in a supply chain, then devel-
in 2020, hurricane Harvey in Houston in 2017, hur- oping strategies to mitigate disruptions by making the
ricane Maria in Puerto Rico in 2017, the 2010 erup- supply chain more resilient and responsive (Brusset and
tion of a volcano in Iceland, and the 2011 Japanese Teller 2017; Ivanov 2020a; Jüttner, Peck, and Christopher
tsunami caused serious damage and financial losses to 2003).
several industries such as the auto and computer indus- There has been increased awareness regarding the
tries (Chen, Xu, and Zhou 2020; Dolgui, Ivanov, and importance of supply chain responsiveness and miti-
Sokolov 2018; Dolgui, Ivanov, and Rozhkov 2020; Duong gating supply chain disruptions (Ciccullo et al. 2018;
and Chong 2020; Esmaeili-Najafabadi et al. 2019; Fiksel Dubey, Gunasekaran, and Childe 2018; Elzarka, 2013;
et al. 2015; Hosseini et al. 2019b; Inman and Blumenfeld Parast, Sabahi, and Kamalahmadi 2019; Paul et al. 2019;
2014; Ivanov et al. 2016; Olivares Aguila and ElMaraghy Rezapour et al. 2018). After being disrupted, a resilient
2018; Pavlov et al. 2019b; Shekarian, Nooraie, and Parast supply chain can quickly return to its original state or
2020). Consequently, for firms with global supply chains, even move to a new, more desirable state (Behzadi,
the effectiveness and responsiveness of traditional cost- O’Sullivan, and Olsen 2020; Fattahi, Govindan, and Mai-
driven supply chains designed for stable environments hami 2020). A study by Aon Risk Solutions found that

CONTACT Mahour Mellat Parast mahour.parast@asu.edu Ira A. Fulton Schools of Engineering, Arizona State University, Tempe, AZ 85281

© 2021 Informa UK Limited, trading as Taylor & Francis Group


2 M. KAMALAHMADI ET AL.

the percentage of global companies reporting a loss of creating redundancies across their supply chains. Hav-
income because of supply chain disruptions increased ing multiple suppliers, safety stock, overcapacity, and
from 28% in 2011–42% in 2013 (Sáenz and Revilla 2014). backup suppliers are instances of redundancies in a sup-
Based on Gunessee, Subramanian, and Ning (2018), ply chain (Kamalahmadi and Parast 2017; Tang 2006b;
enterprises that have little flexibility and agility expose Tomlin 2006). Alternatively, firms can instill flexibility
themselves to the disastrous consequences of changes in their supply base. Flexibility is defined as the ability
due to facing catastrophic vulnerability. In addition, to take different positions to better respond to abnor-
Shekarian, Nooraie, and Parast (2020) showed that in mal situations and rapidly adapt to significant changes
a rapidly changing environment, firms relying on their in the supply chain (Lee 2004; Sreedevi and Saranga
supply chain to remain competitive must be more agile in 2017). Having flexible transportation systems or adding
perceiving and developing opportunities, more respon- volume flexibility based on suppliers’ production capac-
sive to disruptions, and more resilient against external ities are examples of the ways flexibility can be added
threats. A recent study by Shih (2020) indicates that to a firm’s supply chain (Tang and Tomlin 2008; Tomlin
the COVID-19 contagion has had a major impact on 2006).1
Chinese manufacturers, and because of the central role The studies that examine the effectiveness of flexibil-
many Chinese companies play in the supply chains of ity and redundancy as responsive strategies are mainly
the world’s companies, the impact is being felt around anecdotal and conceptual, with some recent efforts to
the world: ‘The disruption is particularly acute in the quantify the cost and benefits of each. Some researchers
electronics and auto industries, but it is also affecting support flexibility over redundancy (Christopher and
pharmaceutical, metals, and a wide range of consumer Holweg 2011; Datta, Christopher, and Allen 2007; Sheffi
and industrial products, including surgical gowns and 2005). Others argue that redundant resources such as
masks.’ The COVID-19 outbreak has already impacted safety stock and emergency backup suppliers are pri-
operations and supply chain activities at a large scale. It mary solutions for enhancing supply chain responsive-
is reported that 94% of the companies in the Fortune ness (Azadeh et al. 2014; Blackhurst, Dunn, and Craig-
1000 list were facing supply chain disruptions due to head 2011; Jüttner and Maklan 2011; Ratick, Meacham,
the COVID-19 (Queiroz et al. 2020). Another study by and Aoyama 2008). Very few studies compare the two
Sáenz and Revilla (2014) shows that approximately 60% practices (Carvalho et al. 2012a; Parast and Shekarian
of the managers either do not actively work on SCRM 2019; Zsidisin and Wagner 2010). Fundamental ques-
or do not believe that their company’s risk management tions remain regarding the relative effectiveness of redun-
practices are effective. Thus, whether firms can benefit dant resources versus flexible resources, and under what
from their investments in supply chain disruption man- conditions each practice should be emphasized, given
agement practices remains debatable, which warrants the limited resources of organisations to develop capa-
further study to examine the overall impact of SCRM bilities associated with both flexibility and redundancy
practices. (Ghadge et al. 2013; Kamalahmadi and Parast 2016b).
Risks in a firm’s supply chains can arise from sources Besides, flexibility and redundancy have been studied in
such as the firm’s operations, suppliers, customers, and isolation. For example, Kamalahmadi and Parast (2017)
the environment (Christopher and Peck 2004; Ivanov examined the impact of redundancy practices on supply
et al. 2015). In this study, we focus on risks that chain resilience under supply and environmental risks. In
threaten supply chain flows by affecting the availabil- addition, Shekarian, Nooraie, and Parast (2020) exam-
ity of suppliers. These include supplier-induced risks ined the impact of lexibility and agility on supply chain
(e.g. the fire in a plant of one of Ericsson’s semi- responsiveness under demand and supply risks. No study
conductor suppliers, which cost Ericsson more than has examined the impact of both flexibility and redun-
400 million dollars (Baghalian, Rezapour, and Fara- dancy on supply chain resilience under supply and envi-
hani 2013)), and environmentally-induced risks (e.g. ronmental risks, to investigate the trade-off among them.
hurricanes such as Harvey, Irma, and Maria in the U.S. In this study, we consider both redundancy and flexibility
that disrupted a number of supply chains by making sup- in one model and propose a hybrid policy. Furthermore,
pliers in the affected areas unavailable for a period of those studies are based on one specific example; to add
time). To cope with such risks, traditional supply chains credibility to generalising our results, we ran a number
need to be redesigned with a focus on adding redundancy of simulations designed to reveal any potential bias from
and/or flexibility to their supply bases (Christopher and certain values of parameters. Thus, to the best of our
Peck 2004; De Giovanni and Massabò 2018; Fiksel et al. knowledge, this is the first study to examine the impact of
2015; Sreedevi and Saranga 2017). Firms can prepare for flexibility and redundancy in a supply chain under supply
disruptions and enhance their responsiveness through and environmental risks.
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 3

Using a two-stage mixed-integer programming model, This is followed by simulations and sensitivity analyses
this study contributes to the literature on supply chain on parameters in order to further examine the dynamics
flexibility and redundancy in three ways: 1) It provides of the model. Finally, we provide theoretical and manage-
insights into the ways a firm’s supplier selection and rial implications of this research, limitations of the study,
demand allocation are affected by the selection and devel- directions for future research, and the conclusions of this
opment of capabilities based on volume flexibility or study.
backup suppliers. 2) It investigates trade-offs among sup-
ply chain resilience, risk, and cost to determine how
2. Literature review
they interact. 3) It develops a more realistic supply chain
model where both primary suppliers and backup suppli- The study by Christopher and Peck (2004) was one of
ers are exposed to two types of disruptions: each supplier the first studies to recommend flexibility and redundancy
is exposed to the risk of an internal disruption; and sup- as key hedges against disruptions. Redundancy refers
pliers in the same region are exposed to the risk of an to practices in which a firm takes action in advance of
environmental disruption that could make all of them a disruption, incurring the cost of the action regard-
simultaneously unavailable. less of whether a disruption occurs. The most common
In line with Kamalahmadi and Parast (2017) and redundancy practices are keeping inventory, contracting
Kamalahmadi and Parast (2016a), the two-stage mixed- with backup suppliers, and investment in supplier pro-
integer programming model in this study simultaneously tection (Knemeyer, Zinn, and Eroglu 2009; Sodhi and Lee
examines the impact of supplier disruptions and environ- 2007; Tomlin 2006). Pre-disruption investment in adding
mental disruptions on a firm’s supply chain. We choose redundancy has made these practices less attractive for
backup suppliers as a redundancy practice and volume managers (Chopra and Sodhi 2014). A more attractive
flexibility as a flexibility practice. We use a numerical alternative is to add flexibility to the supply base, which
example from the literature as a motivating example for entails a cost that is incurred only in the event of a disrup-
our model. To examine the generalizability of our find- tion. (Fiksel et al. 2015; Tang and Tomlin 2008; Tomlin
ings, we extend our numerical example by creating mul- 2006). Table 1 presents a list of studies that emphasize
tiple settings, using simulations to test the robustness of flexibility and redundancy with respect to supply chain
our results. We quantify the costs and benefits of backup responsiveness.
suppliers and flexible suppliers, then compare them to The literature that compares flexible resources and
each other and to a supply chain without flexibility or redundant resources can be divided into two groups.
redundancy. Studies in the first group support flexibility over redun-
Moreover, we consider a hybrid practice where backup dancy. A theoretical study by Sheffi and Rice Jr (2005)
suppliers and volume flexibility are added to a supply measured supply chain disruptions by company perfor-
base together. We find that adding some level of flex- mance in terms of sales, production level, profits, and
ibility to the backup-suppliers practice is more desir- customer service. They provided a supply-chain view of
able in terms of lowering pre-disruption investment and a resilient enterprise and used several examples from
increasing supply chain responsiveness, while lowering practice to argue that while investing in flexibility and
expected total cost. To the best of our knowledge, this investing in redundancy both increase cost, investing
is the first study to provide an integrated, simultaneous in flexibility is the most important step a company can
assessment of the effectiveness of flexibility and redun- take to be more resilient, because it helps the firm not
dancy in improving firm responsiveness to supply chain only with mitigating disruptions but also with day-to-day
disruptions. We show that a hybrid flexible-backup sup- operations. By considering customer-service level, pro-
ply base can be a desirable practice for improving firm duction changes over time, the average inventory at each
responsiveness to supply chain disruptions. Our analyses distribution centre, and the total average network inven-
provide a more nuanced understanding of the impact of tory as the performance measures, Datta et al. (2007)
flexibility versus redundancy on mitigating supply chain developed an agent-based simulation model and after
disruptions. using a case study, they drew two conclusions: flexibil-
The rest of this paper is organised as follows. First, ity is needed in all elements of a supply network; and
we discuss the literature on flexibility and redundancy. balancing flexibility and redundancy in production and
Then, the model’s indices, parameters, and variables are dispatching capabilities is required to enhance the opera-
specified, and a two-stage mixed – integer programming tional resilience of a real-world supply network. Manuj
model is developed. In order to examine how the model and Mentzer (2008) extensively reviewed the literature
behaves, a numerical example is presented to show the and interviewed several senior supply chain executives,
trade-off among supply chain resilience, risk, and cost. then used as metrics of performance a reduction in cycle
4 M. KAMALAHMADI ET AL.

Table 1. Flexibility vs. Redundancy.


Author(s) and year Flexibility Redundancy Author(s) and year Flexibility Redundancy
Rice and Caniato 2003 ∗ ∗ Carvalho, Azevedo, and Cruz-Machado 2012b ∗ ∗
Christopher and Peck 2004 ∗ ∗ Klibi and Martel 2012 ∗ ∗
Chopra and Sodhi 2004 ∗ ∗ Ponis and Koronis 2012 ∗ ∗
Peck 2005 ∗ ∗ Wicher and Lenort 2013 ∗ ∗
Sheffi 2005 ∗ ∗ Pettit, Croxton, and Fiksel 2013 ∗ ∗
Sheffi and Rice Jr 2005 ∗ ∗ Johnson, Elliott, and Drake 2013 ∗ ∗
Sheffi 2006 ∗ ∗ Pal, Torstensson, and Mattila 2014 ∗ ∗
Tang 2006b ∗ ∗ Urciuoli et al. 2014 ∗ ∗
Tomlin 2006 ∗ ∗ Azadeh et al. 2014 ∗ ∗
Falasca, Zobel, and Cook 2007 ∗ ∗ Melnyk et al. 2014 ∗ ∗
Datta et al., 2007 ∗ ∗ Torabi, Baghersad, and Mansouri 2015 ∗ ∗
Ponomarov and Holcomb 2009 ∗ ∗ Kamalahmadi and Parast 2016a ∗ ∗
Briano et al. 2010 ∗ ∗ Kamalahmadi and Parast 2016b ∗ ∗
Pettit, Fiksel, and Croxton 2010 ∗ ∗ Kamalahmadi and Parast 2017 ∗ ∗
Zsidisin and Wagner 2010 ∗ ∗ Jain et al. 2017 ∗ ∗
Jüttner and Maklan 2011 ∗ ∗ Parast and Shekarian 2019 ∗
Carvalho, Duarte, and Cruz-Machado 2011 ∗ Pavlov et al. 2019a ∗
Gunasekaran, Rai, and Griffin 2011 ∗ Tan, Cai, and Zhang 2019 ∗
Christopher and Holweg 2011 ∗ Hosseini et al. 2019b
Blackhurst, Dunn, and Craighead 2011 ∗ Ivanov and Dolgui 2019
Carvalho et al. 2012a ∗ Shekarian, Nooraie, and Parast 2020

time, higher resilience, a higher level of relationship com- The researchers suggested a new perspective in supply
mitment, and increased flexibility; the researchers con- chain disruption management called low-certainty-need
cluded that supply chain flexibility is one of the impor- supply chains, and their study identified three key char-
tant factors that affect the selection of risk-management acteristics of low-certainty-need supply chains: structural
strategies. Christopher and Holweg (2011) used case complexity reduction, process and resource utilisation
study evidence of firms that responded to turbulence flexibility, and non-expensive parametric redundancy.
by adjusting their supply chains to emphasize flexibility Shekarian, Nooraie, and Parast (2020) developed a multi-
as the key in responding to disruptions. By consider- objective mixed-integer programing model by using mul-
ing stock-out costs, poor capacity utilisation, and costly tiple sites, multiple transportation channels, multiple
buffers as the metrics of performance, the researchers product planning, and multiple periods to examine the
showed that flexibility creates supply chain resilience by effect of supply chain flexibility and agility on respon-
enhancing prompt adaptability during turbulence. A lit- siveness under supply and demand risks. The researchers
erature review by Parast and Shekarian (2019) showed showed that investing in supply chain flexibility is more
that flexibility, redundancy, collaboration, and agility are profitable for firms in comparison with investing in
the most important antecedents of supply chain respon- agility.
siveness to cope with supply chain disruptions. They Studies in the second group support investment in
showed that more than 40% of the literature empha- redundancy over improved flexibility. Ratick, Meacham,
sized the crucial role of flexibility in enhancing supply and Aoyama (2008) developed a set cover location model
chain resilience. Pavlov et al. (2019a) developed a prob- for a supply chain that relies on emergency backup
lem of contingency-plan optimisation for seaport oper- and storage facilities, using minimisation of the separa-
ations under supply and network structural dynamics tion distance between the facilities as the performance
as a trade-off between sustainable resource utilisation measure. They found that this strategy is a powerful
and supply chain resilience. The researchers used a flex- tool to help firms achieve supply chain responsiveness.
ible policy that allowed utilisation of network potential Using systems theory and a resource-based view of firms
capacities according to structural dynamics and ratio- as the theoretical underpinnings, Blackhurst, Dunn, and
nal use of the structural functional network redundancy, Craighead (2011) conducted an empirical analysis of
which in turn made it possible to increase both the seven firms and reported that five of them considered
effectiveness and efficiency of network operations and holding safety stock throughout the supply chain as a
also to balance resilience and sustainability. In addi- primary strategy to mitigate the impact of disruptions.
tion, a theoretical study by Ivanov and Dolgui (2019) Sawik (2013) developed a mathematical model in order
suggested integrating the concepts of efficiency and to achieve the minimum cost for a firm that is inter-
resilience to enhance each other based on synergic effects. ested in investment in supplier protection and keeping
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 5

emergency inventory in the firm’s facilities; the model was a sequential optimisation process. In addition, whole
used to show the effectiveness of these strategies. Using sales contracts, option contracts, and purchases from the
a combination of optimisation and modeling approaches spot market were considered by Keyvanloo, Kimiagari,
with regard to average time in system, utility of resources, and Esfahanipour (2015) as sourcing alternatives for the
number of breakdowns, and total cost as the output met- buyer in a multi-period problem with time-correlated
rics, Azadeh et al. (2014) concluded that redundancy uncertain demand and prices. In another study, Chen,
and visibility are the most important principles of supply Zhao, and Zhou (2012) considered a periodic-review
chain responsiveness. Using a simulation-based analy- inventory system with an unreliable supplier that may
sis for decision makers to determine the strategies and be disrupted and a backup supplier (with higher prices
parameters best suited for building resilience in a sup- and limited capacity) that can be used during disruptions.
ply chain network, Tan, Cai, and Zhang (2019) showed They developed a mathematical model to use to deter-
that mitigation strategies are more suitable for long- mine the optimal order practice in the designed environ-
term disruptions, and contingency strategies are more ment. Torabi, Baghersad, and Mansouri (2015) studied
effective for short-term disruptions. In addition, they a supplier selection and allocation decision; they devel-
found that a strategy of redundant capacity enables a oped a model considering the three strategies of business
supply chain network to have lower time-to-recover, continuity plan, supplier fortification, and backup sup-
whereas the backup supply chain strategy has the lowest pliers. In their study, a backup supplier is a supplier
cost among all strategies. Hosseini et al. (2019b) devel- that provides items for the firm with a higher price and
oped a stochastic bi-objective mixed-integer program- longer lead time in the case of disruption in a primary
ming model for supplier selection and order allocation supplier. They showed that the backup-supplier strategy
to support decision-making in how and when to use along with other strategies provides a powerful tool for
proactive or reactive resilience strategies including sup- the firm to respond to disruptions. Kamalahmadi and
plier segregation, backup supplier contracts, supplier reli- Parast (2017) investigated the effectiveness of incorporat-
ability, surplus inventory, and supplier restoration capa- ing three types pf redundancy practices (pre-positioning
bility. Their results can help suppliers to find the opti- inventory, backup suppliers, and protected suppliers) into
mal set of operational decisions that enhance resilience a firm’s supply chain that is exposed to supply and
capabilities. environmental risks. Using a two-stage mixed-integer
This study evaluates three practices. The first prac- programming model, they considered contracting with
tice uses backup suppliers, a common implementation of backup suppliers by incorporating a fixed cost for reserv-
redundancy. The second practice uses flexible suppliers ing that supplier that was equal to the cost of building
(suppliers with volume flexibility), a common implemen- a warehouse. In this paper, we use an approach simi-
tation of flexibility. The third practice, called a hybrid lar to the approach that Kamalahmadi and Parast (2017)
practice, combines features of both backup suppliers and used for incorporating contracting cost. The main advan-
flexible suppliers. The following section reviews the liter- tage of having a backup supplier in all of these studies is
ature on the use of backup suppliers and the literature on that the firm contracts with a given supplier prior to any
the use of flexible suppliers. disruptions in primary suppliers. Consequently, when
disruptions happen, the firm is able to shift its demand
to the available backup suppliers.
2.1. Practice of using backup suppliers (Investment
in redundancy)
2.2. Practice of using flexible suppliers (Investment
A backup supplier is a secondary supplier that is used
in flexibility)
only when a primary supplier is disrupted. Contracting
with a backup supplier ensures that the flow of mate- Tomlin (2006) states that a contingent response to dis-
rial does not stop if disruption happens in other sources ruptions is viable only if suppliers have volume flexibility,
(Sodhi and Lee 2007; Tomlin 2006). Since the firm makes which is the capability to temporarily increase their pro-
the contract prior to any disruption, this strategy is classi- duction. Therefore, the manufacturer needs to reserve a
fied as a redundancy practice. Previous studies in backup- portion of the suppliers’ capacity to respond to increases
supplier strategies emphasize the importance of this strat- in demand due to disruptions. Ruiz-Torres and Mah-
egy in supply chain risk management. Hou, Zeng, and moodi (2006) presented a decision model to optimise the
Zhao (2010) studied a buy-back contract between a buyer demand allocation across a set of suppliers by evaluating
and a backup supplier when the main supplier experi- different costs associated with a supply network. Their
ences disruptions. Using an expected profit function, the model takes into account the capability of each supplier to
optimal decisions of the contract are derived through increase its output in case of disruption in other suppliers.
6 M. KAMALAHMADI ET AL.

Moritz and Pibernik (2008) extended the model devel- some level of flexibility to the backup-suppliers practice
oped by Ruiz-Torres and Mahmoodi (2006) by assuming (a practice that we call hybrid flexible backup) provides
different types of compensation mechanisms for suppli- a higher value for the firm compared to the backup-
ers and different purchasing costs, depending on the vol- suppliers practice alone. To the best of our knowledge,
ume of the orders. Yang and Qian (2008) advanced previ- our study is the first to propose such a hybrid strategy.
ous models through incorporating compensation mech-
anisms into the decision model, where they incorporate
3. Model development
the capability of suppliers to fulfill the shortfall caused by
other suppliers’ failures. They found the optimal number To investigate the effectiveness of redundant practices
of suppliers with their new extensions. Ruiz-Torres, Mah- and flexible practices, we develop a decision-making
moodi, and Zeng (2013) extended the previous models model to assess disruption risks using data from a real
and considered different demand points and different organisation. The supply chain that is modeled here is an
suppliers with different costs and reliability character- extension of the model proposed by Ruiz-Torres, Mah-
istics; the researchers expanded the traditional trans- moodi, and Zeng (2013). The model presented here is
portation problem to develop a mathematical model that based on the hybrid flexible-backup practice. In the next
considers contingency planning in the decision process. section, when we compare the practices numerically, we
Later, Kamalahmadi and Parast (2016a) extended Ruiz- adjust the model for the respective backup practice or
Torres, Mahmoodi, and Zeng (2013) by adding regional flexibility practice.
disruptions and the availability of different transporta- The manufacturer (firm) procures a single item in a
tion channels during disruptions into the model. The single period from a set of h possible suppliers: H =
results of these two studies emphasize the effectiveness {1, . . . , h}. All suppliers are first-tier suppliers; therefore,
of reserving volume flexibility in suppliers’ capacity to they provide the same items with different capacities,
enable available suppliers to deliver more items during costs, and reliabilities. These suppliers may be located in
disruptions. These studies on the use of flexible suppli- different regions. We distinguish suppliers of each region
ers suggest that the main advantage of reserving flexible by dividing them into separate sets. H r is defined as the
capacity is that without prior investment, firms can still set of suppliers in region r, R = {1, . . . , r}, where H 1 ∪
cope with disruptions. H 2 ∪ . . . ∪ H r = H. There is only one demand point, and
We contribute to the SCRM literature by develop- the demand is Q per cycle. Q belongs to only one cycle,
ing a decision-making model that examines the impact and demand is independent across cycles. Failure to sat-
of two common redundancy and flexibility practices on isfy the demand results in a per unit financial loss, M
improving firm responsiveness to supply chain disrup- (no backlogging). Definitions of other parameters and
tions: backup suppliers and flexible suppliers (volume variables are provided below, as well as in Table 2.
flexibility). We choose these two practices because the Suppliers’ characteristics: Adding each supplier h to
respective settings required for developing them are quite the supply base has a fixed cost, Fh , which is the manage-
similar; this provides an opportunity to study the rela- ment cost of working with that supplier (Berger, Gersten-
tive value of investment in each of them compared to feld, and Zeng 2004). The maximum capacity for supplier
the other. We develop a model for a single firm with h, Caph , is defined as the maximum number of items
a two-tier supply chain and multiple first-tier suppli- supplier h can produce per cycle. ch and th are the unit
ers with different costs and reliabilities. Our suppliers purchasing and transportation cost for each item ordered
are exposed to two types of disruption: supplier disrup- and transported from supplier h, respectively (Table 2). If
tion and environmental disruption. A supplier disruption the firm decides to select supplier h as a backup supplier,
disrupts a single supplier; an environmental disruption it incurs some costs regardless of whether a disruption
makes a number of suppliers in the same region unavail- occurs (Kamalahmadi and Parast 2017). Bh is defined
able. By simultaneously considering the risks of these as the fixed cost of contracting with a backup supplier
two types of disruption, we provide a more realistic sce- h. Furthermore, each supplier has a separate capacity,
nario that makes a clear distinction between the relia- ECaph , that restricts the maximum number of emergency
bility of a supplier and the reliability of the supplier’s items it can satisfy. With respect to the flexibility policy,
region. We find that both practices are effective in terms each supplier has a flexibility parameter, dh , which speci-
of improvements in cost reduction and service level; fies the quantity of additional items that the supplier can
however, the backup-suppliers practice outperforms the deliver above its normal order allocation, in case of dis-
flexible-suppliers practice for all the performance mea- ruption in other suppliers. (Ruiz-Torres, Mahmoodi, and
sures we consider. We then consider backup suppliers Zeng 2013). In other words„ dh is the extra amount of
and flexible suppliers together and show that adding deliveries that a supplier can provide to their order if any
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 7

Table 2. Decision Variables, Input Parameters, and Associated Types of events: There are two types of events that
Costs. affect suppliers’ operations. The first type is a semi-
Decision variables: super event, which is assumed to be location-specific
Zh A binary variable; 1 if supplier h is in supply base, 0 (Kamalahmadi and Parast 2016a; Sarkar and Mohapa-
otherwise. tra 2009; Sawik 2014). Due to the occurrence of such
Yh A binary variable; 1 if supplier h is selected as a
backup supplier, 0 otherwise. an event, all suppliers in that specific geographical area
ah Number of items allocated to normal supplier h. will be disrupted. The second type is a unique event
qs,h Quantity of items that flow from normal supplier h
in scenario s. that affects an individual supplier and causes disrup-
Eqs,h Quantity of emergency items that flow from tion. Parameters Pr∗ and Ph are assigned to each region
supplier h in scenario s above its base allocation.
vs,h Quantity of emergency items that flow from
and each supplier as the probability of occurrence of
backup supplier h in scenario s. semi-super and unique events, respectively (Table 2).
Us Number of unsatisfied items in scenario s. Disruption parameters: There are s possible scenar-
Input Parameters: ios in the problem, S = {1, . . . , s}, where each scenario is
S Set of scenarios.
R Set of regions. defined by a vector of disruption status of all suppliers
H Set of suppliers. regardless of the source of disruption (regional or sup-
Hr Set of suppliers in region r.
Fh Fixed management cost for supplier h.
plier disruption). In each scenario, when there is a disrup-
Bh Fixed backup contracting cost for supplier h. tion, the supplier(s) and/or the regions(s) can or cannot
Caph Capacity of supplier h. deliver their allocation. Let Sups.h represent the deliver-
ECaph Emergency capacity of supplier h when it is
selected as a backup supplier. ing status for supplier h in scenario s. Thus, Sups,h = 1
ch Purchasing cost of each item from supplier h in for available suppliers in scenario s, and Sups,h = 0 for
normal situation.
gh Purchasing cost of each item from supplier h in disrupted ones. The model considers all of the possible
disruption, when the supplier delivers items scenarios. Each scenario s has a probability of occurrence,
above its base allocation.
th Transportation cost of each item from supplier h.
πs , determined in a three-step process:
dh Supplier flexibility; ability of supplier h to satisfy Step 1: Calculate Psr , the probability of realising a
M
more items than its allocation.
Loss cost for each item.
disruption in suppliers located in region r, h ∈ H r , in
Q Demand of the buying firm. scenario s:
Sups,h Binary parameter related to the status of supplier 
h in scenario s. 1 if the supplier delivers, 0 Psr = [(1 − Sups,h )Ph + Sups,h (1 − Ph )] (1)
otherwise.
Ph Probability of occurrence of unique event in h∈H r
supplier h.
Pr∗ Probability of occurrence of semi-super event in Step 2: Calculate πsr , the probability of realising disrup-
region r. tions in region r and its suppliers,h ∈ H r , in scenario s:
Psr Probability of realising disruptions in suppliers
located in region r, h ∈ Hr , in scenario s. ⎧
πsr ⎪
⎪ (1−Pr∗ )Psr If there is no disruption in region r,
Probability of realising disruptions in region r and ⎨
its suppliers, h ∈ Hr , in scenario s. there is possible disruption in
πs πsr =
Probability of occurrence for scenario s. ⎪
⎪ individual suppliers there
Associated Costs:
⎩ ∗
Pr If there is disruption in region r
SMC Supplier management costs.
BCC Backup contracting costs. (2)
OCs Order cost in scenario s. Step 3: Calculate πs , the probability of occurrence of
TCs Transportation cost in scenario s.
scenario s:
PCs Premium cost in scenario s. 
LCs Loss cost in scenario s.
πs = πsr (3)
Performance Measures: r∈R
ETC Expected total cost
 
SL Expected service level It should also be noted that Psr = 1and Ps∗ +
ELS Expected number of lost sales
S S
(1 − Pr∗ )Psr = 1.
An example of how πs is calculated is shown in
Figure 1. Note that the scenario in which all suppliers in
disruption happens in other suppliers. The firm does not a particular region are disrupted yields results similar to
need to make significant investments to employ a flexibil- the scenario in which a regional disruption in that region
ity strategy; however, each supplier charges a higher unit makes all of the suppliers unavailable. The only difference
purchasing price for each item that it delivers above its between these two scenarios would be their probabilities
base allocation to make up for the failures of other sup- of occurrence. For example, when there is only one sup-
pliers. gh is defined as the unit purchasing cost of each plier in region 1 (Figure 1), although disruption in this
item ordered from supplier h above its base allocation. supplier yields a result similar to regional disruption in
8 M. KAMALAHMADI ET AL.

Figure 1. Decision Tree for Two Regions and One Supplier in Each Region.

region 1, the probabilities of occurrence for these two emergency items that flow from flexible (normal) sup-
scenarios are different. Thus, in order to aggregate iden- plier h above their base allocation in scenario s. Finally, let
tical scenarios with different probabilities of occurrence Us be the number of items not delivered to the demand
into one, and to avoid redundant scenarios, we divided point in scenario s. A list of the notations and symbols
the calculation of πsr into two methods in (2). In the first used throughout the paper is summarised in Table 2.
method, the probabilities of identical scenarios are added Associated costs: There are six different costs. All of
to combine two separate scenarios into one. However, in the costs are scenario dependent except supplier man-
the second method, other scenarios are considered. agement cost (SMC), which is the summation of man-
Variables: Our model has two stages. In the first stage, agement costs of working and collaborating with all sup-
the firm selects its suppliers and allocates the demand pliers, and backup contracting cost (BCC), which is the
among them. The backup-supplier selection or volume- summation of contracting costs with all of the backup
flexibility decisions are made at this stage as well. Zh and suppliers. OCs is the order cost in scenario s; it is based on
Yh are two binary variables related to the selection of sup- the quantity of items that flow from each supplier in each
plier h either in the supply base (as a flexible supplier) scenario as its normal allocation and backup allocation.
or in the backup-supplier base (as a backup supplier), TCs is the transportation cost in scenario s; it is based on
respectively. Variable as represents the normal alloca- the quantity of items that flow from each supplier in that
tion of suppliers in the supply base. The aforementioned scenario. (It can be either the quantity of normal items
variables constitute the first-stage decisions. In the sec- or the quantity of emergency items.) PCs is the premium
ond stage, different scenarios take place due to supplier cost in scenario s; it is defined as the costs that the firm
and/or regional disruptions. Based on the decisions in the needs to pay to order emergency items from flexible sup-
first stage, the manufacturer reallocates the demand in pliers above their base allocations. LCs is the loss cost for
order to cope with disruptions. Subsequently, the man- the firm in scenario s, calculated based on the number of
ufacturer incurs costs such as procurement, transporta- unsatisfied demand in that scenario. The associated costs
tion, and loss costs. Thus, the second-stage decisions we discussed before are calculated in (4)–(9) as shown
are conditioned on the first-stage decisions. These vari- below:
ables are scenario dependent; it means that the values of

these variables may change in each scenario. For exam- SMC = F h Zh (4)
ple, the number of items that flow from each supplier h∈H
to the demand point can be changed in each scenario. 
Let qs,h be the quantity of items that flow from normal BCC = Bh Yh (5)
supplier h in scenario s. vs,h represents the quantity of h∈H

emergency items that flow from backup supplier h in OCs = ch (ah + vs,h ) (6)
scenario s. Furthermore,Eqs,h , represents the quantity of h∈H
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 9


TCs = th (qs,h + vs,h ) (7) quantities that are supplied by available suppliers in dif-
h∈H ferent scenarios. Its values for each scenario would be
 equal to the difference between the quantity that flows
PCs = gh Eqs,h (8) from the supplier in that scenario and the normal alloca-
h∈H tion of that supplier. Constraint (C7) sets an upper bound
LCs = M ∗ Us (9) for the quantity of emergency items that are supplied by
the backup supplier based on its availability and its emer-
The expected total cost in the second stage would be: gency capacity. Constraint (C8) restricts the utilisation of
 emergency items from backup suppliers to the situation
Expected Total Cost = πs (OCs + TCs + PCs + LCs ) where at least one normal supplier is disrupted. There-
s∈S fore, when there is no disruption, there is no need to order
(10) emergency items from backup suppliers. Constraint (C9)
Here is the two-stage mixed-integer programming for determines the value for unsatisfied demand in each sce-
supplier selection and demand allocation, with the objec- nario; this is equivalent to the difference between the
tive of minimising expected total cost. total demand and the summation of quantities that flow
from all suppliers in the supply base and all backup sup-
Minimize SMC + BCC pliers. The last constraint determines the type of each
variable.

S
+ πs (OCs +TCs + PCs +LCs ) (11) The proposed model is a two-stage mixed-integer pro-
s=1 gramming model. In two-stage mixed-integer program-
ming (MIP), the first-stage decisions are determined
Subject to: through incorporating uncertainty of the future out-
 comes. The second-stage decisions are then evaluated
ah = Q (C1) after the potential future outcomes are assessed; thus, the
h∈H second-stage decisions are conditional on the first-stage
ah (1 + dh ) ≤ Zh Caph ∀h ∈ H (C2) decisions (Birge and Louveaux 2011; Rawls and Turn-
quist 2010). In our model, the Zh variables (the selection
Zh + Yh ≤ 1 ∀h ∈ H (C3) of suppliers), the Yh variables (the selection of backup
qs,h ≤ ah (1 + dh )Sups,h ∀s ∈ S&h ∈ H (C4) suppliers), and the ah variables (allocation of suppliers),
are first-stage decisions, which are determined before
qs,h ≥ ah Sups,h ∀s ∈ S&h ∈ H (C5)
there is knowledge of any specific scenario outcome. The
Eqs,h = (qs,h − ah )Sups,h ∀s ∈ S&h ∈ H (C6) objective function in the first stage is minimisation of
vs,h ≤ Yh Sups,h ECaph ∀s ∈ S&h ∈ H (C7) supplier management costs (SMC), backup contracting
  costs (BCC), and expected total cost of ordering, trans-
vs,h ≤ ah (1 − Sups,h ) ∀s ∈ S (C8) portation, premium, and loss costs of supplies over all
h∈H h∈H possible disruptions. The second-stage decisions are the

Us = Q − (qs,h + vs,h ) ∀s ∈ S (C9) qs,h variables (quantity flow from each normal supplier
h∈H
in each scenario), vs,h variables (quantity flow from each
backup supplier in each scenario), Eqs,h variables (quan-
Zh = {0, 1}, Yh = {0, 1}, ah ≥ 0, qs,h ≥ 0, vs,h ≥ 0, tity flow from each supplier above its base allocation
Eqs,h ≥ 0, Us ≥ 0 (C10) in each scenario), and Us variables (quantity of unsat-
isfied items in each scenario). The objective function
Constraint (C1) ensures that the demand will be allo- in the second stage is minimisation of the sum of the
cated to the suppliers. Constraint (C2) defines the capac- scenario-dependent costs including order, transporta-
ity utilisation limit for the flexible supplier as part of the tion, premium, and loss costs, considering the probability
contingency plan to respond to additional orders during of occurrence of each scenario.
disruptions. Constraint (C3) indicates that a supplier can The equivalent deterministic programme of this
be selected either as a flexible supplier or as a backup model is a mixed-integer programme (MIP) where the
supplier. Constraints (C4) and (C5) set an upper bound first–stage and second-stage decisions are integrated,
and a lower bound for the quantities that flow from each and the model is being solved in the extensive form.
normal supplier in each scenario based on the supplier’s The solution to the relatively small problems can be
allocation, flexibility rate, and its availability in that sce- determined by creating the equivalent deterministic MIP
nario. Constraint (C6) determines the values of extra model. This can be accomplished with commercial MIP
10 M. KAMALAHMADI ET AL.

software (Qin, Liu, and Tang 2013; Rawls and Turn- 4.1. Computational study
quist 2010). In this study, CPLEX 12.6 was used for
4.1.1. Case study
solving MIP problems. Our model first calculates the
We used the data provided by Ruiz-Torres, Mahmoodi,
probability of occurrence for each scenario utilising the
and Zeng (2013) and Kamalahmadi and Parast (2016a).
3-step algorithm explained. Next, the two-stage mixed-
This data is based on the observation of a manufacturer
integer programming provides the optimal solution for
of appliances. We assume three regions exist: H 1 , H 2 ,
the selection of flexible suppliers, backup suppliers, and
and H 3 . In each region, there are two suppliers with
the demand allocation among them, in order to minimise
different characteristics: H 1 = {1, 2}, H 2 = {3, 4}, H 3 =
the expected total network costs. In addition, the contin-
{5, 6}. In both normal and disruption situations, suppli-
gency plans for each scenario are included in the opti-
ers in region 3 have the lowest order price for items,
mal solution. In the next section, we conduct numerical
the largest production capacity, and the highest flexi-
analyses to evaluate the impact of each of the three prac-
bility rates. However, they are least reliable, and their
tices (using backup suppliers, using flexible suppliers, and
geographic distance from the firm gives them the highest
using a hybrid of both) on improving firm responsiveness
transportation rates. In addition, the probability of fail-
to disruptions.
ure due to the occurrence of a semi-super event is higher
for suppliers in region 3, due to geographical dispersion
and distance. The fixed backup contracting cost would
4. Numerical analysis be the same for all suppliers, and it would be five times
as much as the fixed management cost of suppliers. The
In Section 4.1, we show the application of our model
emergency capacity of each supplier is also assumed to be
using data from a company, which we call a case study.
equal to the maximum flexible capacity of that supplier
We show the relative effectiveness of each practice for the
(ECaph = Caph ∗ dh ). Suppliers’ information and char-
firm. In Section 4.1.2, to generalise the findings of the case
acteristics are shown in Table 3. In Section 4.3, we run a
study in Section 4.1.1 and to provide more insight, we run
number of sensitivity analyses to show that the results do
a number of simulations to focus on the core issue of flex-
not stem particularly from the choice of parameter val-
ibility vs. redundancy. In Section 4.2, we run a number
ues. The values for demand and loss cost are assumed to
of sensitivity analyses, where we start with the data used
be 4000 items and $400 per item, respectively.
in Section 4.1 and select different individual parameters
Practice 1. No Backup, No Flexibility
to vary, to show each parameter’s influence on the out-
First, consider the practice with no backup and no
come. Then we discuss the underlying factors that affect
flexibility. In this case, there is no delivery during disrup-
the optimal choice of flexible and backup suppliers. In all
tion, and there is no pre-disruption investment. The opti-
three sections, we use the following measures to compare
mal solution is the allocation of 1000 items to supplier 1,
the effectiveness of these strategies: 1) expected total cost
1500 items to supplier 2, and 1500 items to supplier 3.
(ETC); 2) expected service level (SL), defined as the prob-
(So 2500 items are allocated to region 1 and 1500 items
ability of zero loss to the manufacturer (Meena, Sarmah,
to region 2.) This result indicates that if the focal firm is
and Sarkar 2011); and 3) expected number of lost sales
not interested in implementing any responsive strategies
(ELS).

Table 3. Suppliers’ Information.


Regions Region 1 Region 2 Region 3
Pr∗ :Semi-super event probability 0.5% 1% 2%
Suppliers S1 S2 S3 S4 S5 S6
Caph : Capacity (items) 1000 1500 2000 2500 3000 3500
ECaph :Emergency Capacity (items) 300 600 1000 1500 2100 2800
dh :Supplier flexibility 30% 40% 50% 60% 70% 80%
ch : Normal purchasing cost ($/item) 25 22 20 15 12 10
gh : Disruption purchasing cost 35 32 30 25 22 20
($/item) 5 6 7 8 9 10
th :Transportation cost ($/item) 1% 2% 4% 6% 8% 10%
Ph :Supplier failure probability 1000 1000 1000 1000 1000 1000
Fh :Fixed management cost ($) 5000 5000 5000 5000 5000 5000
Bh :Fixed backup contracting cost ($)
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 11

Table 4. Suppliers’ and Regions’ Allocations in the Flexibility Table 5. Suppliers’ and Regions’ Allocations in the Backup
Setting. Setting.
Region 1 Region 2 Region 3 Region 1 Region 2 Region 3
Regions’ Allocations 1525 1250 1225 Regions’ Allocations 1500 - 2500
Suppliers’ Allocations S1 S2 S3 S4 S5 S6 Suppliers’ Allocations S1 S2 S3 S4 S5 S6
760 765 580 670 630 595 - 1500 1000 1500 2100 2500

for mitigating disruption, then in the selection of sup- $111, 468. This is a 31.5% reduction in expected total cost
pliers, reliability factors outweigh cost factors. As we see, compared to the first practice, which has no backup and
expensive but reliable suppliers are selected in the supply no flexibility. The result shows that by creating a region-
base; less expensive and less reliable suppliers, which are alised supply base, the set of suppliers in the second plan
mainly located in the third region, are not selected. For a would be able to meet the requirement in all scenarios
practice with no backup and no flexibility, the expected when there is a single or double supplier failure or a sin-
total cost is $162, 640.80, the expected service level is gle regional disruption; however, when more than one
91.7%, and the expected lost sales is 127.4 units. region fails or more than two suppliers fail, some part
Practice 2. Only Flexibility of the demand will remain unsatisfied at the end of the
When all of the contingency characteristics related to cycle. The findings show that the service level (success
volume flexibility are added to the problem, the optimal rate in mitigating supply disruption) is 99%. This is much
solution is a selection of all suppliers in the supply base. higher than the 91.7% service level in the first practice.
The allocation for each supplier and consequently each Furthermore, the expected lost sales was reduced from
region is shown in Table 4. 127.4 units in the first practice to 8.1 units in the flexibility
Figure 2 presents the optimal demand allocation (nor- practice, a 93.7% reduction.
mal allocation) of the suppliers in normal conditions Practice 3. Only Backup
and their maximum output in disruptions (i.e. the quan- When all of the redundancy characteristics related to
tity that flows from each supplier changes in the ranges backup suppliers are added to the model, the optimal
shown in Figure 2 in different scenarios). The result indi- solution is the selection of two suppliers (S2 and S6) in
cates that if the focal firm decides to design a practice with the supply base and three suppliers as backup suppliers
volume flexibility in suppliers, there is a balance between (S3, S4, and S5). Table 5 shows the result, where backup
reliability, flexibility, and costs. In normal conditions, the suppliers are highlighted in gray.
reliability factor determines the allocation of suppliers; The result indicates that with this practice, expensive
highly reliable suppliers receive more demand allocation. supplier 1 was dropped from the supply base. The num-
In disruptions, cost factors and flexibility rates determine ber of primary suppliers was reduced to two. Supplier 6,
the allocation of suppliers. As Figure 2 shows, the maxi- the least reliable and least expensive supplier, was selected
mum output for all suppliers is fairly equal (except sup- along with supplier 2, a reliable and expensive supplier.
plier 3, which has the minimum allocation). This means With this selection, the reliability and cost is balanced in
that the range of flexible allocation is wider for suppliers the supply base. Then, we have three backup suppliers. It
in the third region in comparison with suppliers in the suggests that the optimal solution is to keep fewer sup-
first and second regions. In other words, higher flexibil- pliers in the supply base with higher allocations to each
ity leads to lower normal allocation and higher utilisation (compared to the previous two runs), while pre-investing
in disruptions. The expected total cost for this model is to contract with two or more backup suppliers to ensure

Figure 2. Optimal Allocation and Maximum Output in the Flexibility Practice.


12 M. KAMALAHMADI ET AL.

Table 6. Suppliers’ and Regions’ Allocations in the Hybrid Prac- other hand, reliable suppliers of each region (except sup-
tice. plier 1) were selected as backup suppliers. Disruptions
Region 1 Region 2 Region 3 are unavoidable; therefore, the optimal solution ensures
Regions that when disruption occurs in primary suppliers, reliable
Allocations 1071 1539 1390 backup suppliers are available to prevent interruption
Suppliers S1 S2 S3 S4 S5 S6 in the flow of goods. The number of backup suppliers
Allocations - 1071 1000 1539 2100 1390 was reduced from 3 in the backup practice to 2 in the
Maximum output - 1499 1000 2462 2100 2502
hybrid practice, which in turn reduces the pre-disruption
investment costs.
The hybrid practice with both flexible and backup
that the flow of goods does not stop when primary sup- resources addressed shortcomings of the previous polices.
pliers are disrupted. The reason that three suppliers were In the flexibility practice, since the flexible capacity of
selected as backups is that in our model, backup suppli- suppliers was limited, they would not be able to meet
ers are also exposed to supplier and regional disruptions. manufacturer demand when more than two suppliers (or
Our model then considers all of the combinations of one region) were disrupted simultaneously. In the backup
failures in primary and backup suppliers, and suggests practice, when backup suppliers were not able to make
contingency plans for each scenario. In the current result, up the failure of one of the primary suppliers, the other
supplier 5 is in the priority to be used in disruptions, primary supplier did not have the ability to increase its
because of its low prices. However, in case the emergency output, as its capacity was already fully assigned to sat-
capacity of this supplier is not sufficient or this supplier isfy the manufacturer’s order. However, in this hybrid
is unavailable due to a disruption, suppliers 4 and 3 will practice, the combination of backup and flexible sup-
be used to make up for the failure of primary suppliers. pliers enables the manufacturer to increase its expected
The expected total cost is $111, 483.50 for the backup service level with a lower pre-disruption investment. In
practice. This is a 31.5% reduction in expected total cost case of a disruption in a primary supplier, the manufac-
compared to the first practice, which has no backup and turer would check with backup suppliers. If they were
no flexibility. The result shows that by creating a region- available and could completely make up for the failure
alised supply base with primary and backup suppliers, the of primary suppliers, their emergency allocations would
set of suppliers in this plan would be able to meet the be used. However, if backup suppliers were not available
requirement in all scenarios except when more than one or did not have sufficient capacity, the flexible capacity
of the backup suppliers is disrupted simultaneously with of available primary suppliers would be used to cover
one of the primary suppliers in the supply base. The find- the failure of other suppliers. The expected total cost for
ings show that the expected service level is 99.5%. This is this practice is $107, 944.5. This is about a 34% reduc-
much higher than the 91.7% service level in the first prac- tion in expected total cost compared to the first practice
tice. The expected lost sales was reduced from 127.4 units with no disruption plans and a 3% reduction compared
in the first practice to 5.9 units in the backup practice, a to the second and third practices with only flexibility or
95.4% reduction. backup. The result shows that by creating a regionalised
Practice 4. Hybrid Practice with Backup and Flexibility supply base with flexible and backup suppliers, the set of
Here, all of the characteristics related to flexible and suppliers in this plan would be able to meet the require-
backup suppliers are added to the model. The optimal ments of the manufacturer in all scenarios except in the
solution is the selection of three primary flexible suppli- rare event of simultaneous disruptions in more than three
ers (S2, S4, and S6) and two backup suppliers (S3 and suppliers. The findings show that the expected service
S5). Table 6 shows the result, where backup suppliers are level is 99.6%. This is much higher than the 91.7% service
highlighted in gray. level when there was no plan for disruptions. Further-
The result indicates that the set of suppliers selected more, the expected lost sales was reduced from 127.42 in
for Practice 4 is similar to the set for Practice 3 (which the first practice to 4.32 in the hybrid practice, a 96.7%
has only backup suppliers), except that when volume flex- reduction.
ibility was allowed in Practice 4, supplier S4 was switched The summary results for the expected total cost,
from the backup supplier base to the primary flexible sup- expected service level, and expected lost sales for all
pliers. The allocations of S2 and S6 were lowered and a four practices are shown in Figure 3. It is shown how
part of their capacity was reserved for disruptions. The the hybrid practice outperforms all other practices in all
emergency allocations of suppliers S3 and S5 did not three measures of interest.
change. We observe that the least expensive supplier in In the following section, we develop a simulation
each region was selected as a primary supplier. On the model in a number of different settings, to see if the
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 13

Figure 3. Results of the Computational Study.

results obtained in this example are robust and can be Table 7 shows the VSS and EVPI metrics for the case
generalised. study. The VSS metrics are significant, ranging from
1.55% to 4.3% based on each practice. Therefore, the
4.1.2. Stochastic measures effort to create and solve the stochastic model in this
Two important aspects of formulating and processing problem (rather than a deterministic problem) can con-
stochastic programming (SP) problems are to obtain siderably decrease the company’s expected total cost. In
stochastic information regarding the family of models addition, the EVPI metrics are also significant, ranging
and to compare the related bounds (Valente et al. 2009). from 2.58% to 8.55% based on each practice. Therefore,
The expected value of perfect information (EVPI) and the better forecasting and efforts to gain more information
value of stochastic solution (VSS) are the motivation for about the future and more control over stochastic param-
using SP methods (Nazemi, Parragh, and Gutjahr 2020). eters can considerably decrease the company’s expected
Assuming that it is possible to indicate beforehand which total cost.
scenario will occur, it will be possible to make a decision
according to that specific scenario and obtain the opti- 4.1.3. Simulations
mal solution for that particular scenario. This situation in The goals of this section are to generalise the findings of
SP is called a ‘perfect information’ situation, and the cor- the last section into a broader setting and also to delve
responding solution is called a ‘wait-and-see’ (WS) solu- deeper into the core issue of flexibility vs. redundancy.
tion. The difference between the solutions of SP and WS We run a number of simulations in different settings to
is called EVPI. EVPI shows the loss of profits caused by provide insight into the performance when using only
not being completely ready to cope with disruption. On flexible suppliers, only backup suppliers, and the hybrid
the other hand, instead of spending time to develop an SP model.
model, a deterministic model based on the mean value We set our parameters to get values randomly from
of the parameters in each scenario can be used in order some known distributions. The first case considers a sim-
to obtain the solutions that are called ‘expected value’ ple structure for the supply chain, with a manufacturer
(EV) solutions. This method can lead to high costs or and two suppliers: a reliable but expensive supplier from
adverse consequences in optimising the solutions. There- a reliable region, and an unreliable but inexpensive sup-
fore, the loss incurred by ignoring stochastic parameters, plier (with higher capacity) from an unreliable region.
which is called VSS, is equal to the difference between the We define the values for the parameters in a way that we
expected performance of the EV solution (i.e. EEV) and provide a trade-off between risks (reliability) and costs
the solution of SP. VSS and EVPI can also be shown as in the choice between the two suppliers. The values for
follows (Khalilabadi, Zegordi, and Nikbakhsh 2020): the parameters were drawn randomly from uniform dis-
tributions as shown in Table 8. Demand is fixed at 8000
VSS = ZEEV − ZSP (12) items, loss costs are fixed at $400 per unit, and fixed man-
agement cost per supplier is fixed at $1000. The values
EVPI = ZSP − ZWS (13) for fixed backup contracting cost and flexibility rate are

Table 7. EVPI and VSS metrics for the case study.


Practice SP ($) EEV ($) WS ($) VSS ($) EVPI ($) VSS (%) EVPI (%)
No Backup/ No Flexibility 162,640 165,160 148,734 2,520 13,905 1.55 8.55
Only Flexibility 111,468 114,667 106,730 3,199 4,737 2.87 4.25
Only Backup 111,483 114,994 107,045 3,511 4,437 3.15 3.98
Backup and Flexibility 107,944 112,585 105,159 4,641 2,784 4.30 2.58
14 M. KAMALAHMADI ET AL.

Table 8. Regions’ and Suppliers’ Characteristics – Simulation As shown in Figure 4, the flexibility and backup prac-
with 2 Suppliers. tices both reduce the expected total cost in almost all
Parameters Region 1 Region 2 trials; the backup practice is more effective, producing the
Pr∗ :Semi-super lowest ETC in 28 of the 30 trials. Furthermore, although
event probability ∼ U[0,0.05] ∼ U[0.05,0.1]
the expected service level (probability of zero loss) is
Suppliers Supplier 1 Supplier 2
Ph :Supplier failure ∼ U[0,0.05] ∼ U[0.05,0.2] constant across all three practices, the expected number
probability ∼ U[8000,12000] ∼ U[10000,15000] of lost sales has been lowered significantly using either
Caph : Capacity (items) ∼ U[4000,6000] ∼ U[5000,7500]
ECaph :Emergency capacity ∼ U[20,30] ∼ U[10,20] backup or flexibility, with the backup practice being more
(items) ∼ U[30,40] ∼ U[20,30] effective than the flexibility practice. The results of the
ch : Normal purchasing ∼ U[5,10] ∼ U[10,15]
cost ($/item) ∼ U[5000,10000] ∼ U[5000,10000]
two-sample t-test for the means of ETC and ELS con-
gh : Disruption purchasing ∼ U[0,1] ∼ U[0,1] firm that the mean expected total cost and mean expected
cost ($/item) lost sales are lower for the backup-suppliers practice com-
th :Transportation cost
($/item) pared to the flexible-suppliers practice and having no
Bh :Fixed backup SCRM plan (p < .01). Furthermore, the mean ETC and
contracting cost ($)
dh :Supplier flexibility mean ELS are lower for the flexibility practice compared
to having no SCRM plan (p < .01).
In Figure 4, we see only minor improvements in terms
drawn from the same distributions for both suppliers to of both ETC and ELS in trials 12, 24, 25, and 30. On the
see how the model picks suppliers based on the other cost other hand, significant improvement is observed in tri-
and risk parameters. Backup contracting cost per supplier als 13 and 14. To see why, let’s look at the reliability of
is 5–10 times the management cost per supplier. The sup- region 1 and supplier 1 in these trials. In trials 12, 24, 25,
plier flexibility parameter is drawn uniformly from the and 30, the disruption probability of supplier 1 and region
range of [0,1]. Table 8 shows how each of the parameters 1 is less than 1% ((1 − P1∗ )(1 − P1 ) = 0.99). In trials 12
is assigned values. and 13, (1 − P1∗ )(1 − P1 ) = 0.9; i.e. there is a 10% chance
We ran 30 trials generating data as explained in of losing supplier 1 due to regional or supplier disrup-
Table 8. The results for expected total cost and expected tion. This result indicates that when there exists a reliable
lost sales for each trial under each practice are shown in supplier and region, it may not be worthwhile switch-
Figure 4 (see the Appendix for the detailed results). Note ing to flexibility or redundancy practices. However, when
that a hybrid flexible-backup practice is not applicable to uncertainty is high, firms can achieve a lot by pursuing
this case with only two suppliers. Also, all three practices either flexibility or redundancy practices.
yield the same expected service level, so that is not shown Table 9 shows the VSS and EVPI metrics for the first
in Figure 4. simulation of 30 trials. The VSS metrics are significant,

Figure 4. ETC and ELS Comparison for the Three Practices in Each of 30 Trials.
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 15

Table 9. Average EVPI and VSS metrics for the first simulation of 30 trials.
Practice SP ($) EEV ($) WS ($) VSS ($) EVPI ($) VSS (%) EVPI (%)
No Backup/ No Flexibility 410,287 421816.1 378,982.1 11,529.06 31,304.9 2.81 7.63
Only Flexibility 360,510 369126.2 342,700.8 8,616.189 17,809.19 2.39 4.94
Only Backup 308,481 317642.9 295,895 9,161.886 12,586.02 2.97 4.08

ranging from 2.39% to 2.97% based on each practice. In Table 10 shows the distributions used to set the values
addition, the EVPI metrics are also significant, ranging of parameters. Demand, loss cost per item, and supplier
from 4.08% to 7.63% based on each practice. management cost per supplier are again fixed at 8000
Our findings provide evidence that supports the role items, $400 per item, and $1000 per supplier, respectively.
of reliability as the key factor in determining which sup- We ran 30 trials generating data as explained in
plier should be selected as primary supplier. (Recall that Table 10. The results for expected total cost, expected ser-
supplier 1 is more reliable, more expensive, and located in vice level, and expected lost sales for all four practices are
a more reliable region compared to supplier 2.) Supplier shown in Figure 6. (Note that in expected service level
1 is the sole supplier when there is no flexibility prac- and expected lost sales figures, the hybrid and redun-
tice and no backup practice, and supplier 1 is the primary dancy practices have a very similar pattern.) Please see
supplier when the backup practice is used (even though the Appendix for the detailed results.
the backup contracting cost was lower for supplier 1 in The results for expected total cost and expected lost
18 out of 30 trials). When the flexibility practice is used, sales are consistent with the results of the previous anal-
both suppliers are chosen; supplier 1 gets the majority ysis. Here again, the flexible-suppliers practice and the
of the allocation, as shown in Figure 5. In Figure 5, the backup-suppliers practice lead to lower mean expected
suppliers’ allocations and the value of supplier 1’s flex- total cost and lower mean expected lost sales (p < .01),
ibility rate in each trial are shown. Even though other with the backup-suppliers practice outperforming the
factors (such as disruption probabilities) are also con- flexible-suppliers practice (p < .01). However, a compar-
tributing to the changes in allocations, there is a strong ison of service levels provides new insights. We see that
negative correlation (−0.98) between supplier 1’s alloca- the backup-suppliers practice improved the mean service
tion and its flexibility rate. When supplier 1’s flexibility level (p < .01). On the other hand, even though the mean
is higher, less demand allocation goes to this supplier, SL is lower in the flexible-suppliers practice compared
and a larger portion of supplier 1’s capacity is reserved to the base practice, this result is not statistically signif-
for use during disruptions. This makes the supply chain icant (p > .10). Nevertheless, the last graph shows that
more responsive during disruptions by providing a larger mean expected lost sales was reduced with the flexible-
available capacity and thus reducing the loss of sales. suppliers practice (p < .01). Next, when we incorporate a
Consequently, this leads to a reduction in the expected hybrid flexible-backup-suppliers practice in our analysis,
total cost. we see that it can provide the same expected SL and ELS as
Now, in order to analyze the hybrid practice, we extend the backup-suppliers practice, with some improvements
the analyses to the case of two regions, with two suppli- in the mean ETC (95% confidence interval for the dif-
ers in each region. Again, we assume suppliers in the first ference between means: [1407.68, 5760.52]). To further
region (the more reliable region) are more reliable, more analyze why the expected total cost is lower in the hybrid
expensive, and have lower capacities compared to region practice, we present the pre-disruption investment costs
2 suppliers. Backup contracting costs and flexibility rates as a percentage of the overall expected total cost for the
are drawn from the same distributions for all suppliers. hybrid and backup practices in Figure 7.

Figure 5. Suppliers’ Allocation and Supplier 1’s Flexibility Rate Changes in Flexibility Practice.
16 M. KAMALAHMADI ET AL.

Table 10. Regions’ and Suppliers’ Characteristics – Simulation with 4 Suppliers.


Parameters Region 1 Region 2
Pr∗ :Semi-super event probability ∼ U[0,0.05] ∼ U[0.05,0.1]
Suppliers Supplier 1 Supplier 2 Supplier 3 Supplier 4
Ph :Supplier failure probability ∼ U[0,0.05] ∼ U[0.05,0.10] ∼ U[0.05,0.10] ∼ U[0.05,0.2]
Caph : Capacity (items) ∼ U[8000,12000] ∼ U[8000,12000] ∼ U[10000,15000] ∼ U[10000,15000]
ECaph :Emergency capacity (items) ∼ U[4000,6000] ∼ U[4000,6000] ∼ U[5000,7500] ∼ U[5000,7500]
ch : Normal purchasing cost ($/item) ∼ U[20,30] ∼ U[17,27] ∼ U[14,24] ∼ U[10,20]
gh : Disruption purchasing cost ($/item) ∼ U[30,40] ∼ U[27,37] ∼ U[24,34] ∼ U[20,30]
th :Transportation cost ($/item) ∼ U[5,10] ∼ U[7,12] ∼ U[8,13] ∼ U[10,15]
Bh :Fixed backup contracting cost ($) ∼ U[5000,10000] ∼ U[5000,10000] ∼ U[5000,10000] ∼ U[5000,10000]
dh :Supplier flexibility ∼ U[0,1] ∼ U[0,1] ∼ U[0,1] ∼ U[0,1]

Figure 6. ETC, SL, and ELS Comparison for the Four Practices in Each Trial.

In the backup practice, the model often chooses That is why the pre-disruption investment is lower in the
one primary and three backup suppliers. The choice of hybrid model (p < .01).
primary supplier changes in each trial, based on the value The mean pre-disruption investment percentage is
of backup contracting costs and other parameters. 6.98% in the hybrid model and 8.62% in the backup
In the hybrid model though, when we allow for some model. As we see in Figure 7, having two flexible and
level of flexibility, the model chooses two flexible primary two backup suppliers outperforms one primary and three
suppliers and two backup suppliers in most of the cases. backup suppliers in terms of pre-disruption investment.
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 17

Figure 7. Pre-Disruption Investment in Redundancy (Backup Contracting) as a Percentage of Expected Total Cost, for the Hybrid and
Backup Practices.

Basically, a portion of that pre-disruption investment The analysis of this section reinforces the findings of
in contracting with a third backup supplier is used for our computational study:
post-disruption investment in paying extra dollars per
unit that flexible suppliers satisfy above their base allo- 1) Even though new costs are involved when it comes to
cation, and the other portion has been saved. The hybrid investment in flexibility and redundancy, the overall
model may be more interesting for managers, to avoid a benefits of this investment outweigh the costs when
tendency to overlook the effectiveness of investment in all possible states of nature are considered. Benefits
redundancy for potential low-likelihood events (Chopra include reduction in expected total cost, reduction
and Sodhi 2014). in expected lost sales, and improvement in expected
Analyzing the detailed results of each practice, sup- service level.
plier 1 is again chosen as the sole supplier for the case 2) The backup practice is a more effective strategy
with no flexibility and no redundancy. In the flexibility compared to volume flexibility, but it involves pre-
practice, 3 or 4 suppliers are chosen in different trials, investment in redundant resources. Therefore, a
where more flexibility in suppliers leads to an increase backup practice is more appropriate for a risk-averse
in the number of suppliers and a more even distribution firm, and a flexibility practice can be an alternative
of demand across them. Also, region 1 suppliers (reli- for a firm that avoids pre-investment in redundant
able region and suppliers) receive most of the demand resources.
allocation. Finally, in the hybrid model, two primary flex- 3) The hybrid model outperforms the backup strategy
ible suppliers and two backup suppliers were chosen in by achieving the same expected service level and
a way that reliability is balanced between the primary expected lost sales with a lower expected total cost
supply base and backup supply base (one reliable and one and lower pre-disruption investment. The hybrid
unreliable supplier in each base). model shows that a combination of volume flexibil-
Table 11 shows the VSS and EVPI metrics for the ity and backup suppliers is superior to an investment
case study. VSS metrics are significant, ranging from in only one of them, from both a risk and a cost
2.25% to 5.51% based on each practice. The EVPI met- standpoint.
rics are also significant, ranging from 3.08% to 7.87%
based on each practice. Therefore, the efforts to create Finally, we would like to summarise the key observa-
and solve the stochastic model as well as the efforts to tion with regard to the supplier selection and demand
gain more information about the future and more con- allocation of each practice. It turned out that reliability
trol over stochastic parameters can considerably decrease is playing the major role in determining which suppli-
the company’s expected total cost. Thus, a proportion of ers should be selected in either the primary or backup
EVPI can be offered to retailers as an incentive to encour- supply base. We observed that with no SCRM practice in
age them to announce their exact quantity for the next place, the most reliable supplier was chosen, as the firm
period. tries to minimise the cost and maximize the service level.

Table 11. Average EVPI and VSS metrics for the second simulation of 30 trials.
Practice SP ($) EEV ($) WS ($) VSS ($) EVPI ($) VSS (%) EVPI (%)
No Backup/ No Flexibility 402,980 412,047 371,265 9,067 31,714 2.25 7.87
Only Flexibility 345,980 359,369 332,348 13,389 13,632 3.87 3.94
Only Backup 250,572 262,299 241,576 11,727 8,996 4.68 3.59
Backup and Flexibility 243,476 256,891 235,977 13,416 7,499 5.51 3.08
18 M. KAMALAHMADI ET AL.

To design a flexibility practice, the more flexibility that is selected in the supply base even though it is an expen-
suppliers could insert in their capacity, the more even sive one. In all of the first three experiments, the model
the firm’s distribution of demand would be, with a conse- has made a balance between risks and costs in the supply
quent reduction in the expected total costs and expected base by selecting reliable suppliers in the first region and
lost sales. Reliable suppliers would still get more demand cheap suppliers in the third region as the primary suppli-
compared to less reliable suppliers. To design an effec- ers, and selecting mid-reliable mid-expensive suppliers in
tive redundancy practice, having a strong backup supply the second region (S3 and S4) as backup suppliers. Alter-
base seems to be more important than a strong pri- natively, in each of the second three experiments (when
mary supply base. With multiple suppliers in the backup S4, S5, and S6 are respectively set to perfectly reliable),
supply base, the firm can ensure that the flow of mate- there is a supplier that has characteristics of both relia-
rial will not stop when primary suppliers and one of bility and cheap prices. Therefore, this supplier receives
the backup suppliers are unavailable. Finally, an effec- a great portion of the demand. Consequently, the relia-
tive hybrid practice is achieved when a balance between bility of the supply base was increased and the number
costs and reliability is made in each supply base. The of backup suppliers was reduced to one. To minimise the
firm can do that by choosing both reliable and unreli- impact of regional disruptions, in each scenario at least
able (but inexpensive) suppliers in each of the supply one supplier from each region is selected in the supply
bases. base.
In the following section, we run a number of sensitiv-
ity analyses to study the effects of each of the parameters 4.2.2. Sensitivity analysis of the reliability of regions
of our model on the optimal choice of suppliers and In contrast to the previous sensitivity analysis, we assume
allocation of demand for the computational study we that one of the regions is highly volatile and unreliable.
presented in the last section. For each of the three regions, we ran a simulation where
we set the disruption probability of that region to 10%.
The optimal allocations resulting from the three simula-
4.2. Sensitivity analyses tions are shown in Figure 9. In each set, the allocations
of primary flexible suppliers are shown in black, and
Since the results of supplier selection and order allo-
emergency allocations of backup suppliers are shown in
cation are highly dependent on the parameters of our
gray.
model, and to obtain a more nuanced understanding of
Based on Figure 9, in all three experiments, two
the disruption models, we conduct a number of sensitiv-
backup suppliers are selected. In the first case, when the
ity analyses to examine the performance of the model and
first region is highly unreliable, supplier S1 is removed,
the solutions regarding these decisions in different set-
and supplier S6 (very inexpensive) is selected as backup
tings. All sensitivity analyses are conducted in the setting
supplier using its full emergency capacity utilisation. In
of our computational study.
the second case, when the second region is highly unre-
liable, suppliers in this region are selected as backup
4.2.1. Sensitivity analysis of the reliability of suppliers suppliers to avoid the reduction in the reliability of the
In order to analyze the effect of changes in the relia- supply base. Suppliers in the other regions are selected as
bility of suppliers on the final solution, we consider the primary flexible suppliers in the supply base. In the last
availability of one highly reliable supplier. For this pur- scenario, when the third region’s reliability is low, sup-
pose, we put the disruption probability of each of the plier S5 is selected as backup supplier from this region.
suppliers equal to zero, once at a time. The result of the We should keep in mind that supplier S6’s low prices still
optimal allocation of suppliers in each setting is shown keep this supplier in the supply base. Supplier 1, which
in Figure 8; the allocation of primary flexible suppliers is was dropped in the first case, is back in the second and
shown in black, and the allocation of backup suppliers is third scenarios to increase the reliability of the supply
shown in gray. base.
Recall that in the computational study example, the
first three suppliers are reliable but expensive, and the 4.2.3. Sensitivity analysis of the flexibility and
second three suppliers offer cheaper prices with less reli- emergency capacity of suppliers
ability. Now the results of the first three experiments, In this section, the flexibility rate and emergency capac-
where we set the respective disruption probability of sup- ity of suppliers are analyzed to see how changes in these
pliers S1, S2, and S3 equal to zero, have some similarities. parameters can change the final solutions. For the pur-
The only difference is the first experiment, where sup- pose of analysis of flexibility rate, supplier S6 was selected,
plier S1’s reliability outweighs its expensive prices, and it as it has the highest flexibility and the least expensive
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 19

Figure 8. Optimal Allocations of Primary Suppliers (black) and Backup Suppliers (gray) with One Reliable Supplier.

Figure 9. Optimal Allocations of Primary Suppliers (black) and Backup Suppliers (gray) with One Unreliable Region.

prices according to the case study example. The result of Next, we choose supplier S3 to analyze how changes
the analysis is shown in Figure 10, where the black colour in its emergency capacity affect the final solutions. This
represents the allocation of primary suppliers and gray supplier is selected for this analysis as it is mainly selected
colour represents the emergency allocation of backup as a backup supplier in most of the previous analyses. The
suppliers. result is shown in Figure 12.
Based on Figure 10, when the flexibility rate of supplier The result in Figure 12 indicates that when the emer-
S6 is more than 0.5, suggesting that the supplier is able to gency capacity of supplier S3 is low, it is not selected,
increase its output to more than 1.5 of its base allocation, either as a backup supplier or as a primary one. However,
this supplier is selected as a flexible supplier. However, when this supplier is able to deliver a great number of
when its flexibility capability drops, it is removed from items in disruptions, it would be selected as a backup sup-
the primary supply base, and it is selected as a backup plier. We see that the set of suppliers and their allocations
supplier due to its cheap prices. were totally changed in this case.
20 M. KAMALAHMADI ET AL.

Figure 10. Optimal Allocations of Primary Suppliers (black) and Backup Suppliers (gray) with One Flexible/Inflexible Supplier.

4.2.4. Sensitivity analysis of the fixed backup 5.1. Theoretical contributions and managerial
contracting cost of suppliers implications
The result of this analysis is shown in Table 12. As
According to Christopher and Peck (2004), traditional
expected, when the pre-disruption investment in con-
cost-driven supply chains need to be redesigned with an
tracting with backup suppliers is substantial, this invest-
approach that accounts for uncertainties that threaten
ment cannot be justified. Therefore, the number of
supply chain flows, incorporating factors such as suppli-
backup suppliers drops as this cost increases.
ers’ reliabilities and the risk of disruptions due to envi-
ronmental uncertainty. A firm can improve its respon-
4.2.5. Sensitivity analysis of the fixed loss cost siveness to supply chain disruptions by incorporating
Finally, we examine how the fixed loss cost of each item redundant and/or flexible resources into its supply chain
affects the final selection and allocation of flexible pri- design. A firm with redundant and/or flexible resources
mary and backup suppliers. For the purpose of analysis can respond to disruptions by redirecting the flow of
of this section, we change the value of fixed loss cost per materials whenever the normal flow is interrupted. Even
item from $0 to a large number when losing one item can though the awareness for adding such resources into sup-
cause a huge loss of money for the firm. The results are ply chains has increased, recent surveys from practition-
shown in Figure 11 in four sets. ers indicate that companies are still struggling with their
The results shown in Figure 11 indicate that when SCRM plans and whether the value of these resources jus-
there is no charge for lost sales, the three least expensive tifies the investment required (Sáenz and Revilla 2014).
suppliers are selected with no flexibility or redundancy. Thus, the assessment of the value of SCRM practices for
When the fixed loss cost is a small number, three flexi- improving firm performance and the associated costs and
ble suppliers and one backup supplier are selected. When benefits needs to be further examined (e.g. Kamalahmadi
fixed loss cost grows, another backup supplier is added. and Parast 2016b; Tang and Musa 2011).
However, when losing a single sale is seen as a disaster We responded to this need by studying the relative
for the firm, the firm should use all of its capabilities to value of redesigning the supply chain of a firm to incor-
prevent the occurrence of lost sales. In this case, four flex- porate redundant resources or flexible resources. We pre-
ible and two backup suppliers are selected to minimise sented a two-stage mixed-integer programming model
the cost of lost sales and maximize the service level. for several purposes: 1) to provide insights into the ways
a firm’s supplier selection and demand allocation are
affected by the selection and development of capabili-
5. Discussion ties based on volume flexibility or backup suppliers; 2)
This study provides several contributions to the the- to investigate trade-offs among supply chain resilience,
ory and practice of supply chain risk management and risk, and cost to determine how they interact; and 3)
resiliency that are discussed in the following sections. to develop a more realistic supply chain model where
each primary supplier and backup supplier is exposed to
both an individual risk of an internal disruption and a
risk of an environmental disruption that could make all
Table 12. Changes in Backup Suppliers Based on Fixed Backup
suppliers in the same region simultaneously unavailable.
Contracting Cost.
Our model provides important managerial implica-
Fixed backup contracting cost Number of backup suppliers tions for supply chain managers. First, we quantified the
$1000 3 values of the volume-flexibility practice and the backup-
$5000 2
$10000 1 suppliers practice, showing that both practices are valu-
$15000 0 able in terms of cost reduction and improvement in
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 21

Figure 11. Optimal Allocations of Primary Suppliers (black) and Backup Suppliers (gray) with Different Fixed Loss Cost.

Figure 12. Optimal Allocations of Primary Suppliers (black) and Backup Suppliers (gray) with One Supplier with High/Low Capacity.

service level. These findings confirm the results from


Shekarian, Nooraie, and Parast (2020) and Ivanov and
Dolgui (2020), which showed that flexibility and redun-
dancy make supply chain networks less sensitive to
external uncertainties. However, we found that adding
redundancy to the supply chain by designing a backup
supply base provides a better response to disruptions
compared to adding volume flexibility, from the perspec-
tive of both cost and service level. We regard volume
flexibility as an alternative plan for firms that are more
cost-driven and may not see an investment in resources
that are used only in disruptions as a viable option. We Figure 13. Relative Value of Different Practices in Terms of Cost
summarise these results in Figure 13. and Risk Preferences of a Firm.
Second, although we found better results from redun-
dancy compared to flexibility, our results also determined environment (e.g. some regions of East Asia where many
that the relative value of redundancy over flexibility is computer and auto-parts suppliers are located). Although
reduced with increases in the flexibility of suppliers, the some studies such as Ivanov and Dolgui (2020) and
reliability of suppliers, the reliability of the environment, Ivanov (2020c) showed that in the case of epidemic dis-
and the cost of lost sales. From a practical standpoint, ruptions, proactive measures and backup suppliers can
volume flexibility may not be an option in some situa- be less efficient due to quarantining entire regions and
tions (e.g. chemical processing or nuclear power plants) even continents, our results suggest that in the case of
where changing the capacity in a short time is expensive less severe regional disruptions, a firm can minimise the
and difficult, if not impossible. On the other hand, pre- impact of environmental disruptions by selecting sup-
disruption investment in backup resources that are only pliers from different regions. Furthermore, a balance of
used during disruptions is well justified for a risk-averse costs and risks (reliabilities) is required in both the pri-
firm that sources from suppliers operating in a volatile mary and backup supply bases. We showed that if a firm
22 M. KAMALAHMADI ET AL.

does not want to invest in SCRM practices, it should no emergency capacity is very likely to be removed from
select suppliers based on their reliabilities, which con- the supply base. This result confirms the findings from
firms the results by Sawik (2017, 2019), and Sadghiani, Shekarian, Nooraie, and Parast (2020). As the flexibil-
Torabi, and Sahebjamnia (2015). However, by developing ity of the supplier increases, the supplier is added to the
a backup supply base, a firm can reduce the costs by shift- supply base. After further increases in flexibility, at a cer-
ing some demand to inexpensive, unreliable suppliers, as tain point, the supplier’s demand allocation decreases,
long as suppliers who are more reliable are selected in the with the excess capacity reserved to satisfy demand allo-
backup supply base. For the case with volume flexibility, cation during disruptions. In contrast, a supplier with
we showed that an even distribution of demand between large emergency capacity is selected as a backup supplier.
expensive/inexpensive and reliable/unreliable suppliers This suggests that accurate estimates of these parame-
provides the best response. ters play a significant role in a firm’s success to cope with
Third, to further analyze the value of flexible and disruptions.
redundant resources, we then argued that these prac- Finally, the results of this study can also be extended
tices should not be considered in isolation; we pro- to the supply chain disruptions due to COVID-19 pan-
posed a hybrid model that combines flexible suppliers demic. Ivanov and Das (2020) concluded that while
and backup suppliers. Our results showed that a hybrid the full impact of COVID-19 on supply chains remains
flexible-backup-supplier plan outperforms other prac- unknown, the benefits associated with creating flexible,
tices by providing lower pre-disruption investment costs redundant, and real-time supply chains to dynamically
and lower expected total costs, while providing an iden- reallocate demand and supply should be highlighted.
tical or superior service level relative to each of the prac- In addition, Ivanov (2020b) showed that supply chain
tices alone. To the best of our knowledge, our study is managers should use available technologies and rele-
the first to analyze and quantify a hybrid plan utilising a vant techniques such as simulation approach (which is
combination of flexibility practices and redundancy prac- also used in this study) in order to predict the impact
tices. A hybrid plan that adds some level of flexibility to a of COVID-19 on the firm’s supply chains. Moreover,
backup plan may be a more suitable option, especially for Ivanov and Dolgui (2020) emphasized the role of inter-
managers who, as Chopra and Sodhi (2014) mentioned, twined supply networks as resilient networks in mitigat-
find it hard to invest in redundant resources that may ing the effect of COVID-19 disruption on supply chains.
have no use in normal conditions. In our study, the backup suppliers can facilitate build-
Furthermore, the analysis of VSS metric results in ing an intertwined network among the suppliers. Lastly,
this study showed that the efforts to solve the stochastic Queiroz et al. (2020) concluded that in assessing the effect
programming model are of great importance and could of pandemic outbreaks such as COVID-19 on supply
significantly decrease the company’s expected total cost. chains, mathematical optimisation theories and simula-
In addition, the analysis of EVPI metric results demon- tions techniques can be used to develop robust resilience
strated that due to uncertainty in demand, the company plans. Therefore, the approaches that are utilised in this
faces higher costs, and it is recommended to consider study are also appropriate in responding to COVID-19
accurate forecasting methods and better information- disruption.
sharing mechanisms. Our findings challenge some of the
popular management thinking and practices such as JIT
and lean in their ability to address a firm’s capabilities
5.2. Limitations and future research
to mitigate disruptions (Birkie 2016; Davis, Brown, and
Feander 2015; Maslaric et al. 2013). We showed that even In this study, we considered disruption-mitigation prac-
though more entities were added to the supply base in tices based on volume flexibility, backup suppliers, or
each setting (flexible suppliers and backup suppliers), the both. We examined their impact on firm performance for
overall impact on supply chain cost was reduced when a single-period product with a supply chain exposed to
supply disruptions were incorporated into the model. two types of disruption: individual suppliers and regions.
Furthermore, we showed that there is not one optimal We showed adding flexibility to the backup practice can
solution for all firms in terms of supplier selection and result in more benefits, but we were unable to capture
demand allocation. Our numerical analyses and sensitiv- the optimal combination of the two practices. Finding
ity analyses highlight the need for managers to carefully the optimal combination requires a rigorous analytical
quantify the value of parameters in order to get the best model, which can be a goal for future research. Another
out of their SCRM strategy. We showed that the avail- direction for future studies is to examine the cost/benefits
ability of an extremely reliable supplier can decrease the of pursuing SCRM practices on a multi-horizon time
size of the supply base. A supplier with no flexibility and scale, to see how the decisions can change when there is
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 23

a possibility of changes in the environment or suppliers’ Disclosure statement


reliabilities over time. No potential conflict of interest was reported by the author(s).
With respect to our data, we used the data used by
Ruiz-Torres, Mahmoodi, and Zeng (2013), and Kamalah-
madi and Parast (2016a). This data is based on the obser- Funding
vation of a manufacturer of appliances. We acknowledge This work was supported by National Science Foundation
that a more detailed data set could have made our results (NSF) .
stronger. Finally, in this study, firm performance was
assessed using three performance measures: expected
total costs, expected lost sales, and expected service level. Notes on contributors
Other performance measures could be used to investi- Masoud Kamalahmadi is an assistant professor in the Depart-
gate this problem from a different perspective. It would ment of Management Science at University of Miami’s Herbert
be interesting to develop a profit-maximization model to Business School. He received his Ph.D. in Operations Manage-
ment and Decision Sciences from Indiana University’s Kelley
capture the impact of disruptions on firm profit, taking
School of Business. He has done extensive research in the field
into account both the revenue and cost dimensions. of supply chain risk management. His other research inter-
ests are healthcare operations management and behavioural
operations management.

6. Conclusion Mansoor Shekarian is a Ph.D. student in Operations and Busi-


ness Analytics Department at the University of Oregon. He
This study examined an integrated approach to sup- received his B.S. in Industrial Engineering from Sharif Uni-
plier selection, demand allocation, and the development versity of Technology and his M.S. in Technology Manage-
of capabilities based on redundancy and flexibility, to ment from North Carolina A&T State University. His research
increase the responsiveness of a supply chain that is interest includes service operations management, supply chain
management, and applied econometrics. He is a member of
exposed to supply risk and environmental risk. The the Institute for Operations Research and the Management
proposed model enhances our understanding of sup- Sciences (INFORMS), the Production and Operations Man-
ply chain risk management, supply chain responsive- agement Society (POMS), and the Manufacturing and Service
ness, and the trade-off between redundancy and flexi- Operations Management Society (M&SOM).
bility capabilities. Using a two-stage mixed-integer pro-
Mahour Mellat Parast is an Eminent Scholar at the Ira
gramming model, the importance of considering con- A. Fulton Schools of Engineering at Arizona State Univer-
tingency planning in order to mitigate the negative sity. His current research is focused on supply chain risk
impact of disruptions was discussed. We proposed a and resilience management, and process and product inno-
hybrid practice combining capabilities from the flexible- vation. His scholarly works have appeared in several peer-
suppliers practice and the backup-suppliers practice, and reviewed journals such as Journal of Operations Management,
Decision Sciences Journal, International Journal of Produc-
we showed that the hybrid practice is more effective tion Research, International Journal of Production Economics,
than either of those practices alone. We also empha- International Journal of Logistics Management, International
sized the effects of several factors on management deci- Journal of Logistics Research and Applications, Supply Chain
sions regarding selection of flexible and backup suppli- Management: An International Journal, and Production Plan-
ers: the reliability of regions and suppliers, the flexibil- ning & Control. He is a member of the Decision Sciences
ity and emergency capacity of suppliers, the associated Institute (DSI), the Production and Operations Management
Society (POMS), and the Academy of Management (AOM).
costs of adding resilience capabilities, and the alloca- Mahour received his Ph.D. in Industrial & Management Sys-
tion of normal and emergency items. The findings can tems Engineering from University of Nebraska-Lincoln, M.S.
help scholars and practitioners design more resilient sup- from Iran University of Science and Technology (Indus-
ply chains that are more responsive to supply chain trial Engineering), and B.S. from Sharif University (Industrial
disruptions. Engineering).

Note
Acknowledgments 1. From the perspective of resource dependency theory, these
This research is based on work supported by the National two strategies are also referred to as Buffering (keeping
Science Foundation (NSF) under grant number 1238878. additional inventory) and Bridging (establishing strong
Any opinions, findings, and conclusions or recommendations linkages with the exchange partners); buffering is more
expressed in this material are those of the author(s) and externally-oriented, while bridging is more internally-
do not necessarily reflect the views of the National Science driven (Oliver 1991; Carroll, 1993; Gresov and Drazin
Foundation. 1997; Bode et al. 2011; Talluri et al. 2013). Because both
24 M. KAMALAHMADI ET AL.

of our mitigation strategies are externally oriented, Redun- Chopra, S., and M. Sodhi. 2004. “Managing Risk to Avoid
dancy and Flexibility appear to be more suitable terms in Supply-Chain Breakdown.” MIT Sloan Management Review
this context. 46 (1): 53–61.
Chopra, S., and M. Sodhi. 2014. “Reducing the Risk of Supply
Chain Disruptions.” MIT Sloan Management Review 55 (3):
References 72–80.
Christopher, M., and M. Holweg. 2011. ““Supply Chain 2.0”:
Azadeh, A., N. Atrchin, V. Salehi, and H. Shojaei. 2014.
Managing Supply Chains in the era of Turbulence.” Interna-
“Modelling and Improvement of Supply Chain with Impre-
tional Journal of Physical Distribution & Logistics Manage-
cise Transportation Delays and Resilience Factors.” Interna-
ment 41 (1): 63–82.
tional Journal of Logistics Research and Applications 17 (4):
Christopher, M., and H. Peck. 2004. “Building the Resilient
269–282.
Supply Chain.” The International Journal of Logistics Man-
Baghalian, A., S. Rezapour, and R. Z. Farahani. 2013. “Robust
agement 15 (2): 1–14.
Supply Chain Network Design with Service Level Against
Ciccullo, F., M. Pero, M. Caridi, J. Gosling, and L. Purvis.
Disruptions and Demand Uncertainties: A Real-Life Case.”
2018. “Integrating the Environmental and Social Sustainabil-
European Journal of Operational Research 227 (1): 199–215.
ity Pillars Into the Lean and Agile Supply Chain Manage-
Behzadi, G., M. J. O’Sullivan, and T. L. Olsen. 2020. “On Metrics
ment Paradigms: A Literature Review and Future Research
for Supply Chain Resilience.” European Journal of Opera-
Directions.” Journal of Cleaner Production 172: 2336–2350.
tional Research 287 (1): 145–158.
Craighead, C. W., J. Blackhurst, M. J. Rungtusanatham, and R.
Berger, P. D., A. Gerstenfeld, and A. Z. Zeng. 2004. “How Many
B. Handfield. 2007. “The Severity of Supply Chain Disrup-
Suppliers are Best? A Decision-Analysis Approach.” Omega
tions: Design Characteristics and Mitigation Capabilities.”
32 (1): 9–15.
Decision Sciences 38 (1): 131–156.
Birge, J. R., & Louveaux, F. (2011). Introduction to Stochastic
Datta, P. P., M. Christopher, and P. Allen. 2007. “Agent-
Programming. Springer Science & Business Media.
based Modelling of Complex Production/Distribution Sys-
Birkie, S. E. 2016. “Operational Resilience and Lean: in Search
tems to Improve Resilience.” International Journal of Logis-
of Synergies and Trade-Offs.” Journal of Manufacturing Tech-
tics Research and Applications 10 (3): 187–203.
nology Management 27 (2): 185–207.
Davis, C., W. Brown, and L. Feander. 2015. Making sure lean
Blackhurst, J., K. S. Dunn, and C. W. Craighead. 2011.
stocks are properly nourished, AT Kearney, Available at:
“An Empirically Derived Framework of Global Supply
http://www.atkearney.co.uk/paper/-/asset_publisher/dVxv4
Resiliency.” Journal of Business Logistics 32 (4): 374–391.
Hz2h8bS/content/making-sure-lean-stocks-are-properly-no
Bode, C., S. M. Wagner, K. J. Petersen, and L. M. Ellram.
urished/10192?_101_INSTANCE_dVxv4Hz2h8bS_redirect
2011. “Understanding Responses to Supply Chain Disrup-
= %2Fissue-papers-perspectives.
tions: Insights from Information Processing and Resource
De Giovanni, D., and I. Massabò. 2018. “Capacity Investment
Dependence Perspectives.” Academy of Management Journal
Under Uncertainty: The Effect of Volume Flexibility.” Inter-
54 (4): 833–856.
national Journal of Production Economics 198: 165–176.
Briano, E., C. Caballini, P. Giribone, and R. Revetria. 2010.
Dolgui, A., D. Ivanov, and M. Rozhkov. 2020. “Does the Ripple
“Objectives and Perspectives for Improving Resiliency in
Effect Influence the Bullwhip Effect? An Integrated Analy-
Supply Chains.” WSEAS Transactions on Systems 9 (2):
sis of Structural and Operational Dynamics in the Supply
136–145.
Chain.” International Journal of Production Research 58 (5):
Brusset, X., and C. Teller. 2017. “Supply Chain Capabilities,
1285–1301.
Risks, and Resilience.” International Journal of Production
Dolgui, A., D. Ivanov, and B. Sokolov. 2018. “Ripple Effect in
Economics 184: 59–68.
the Supply Chain: an Analysis and Recent Literature.” Inter-
Carroll, G. R. 1993. “A Sociological View on why Firms Differ.”
national Journal of Production Research 56 (1-2): 414–430.
Strategic Management Journal 14 (4): 237–249.
Dubey, R., A. Gunasekaran, and S. J. Childe. 2018. “Big Data
Carvalho, H., S. G. Azevedo, and V. Cruz-Machado. 2012b.
Analytics Capability in Supply Chain Agility: The Mod-
“Agile and Resilient Approaches to Supply Chain Man-
erating Effect of Organizational Flexibility.” Management
agement: Influence on Performance and Competitiveness.”
Decision 37 (1): 10–36.
Logistics Research 4 (1-2): 49–62.
Duong, L. N. K., and J. Chong. 2020. “Supply Chain Col-
Carvalho, H., A. P. Barroso, V. H. Machado, S. Azevedo,
laboration in the Presence of Disruptions: a Literature
and V. Cruz-Machado. 2012a. “Supply Chain Redesign for
Review.” International Journal of Production Research 58
Resilience Using Simulation.” Computers & Industrial Engi-
(11): 3488–3507.
neering 62 (1): 329–341.
Elzarka, S. M. 2013. “Supply Chain Risk Management: the
Carvalho, H., S. Duarte, and V. Cruz-Machado. 2011. “Lean,
Lessons Learned from the Egyptian Revolution 2011.” Inter-
Agile, Resilient and Green: Divergencies and Synergies.”
national Journal of Logistics Research and Applications 16 (6):
International Journal of Lean Six Sigma 2 (2): 151–179.
482–492.
Chen, J., H. Xu, and P. Zhou. 2020. “Delegation vs. Direct Sourc-
Esmaeili-Najafabadi, E., M. S. F. Nezhad, H. Pourmohammadi,
ing Revisited: Contract Types Under Correlated Supply Risks
M. Honarvar, and M. A. Vahdatzad. 2019. “A Joint Sup-
and Asymmetric Cost Information.” International Journal of
plier Selection and Order Allocation Model with Disruption
Production Research 58 (22): 7005–7022.
Risks in Centralized Supply Chain.” Computers & Industrial
Chen, J., X. Zhao, and Y. Zhou. 2012. “A Periodic-Review
Engineering 127: 734–748.
Inventory System with a Capacitated Backup Supplier for
Falasca, M., Zobel, C. W., & Cook, D. (2008, May). A deci-
Mitigating Supply Disruptions.” European Journal of Oper-
sion support framework to assess supply chain resilience.
ational Research 219 (2): 312–323.
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 25

In Proceedings of the 5th International ISCRAM Conference Risks and Resilience.” International Journal of Production
(pp. 596-605). Research 57 (15-16): 5119–5136.
Fattahi, M., K. Govindan, and R. Maihami. 2020. “Stochastic Ivanov, D., and A. Dolgui. 2020. “Viability of Intertwined
Optimization of Disruption-Driven Supply Chain Network Supply Networks: Extending the Supply Chain Resilience
Design with a new Resilience Metric.” International Journal Angles Towards Survivability. A Position Paper Motivated by
of Production Economics 230: 107755 (in press). COVID-19 Outbreak.” International Journal of Production
Fiksel, J., M. Ployviou, K. Croxton, and T. Pettit. 2015. “From Research 58 (10): 2904–2915.
Risk to Resilience: Learning to Deal with Disruption.” MIT Ivanov, D., R. Hartl, A. Dolgui, A. Pavlov, and B. Sokolov.
Sloan Management Review 56 (2): 79–86. 2015. “Integration of Aggregate Distribution and Dynamic
Ghadge, A., S. Dani, M. Chester, and R. Kalawsky. 2013. “A Transportation Planning in a Supply Chain with Capacity
Systems Approach for Modelling Supply Chain Risks.” Sup- Disruptions and the Ripple Effect Consideration.” Inter-
ply Chain Management: An International Journal 18 (5): national Journal of Production Research 53 (23): 6963–
523–538. 6979.
Ghadge, A., S. Dani, and R. Kalawsky. 2012. “Supply Chain Risk Ivanov, D., A. Pavlov, A. Dolgui, D. Pavlov, and B. Sokolov.
Management: Present and Future Scope.” The International 2016. “Disruption-driven Supply Chain (re)-Planning and
Journal of Logistics Management 23 (3): 313–339. Performance Impact Assessment with Consideration of pro-
Gresov, C., and R. Drazin. 1997. “Equifinality: Functional Active and Recovery Policies.” Transportation Research Part
Equivalence in Organization Design.” Academy of Manage- E: Logistics and Transportation Review 90: 7–24.
ment Review 22 (2): 403–428. Jain, V., S. Kumar, U. Soni, and C. Chandra. 2017. “Supply
Gunasekaran, A., B. K. Rai, and M. Griffin. 2011. “Resilience Chain Resilience: Model Development and Empirical Anal-
and Competitiveness of Small and Medium Size Enterprises: ysis.” International Journal of Production Research 55 (22):
an Empirical Research.” International Journal of Production 6779–6800.
Research 49 (18): 5489–5509. Johnson, N., D. Elliott, and P. Drake. 2013. “Exploring the Role
Gunessee, S., N. Subramanian, and K. Ning. 2018. “Natural Dis- of Social Capital in Facilitating Supply Chain Resilience.”
asters, PC Supply Chain and Corporate Performance.” Inter- Supply Chain Management: An International Journal 18 (3):
national Journal of Operations & Production Management 38 324–336.
(1): 129–148. Jüttner, U., and S. Maklan. 2011. “Supply Chain Resilience
Hosseini, S., D. Ivanov, and A. Dolgui. 2019a. “Ripple Effect in the Global Financial Crisis: an Empirical Study.” Sup-
Modelling of Supplier Disruption: Integrated Markov Chain ply Chain Management: An International Journal 16 (4):
and Dynamic Bayesian Network Approach.” International 246–259.
Journal of Production Research 58 (11): 3284–3303. Jüttner, U., H. Peck, and M. Christopher. 2003. “Supply
Hosseini, S., N. Morshedlou, D. Ivanov, M. D. Sarder, K. Barker, Chain Risk Management: Outlining an Agenda for Future
and A. Al Khaled. 2019b. “Resilient Supplier Selection and Research.” International Journal of Logistics: Research and
Optimal Order Allocation Under Disruption Risks.” Inter- Applications 6 (4): 197–210.
national Journal of Production Economics 213: 124–137. Kamalahmadi, M., and M. M. Parast. 2016a. “Developing a
Hou, J., A. Z. Zeng, and L. Zhao. 2010. “Coordination with a Resilient Supply Chain Through Supplier Flexibility and
Backup Supplier Through buy-Back Contract Under Supply Reliability Assessment.” International Journal of Production
Disruption.” Transportation Research Part E: Logistics and Research 54 (1): 302–321.
Transportation Review 46 (6): 881–895. Kamalahmadi, M., and M. M. Parast. 2016b. “A Review of
Inman, R. R., and D. E. Blumenfeld. 2014. “Product Com- the Literature on the Principles of Enterprise and Supply
plexity and Supply Chain Design.” International Journal of Chain Resilience: Major Findings and Directions for Future
Production Research 52 (7): 1956–1969. Research.” International Journal of Production Economics
Ivanov, D. 2020a. “‘A Blessing in Disguise ‘or ‘as if it Wasn’t 171: 116–133.
Hard Enough Already’: Reciprocal and Aggravate Vulnera- Kamalahmadi, M., and M. M. Parast. 2017. “An Assessment
bilities in the Supply Chain.” International Journal of Produc- of Supply Chain Disruption Mitigation Strategies.” Interna-
tion Research 58 (11): 3252–3262. tional Journal of Production Economics 184: 210–230.
Ivanov, D. 2020b. “Predicting the Impacts of Epidemic Out- Keyvanloo, M., A. M. Kimiagari, and A. Esfahanipour. 2015. “A
breaks on Global Supply Chains: A Simulation-Based Analy- Hybrid Approach to Select the Best Sourcing Policy Using
sis on the Coronavirus Outbreak (COVID-19/SARS-CoV-2) Stochastic Programming.” Journal of Manufacturing Systems
Case.” Transportation Research Part E: Logistics and Trans- 36: 115–127.
portation Review 136: 101922. Khalilabadi, S. M. G., S. H. Zegordi, and E. Nikbakhsh. 2020. “A
Ivanov, D. 2020b. “Viable Supply Chain Model: Integrating Multi-Stage Stochastic Programming Approach for Supply
Agility, Resilience and Sustainability Perspectives—Lessons Chain Risk Mitigation via Product Substitution.” Computers
from and Thinking Beyond the COVID-19 Pandemic.” & Industrial Engineering 149: 106786 (in press).
Annals of Operations Research 1. Kleindorfer, P. R., and G. H. Saad. 2005. “Managing Disrup-
Ivanov, D., and A. Das. 2020c. “Coronavirus (COVID- tion Risks in Supply Chains.” Production and Operations
19/SARS-CoV-2) and Supply Chain Resilience: A Research Management 14 (1): 53–68.
Note.” International Journal of Integrated Supply Manage- Klibi, W., and A. Martel. 2012. “Modeling Approaches for
ment 13 (1): 90–102. the Design of Resilient Supply Networks Under Disrup-
Ivanov, D., and A. Dolgui. 2019. “Low-Certainty-Need (LCN) tions.” International Journal of Production Economics 135 (2):
Supply Chains: a new Perspective in Managing Disruption 882–898.
26 M. KAMALAHMADI ET AL.

Knemeyer, A. M., W. Zinn, and C. Eroglu. 2009. “Proactive Pettit, T. J., K. L. Croxton, and J. Fiksel. 2013. “Ensuring Sup-
Planning for Catastrophic Events in Supply Chains.” Journal ply Chain Resilience: Development and Implementation of
of Operations Management 27 (2): 141–153. an Assessment Tool.” Journal of Business Logistics 34 (1):
Lee, H. L. 2004. “The Triple-A Supply Chain.” Harvard Business 46–76.
Review 82 (10): 102–113. Pettit, T. J., J. Fiksel, and K. L. Croxton. 2010. “Ensuring Sup-
Manuj, I., and J. T. Mentzer. 2008. “Global Supply Chain Risk ply Chain Resilience: Development of a Conceptual Frame-
Management Strategies.” International Journal of Physical work.” Journal of Business Logistics 31 (1): 1–21.
Distribution & Logistics Management 38 (3): 192–223. Ponis, S. T., and E. Koronis. 2012. “Supply Chain Resilience?
Maslaric, M., T. Backalic, S. Nikolicic, and D. Mircetic. Definition of Concept and its Formative Elements.” The
2013. “Assessing the Trade-off Between Lean and Resilience Journal of Applied Business Research 28 (5): 921–935.
Through Supply Chain Risk Management.” International Ponomarov, S. Y., and M. C. Holcomb. 2009. “Understanding
Journal of Industrial Engineering and Management 4 (4): the Concept of Supply Chain Resilience.” The International
229–236. Journal of Logistics Management 20 (1): 124–143.
Meena, P. L., S. P. Sarmah, and A. Sarkar. 2011. “Sourcing Qin, X., X. Liu, and L. Tang. 2013. “A two-Stage Stochastic
Decisions Under Risks of Catastrophic Event Disruptions.” Mixed-Integer Program for the Capacitated Logistics Fortifi-
Transportation Research Part E: Logistics and Transportation cation Planning Under Accidental Disruptions.” Computers
Review 47 (6): 1058–1074. & Industrial Engineering 65 (4): 614–623.
Melnyk, S. A., D. J. Closs, S. E. Griffis, C. W. Zobel, Queiroz, M. M., D. Ivanov, A. Dolgui, and S. F. Wamba. 2020.
and J. R. Macdonald. 2014. “Understanding Supply Chain “Impacts of Epidemic Outbreaks on Supply Chains: Map-
Resilience.” Supply Chain Management Review 18 (1): 34–41. ping a Research Agenda Amid the COVID-19 Pandemic
Moritz, S., and R. Pibernik. 2008. The optimal number of Through a Structured Literature Review.” Annals of Opera-
suppliers in the presence of volume discounts and differ- tions Research, 1–38.
ent compensation potentials-an analytical and numerical Ratick, S., B. Meacham, and Y. Aoyama. 2008. “Locating Backup
analysis. Facilities to Enhance Supply Chain Disaster Resilience.”
Olivares Aguila, J., and W. ElMaraghy. 2018. “Structural Com- Growth and Change 39 (4): 642–666.
plexity and Robustness of Supply Chain Networks Based on Rawls, C. G., and M. A. Turnquist. 2010. “Pre-positioning of
Product Architecture.” International Journal of Production Emergency Supplies for Disaster Response.” Transportation
Research 56 (20): 6701–6718. Research Part B: Methodological 44 (4): 521–534.
Oliver, C. 1991. “Strategic Responses to Institutional Pro- Rezapour, S., R. Srinivasan, J. Tew, J. K. Allen, and F.
cesses.” Academy of Management Review 16 (1): 145–179. Mistree. 2018. “Correlation Between Strategic and Oper-
Pal, R., H. Torstensson, and H. Mattila. 2014. “Antecedents of ational Risk Mitigation Strategies in Supply Networks.”
Organizational Resilience in Economic Crises—an Empiri- International Journal of Production Economics 201: 225–
cal Study of Swedish Textile and Clothing SMEs.” Interna- 248.
tional Journal of Production Economics 147: 410–428. Rice, J. B., and F. Caniato. 2003. “Building a Secure and Resilient
Parast, M. M., S. Sabahi, and M. Kamalahmadi. 2019. “The Supply Network.” Supply Chain Management Review 7 (5):
Relationship between Firm Resilience to Supply Chain Dis- 22–30.
ruptions and Firm Innovation.” In Revisiting Supply Chain Ruiz-Benítez, R., C. López, and J. C. Real. 2018. “The Lean and
Risk. Springer Series in Supply Chain Management, edited Resilient Management of the Supply Chain and its Impact on
by G. Zsidisin and M. Henke, vol. 7, pp. 279–298. Cham: Performance.” International Journal of Production Economics
Springer. 203: 190–202.
Parast, M. M., and M. Shekarian. 2019. “The Impact of Supply Ruiz-Torres, A. J., and F. Mahmoodi. 2006. “A Supplier Alloca-
Chain Disruptions on Organizational Performance: A Lit- tion Model Considering Delivery Failure, Maintenance and
erature Review.” In Revisiting Supply Chain Risk: Springer Supplier Cycle Costs.” International Journal of Production
Series in Supply Chain Management, edited by G. Zsidisin Economics 103 (2): 755–766.
and M. Henke, 367–389. Cham: Springer. Ruiz-Torres, A. J., F. Mahmoodi, and A. Z. Zeng. 2013. “Sup-
Paul, S. K., R. Sarker, D. Essam, and P. T. W. Lee. 2019. plier Selection Model with Contingency Planning for Sup-
“A Mathematical Modelling Approach for Managing Sud- plier Failures.” Computers & Industrial Engineering 66 (2):
den Disturbances in a Three-Tier Manufacturing Sup- 374–382.
ply Chain.” Annals of Operations Research 280 (1-2): Sadghiani, N. S., S. A. Torabi, and N. Sahebjamnia. 2015. “Retail
299–335. Supply Chain Network Design Under Operational and Dis-
Pavlov, A., D. Ivanov, D. Pavlov, and A. Slinko. 2019a. “Opti- ruption Risks.” Transportation Research Part E: Logistics and
mization of Network Redundancy and Contingency Plan- Transportation Review 75: 95–114.
ning in Sustainable and Resilient Supply Chain Resource Sarkar, A., and P. K. Mohapatra. 2009. “Determining the Opti-
Management Under Conditions of Structural Dynamics.” mal Size of Supply Base with the Consideration of Risks
Annals of Operations Research, 1–30. of Supply Disruptions.” International Journal of Production
Pavlov, A., D. Ivanov, F. Werner, A. Dolgui, and B. Sokolov. Economics 119 (1): 122–135.
2019b. “Integrated Detection of Disruption Scenarios, the Sawik, T. 2013. “Selection of Resilient Supply Portfolio Under
Ripple Effect Dispersal and Recovery Paths in Supply Disruption Risks.” Omega 41 (2): 259–269.
Chains.” Annals of Operations Research, 1–23. Sawik, T. 2014. “Optimization of Cost and Service Level in
Peck, H. 2005. “Drivers of Supply Chain Vulnerability: An Inte- the Presence of Supply Chain Disruption Risks: Single vs.
grated Framework.” International Journal of Physical Distri- Multiple Sourcing.” Computers & Operations Research 51:
bution & Logistics Management 35 (4): 210–232. 11–20.
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 27

Sawik, T. 2017. “A Portfolio Approach to Supply Chain Dis- Tang, C. S. 2006b. “Robust Strategies for Mitigating Sup-
ruption Management.” International Journal of Production ply Chain Disruptions.” International Journal of Logistics:
Research 55 (7): 1970–1991. Research and Applications 9 (1): 33–45.
Sawik, T. 2019. “Two-period vs. Multi-Period Model for Sup- Tang, O., and S. N. Musa. 2011. “Identifying Risk Issues and
ply Chain Disruption Management.” International Journal of Research Advancements in Supply Chain Risk Manage-
Production Research 57 (14): 4502–4518. ment.” International Journal of Production Economics 133
Sáenz, M. J., and E. Revilla. 2014. “Creating More Resilient Sup- (1): 25–34.
ply Chains.” MIT Sloan Management Review 55 (4): 22–24. Tang, C., and B. Tomlin. 2008. “The Power of Flexibility for
Sheffi, Y. 2005. “Preparing for the big one [Supply Chain Man- Mitigating Supply Chain Risks.” International Journal of Pro-
agement].” Manufacturing Engineer 84 (5): 12–15. duction Economics 116 (1): 12–27.
Sheffi, Y. 2006. “Resilience Reduces Risk.” Logistics Quarterly Tomlin, B. 2006. “On the Value of Mitigation and Contingency
12 (1): 12–14. Strategies for Managing Supply Chain Disruption Risks.”
Sheffi, Y., and J. B. Rice Jr. 2005. “A Supply Chain View of the Management Science 52 (5): 639–657.
Resilient Enterprise.” MIT Sloan Management Review 47 (1): Torabi, S. A., M. Baghersad, and S. A. Mansouri. 2015.
41–48. “Resilient Supplier Selection and Order Allocation Under
Shekarian, M., S. V. R. Nooraie, and M. M. Parast. 2020. “An Operational and Disruption Risks.” Transportation Research
Examination of the Impact of Flexibility and Agility on Part E: Logistics and Transportation Review 79: 22–48.
Mitigating Supply Chain Disruptions.” International Journal Urciuoli, L., S. Mohanty, J. Hinsta, and E. G. Boekesteijn. 2014.
of Production Economics 220: 107438. “A Multiple Case Study Approach on oil and gas Supply
Shih, W. 2020. “Is It Time to Rethink Globalized Sup- Chains to Europe.” Supply Chain Management: An Interna-
ply Chains?.” MIT Sloan Management Review March 19, tional Journal 19 (1): 46–63.
2020. Valente, C., G. Mitra, M. Sadki, and R. Fourer. 2009. “Extend-
Sodhi, M. S., and S. Lee. 2007. “An Analysis of Sources of ing Algebraic Modelling Languages for Stochastic Program-
Risk in the Consumer Electronics Industry.” Journal of the ming.” INFORMS Journal on Computing 21 (1): 107–122.
Operational Research Society 58 (11): 1430–1439. Wicher, P., and R. Lenort. 2013. The ways of creating resilient
Sodhi, M. S., B. G. Son, and C. S. Tang. 2012. “Researchers’ Per- supply chains. Proceedings of Carpathian Logistics Congress
spectives on Supply Chain Risk Management.” Production 2013, Cracow, Poland.
and Operations Management 21 (1): 1–13. Xu, S., X. Zhang, L. Feng, and W. Yang. 2020. “Disruption Risks
Sreedevi, R., and H. Saranga. 2017. “Uncertainty and Supply in Supply Chain Management: a Literature Review Based on
Chain Risk: The Moderating Role of Supply Chain Flexibil- Bibliometric Analysis.” International Journal of Production
ity in Risk Mitigation.” International Journal of Production Research 58 (11): 3508–3526.
Economics 193: 332–342. Yang, G., and C. Qian. 2008, October. “The Optimal Number of
Talluri, S., T. J. Kull, H. Yildiz, and J. Yoon. 2013. “Assessing the Suppliers Considering the Quantity Discount and Supplier
Efficiency of Risk Mitigation Strategies in Supply Chains.” Failure.” IEEE International Conference on Service Opera-
Journal of Business Logistics 34 (4): 253–269. tions and Logistics, and Informatics 2008 (2): 2327–2329.
Tan, W. J., W. Cai, and A. N. Zhang. 2019. “Structural- Zhu, Q., H. Krikke, and M. C. Caniëls. 2017. “Integrated Supply
aware Simulation Analysis of Supply Chain Resilience.” Chain Risk Management: A Systematic Review.” The Inter-
International Journal of Production Research 58 (17): 5175– national Journal of Logistics Management 28 (4): 1123–1141.
5195. Zsidisin, G. A., and S. M. Wagner. 2010. “Do Perceptions
Tang, C. S. 2006a. “Perspectives in Supply Chain Risk Man- Become Reality? The Moderating Role of Supply Chain
agement.” International Journal of Production Economics 103 Resiliency on Disruption Occurrence.” Journal of Business
(2): 451–488. Logistics 31 (2): 1–20.
28 M. KAMALAHMADI ET AL.

Appendix
In the following tables, ETC refers to expected total cost, SL refers to service level, and ELS refers to expected lost sales.

Table A1. Detailed Results in Figure 5.


No Flexibility/Backup Only Flexibility Only Backup
Trial ETC SL ELS ETC SL ELS ETC SL ELS
1 395,419.20 0.95 396.80 329,664.23 0.95 229.33 286,762.13 0.95 81.21
2 313,684.80 0.97 238.40 297,940.01 0.97 178.85 298,320.11 0.97 108.38
3 379,619.20 0.95 395.20 316,045.96 0.95 174.81 323,296.20 0.95 179.32
4 433,435.20 0.95 396.80 369,976.94 0.95 240.85 310,774.44 0.95 146.32
5 401,179.20 0.94 473.60 385,021.66 0.94 397.94 308,788.33 0.94 138.86
6 458,318.40 0.92 627.20 380,508.57 0.92 381.47 361,215.43 0.92 281.51
7 365,006.40 0.95 396.80 361,231.08 0.95 381.02 311,333.83 0.95 156.46
8 468,916.80 0.94 473.60 370,592.58 0.94 243.74 337,343.02 0.94 221.89
9 367,080.00 0.96 320.00 323,991.61 0.96 204.52 306,311.49 0.96 132.24
10 353,080.00 0.97 240.00 324,179.96 0.97 175.03 295,294.63 0.97 101.50
11 408,422.75 0.95 432.04 408,575.79 0.95 419.17 311,473.75 0.95 150.49
12 326,155.24 0.98 166.64 326,938.27 0.98 106.97 301,159.98 0.98 74.13
13 519,360.31 0.92 632.86 384,220.86 0.92 252.73 363,623.61 0.92 200.06
14 506,070.88 0.92 676.82 408,425.18 0.92 407.78 375,107.58 0.92 309.32
15 398,796.53 0.96 343.71 388,466.46 0.96 270.26 327,021.37 0.96 131.42
16 475,443.37 0.94 491.58 392,583.46 0.94 280.86 380,177.15 0.94 217.10
17 467,197.95 0.94 468.86 391,974.89 0.94 271.71 375,583.49 0.94 201.07
18 430,776.28 0.95 410.35 365,129.84 0.95 195.57 342,164.25 0.95 144.60
19 476,872.73 0.93 536.64 443,758.87 0.93 443.54 337,195.35 0.93 144.21
20 407,602.17 0.95 367.77 354,250.88 0.95 217.17 345,018.03 0.95 173.92
21 369,682.23 0.95 388.93 309,422.56 0.95 143.28 264,837.56 0.95 85.52
22 338,431.33 0.96 284.18 308,030.47 0.96 157.52 265,984.63 0.96 66.57
23 408,119.94 0.95 369.18 393,835.25 0.95 284.78 332,440.11 0.95 142.12
24 248,094.43 0.99 83.36 272,664.23 0.99 59.09 237,196.82 0.99 33.78
25 312,470.52 0.99 86.22 315,748.10 0.99 53.70 303,063.30 0.99 37.98
26 418,633.86 0.93 520.52 420,152.95 0.93 471.18 305,868.54 0.93 194.99
27 495,645.01 0.92 648.49 422,233.12 0.92 373.76 348,220.75 0.92 235.24
28 345,248.25 0.97 260.84 342,097.17 0.97 210.05 288,111.52 0.97 92.08
29 464,137.10 0.92 603.07 411,470.92 0.92 435.26 300,110.60 0.92 140.25
30 287,061.99 0.99 103.15 289,063.78 0.99 66.25 265,221.24 0.99 28.09
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 29

Table A2. Detailed Results in Figure 6.


No Flexibility/Backup Only Flexibility Only Backup Hybrid Backup and Flexibility
Trial ETC SL ELS ETC SL ELS ETC SL ELS ETC SL ELS
1 394,828.80 0.95 395.20 341,009.84 0.94 203.20 284,069.03 0.99 29.63 282,600.26 0.99 31.95
2 388,948.80 0.96 316.80 310,523.16 0.95 169.43 251,548.80 0.99 39.87 183,822.10 0.99 39.66
3 293,132.40 0.99 119.60 287,934.98 0.99 107.52 241,611.99 0.99 15.25 241,611.99 0.99 15.25
4 358,512.00 0.96 317.60 303,734.55 0.96 157.61 273,253.08 0.99 29.48 265,894.20 0.99 31.83
5 368,016.00 0.97 238.40 315,883.81 0.97 141.29 239,732.66 0.99 25.43 239,732.66 0.99 25.43
6 395,419.20 0.95 396.80 316,254.07 0.94 161.30 279,534.46 0.99 22.67 264,842.62 0.99 23.83
7 481,896.00 0.93 550.40 333,367.80 0.95 170.75 284,889.19 0.99 44.02 284,889.19 0.99 44.02
8 410,845.44 0.97 271.04 298,573.54 0.98 65.10 250,377.31 0.99 27.64 230,292.40 0.99 27.90
9 407,716.42 0.96 326.14 328,711.43 0.94 149.32 252,132.86 0.99 40.42 253,937.78 0.99 41.39
10 289,354.56 0.99 87.92 253,386.53 0.99 40.46 226,919.52 0.99 20.94 223,827.98 0.99 21.20
11 423,868.10 0.94 494.24 314,006.15 0.93 142.97 280,472.85 0.99 46.59 269,018.11 0.99 44.36
12 421,478.35 0.94 508.03 406,073.44 0.92 458.17 244,759.37 0.98 46.36 244,759.37 0.98 46.36
13 402,950.95 0.96 355.06 352,435.48 0.94 221.36 254,210.45 0.99 39.05 254,210.45 0.99 39.05
14 481,163.37 0.92 668.90 435,826.23 0.97 523.70 278,952.55 0.98 60.87 278,952.55 0.98 60.87
15 384,395.86 0.96 325.33 309,193.97 0.96 151.19 235,768.69 0.99 49.29 235,768.69 0.99 49.29
16 397,461.90 0.97 233.97 391,513.05 0.99 220.47 268,171.91 0.99 40.47 268,171.91 0.99 40.47
17 370,384.73 0.96 287.15 295,763.23 0.96 98.68 277,891.18 0.99 31.18 278,171.97 0.99 20.20
18 414,977.21 0.94 470.21 305,966.99 0.95 187.22 235,237.26 0.99 39.28 235,237.26 0.99 39.28
19 512,058.11 0.92 653.16 439,002.56 0.90 500.45 270,139.11 0.98 66.37 271,109.19 0.98 71.39
20 475,455.07 0.92 673.27 401,633.97 0.94 448.25 259,976.23 0.99 38.47 256,763.93 0.99 44.56
21 367,475.83 0.98 172.59 311,525.33 0.98 77.20 281,852.41 0.99 28.95 279,479.83 0.99 30.59
22 524,480.30 0.92 646.92 469,633.25 0.96 523.06 228,879.06 0.99 44.41 228,879.06 0.99 44.41
23 519,180.94 0.92 632.37 381,154.69 0.91 316.56 279,003.91 0.99 52.39 274,307.06 0.99 50.01
24 351,802.83 0.97 257.62 299,906.29 0.97 41.94 263,025.30 0.99 36.70 277,079.70 0.99 25.67
25 399,251.03 0.95 365.81 329,593.46 0.95 197.08 239,042.63 0.99 34.78 243,238.33 0.99 32.41
26 296,507.23 0.97 254.69 299,692.96 0.96 228.61 228,691.10 1.00 18.07 230,567.64 0.99 29.25
27 449,922.88 0.93 544.51 412,645.52 0.92 424.37 258,041.39 0.99 55.50 258,041.39 0.99 55.50
28 424,599.34 0.94 496.21 413,803.15 0.94 457.51 290,893.31 0.98 58.77 290,893.31 0.98 58.77
29 426,518.71 0.93 541.72 312,590.41 0.95 185.23 264,071.19 0.99 49.99 266,315.25 0.99 50.54
30 399,477.98 0.94 469.03 260,720.77 0.93 116.91 215,170.48 0.99 44.90 218,378.14 0.99 41.87

You might also like