Professional Documents
Culture Documents
Supermercado 6
Supermercado 6
Volume 00 Number 0
xxxx 2017
Uzma Raja
Culverhouse College of Commerce, The University of Alabama, Tuscaloosa, AL 35487,
e-mail: uraja@cba.ua.edu
ABSTRACT
Retailers face a major operational challenge in fulfilling online orders while managing
their traditional store-based distribution processes. In this context, the following order
fulfillment options available to retailers are considered: store-facing distribution centers
(DCs), dedicated order fulfillment facilities (DTC), retail stores, and direct-fill by ven-
dors. A framework for the online order fulfillment process is developed to evaluate these
options using operational and financial measures collected from a large U.S. retailer.
The study presents managerial insights regarding each fulfillment option and identifies
operational and cost thresholds where a particular fulfillment option would be preferred.
The results show that due to better order fulfillment efficiency, distribution facilities (DC
and DTC) out-perform fulfillment from stores and vendors. However, retailers can lever-
age their network of stores to overcome the shortfall in store-based fulfillment costs by
focusing on the order delivery process. The analysis presented in this article is useful for
practice as it helps retailers identify options that best suit their order fulfillment strategy.
[Submitted: March 4, 2016. Revised: April 18, 2017. Accepted: April 20, 2017.]
INTRODUCTION
For the past decade, online retail sales have increased at a faster rate than sales
through retail stores. In 2015, online sales in the United States were estimated at
$341 billion (an increase of 14.6% over the previous year), whereas store sales
during the same period increased by only 1.4% (U.S. Census Bureau, 2015).
The growth of online sales has attracted many “pure-play” brick-and-mortar store
retailers to add the online channel to their business strategy. These retailers seek to
align their traditional store-based distribution processes with the requirements of
the online channel through coordinating demand management and order fulfillment
activities (Ishfaq, Defee, Gibson, & Raja, 2016). This realignment of supply chain
† Corresponding author.
1
2 Evaluation of Order Fulfillment Options in Retail Supply Chains
process using real-world data collected from a large U.S. retail firm by considering
various configurations within an experimental design study. Results are analyzed
to develop managerial insights regarding operational factors that affect fulfillment
choices in a retail supply chain.
BACKGROUND
In retailing, the physical (stores) and virtual (online and catalog) channels of-
fer distinct mediums through which customers and retailers interact (Metters &
Walton, 2007). In recent years, customers and retailers have started to engage
these channels in an integrated manner (Bell, Gallino, & Moreno, 2014). In this
context, customers use multiple channels interchangeably during the purchase
process (Verhoef et al., 2015). To serve customers, store-based retailers use the
existing distribution resources organized for product flows in the store network to
fill orders from online customers. In this setting, retailers can use inventory stored
across multiple channels, as a channel-agnostic resource, to serve customers in a
seamless manner. In this context, online orders can be processed and shipped from
anywhere in the distribution network through a mix of fulfillment options selected
on the basis of supply chain economics (Hübner, Wollenburg, Holzapfel, & Mena,
2016).
In this article we are focused on the retailer-side logistics of the order ful-
fillment process. Retailers have different fulfillment options that can be employed
to fill customer demand (see Figure 1). These options include the use of existing
DCs to fill online orders, known as integrated fulfillment (Ishfaq et al., 2016). In
this case, retailers develop unified operational processes that combine warehous-
ing activities for store and online channels. The other option available to retailers
is to use dedicated direct-to-customer order fulfillment centers. This dedicated
fulfillment method requires a significant capital investment, process redesign, and
coordinated product flows (Lummus & Vokurka, 2002). A critical threshold of
online sales is needed to justify the high operating costs and inventory risks due to
Note: Vendors (V), distribution centers (DC), direct-to-customer fulfillment centers (DTC), retail stores (R).
4 Evaluation of Order Fulfillment Options in Retail Supply Chains
LITERATURE REVIEW
The strategic planning of online order fulfillment is a relatively new area of research
in retail supply chain management (Agatz et al., 2008). Most of the current litera-
ture in this area has focused on evaluating system performance while considering a
singular fulfillment option. In these papers, evaluation of the order fulfillment pro-
cess is primarily based on transportation and inventory costs. However, in modern
retail supply chains, the multiple distribution nodes (vendors, fulfillment centers,
store-facing DCs and retail stores) involved in the order fulfillment process require
researchers’ attention to explore the underlying complex issues. This gap in litera-
ture has prompted a call for studies that can help retailers realign their operational
capabilities to integrate online and store channels within their existing distribution
networks (Ishfaq et al., 2016; Xia & Zhang, 2010). This section reviews existing
literature on retail distribution networks and highlights the research contributions
of our article.
One of the earliest works related to our research is Alptekinoğlu and Tang
(2005). This article considered a retail distribution system that serves customers
in the online and in-store channels. Their model combines demand from these
Ishfaq and Raja 5
DCs. Furthermore, the order fulfillment choices in these papers are based solely
on transportation costs.
In light of the discussion presented above, our article fills a number of gaps
in literature (see Table 1). First, our study includes all fulfillment options (retail
stores, dedicated fulfillment centers, store-facing DCs, and vendors) simultane-
ously within a retail supply chain setting. Different options are explicitly included
through multiple product flows, and modeled with parametric details without as-
signing a particular customer segment to a specific fulfillment option. Second,
we develop a modeling framework that selects an optimal mix of fulfillment op-
tions based on the relevant logistics costs and order delivery requirements. Third,
our model incorporates the underlying differences among different fulfillment op-
tions by considering: (i) order fulfilments costs that differ across fulfillment nodes
due to labor wages, facility layouts, and variations in pick-pack-ship operations;
(ii) different order delivery and inventory costs that vary across echelons, that is,
stores versus DCs; and (iii) different delivery considerations, such as pick-from-
store and free shipping. These cost differentials play an important role for retailers
in identifying the best mix of fulfillment options. In this regard, our study extends
the order fulfillment literature by presenting a unified framework of the order ful-
fillment process in retail supply chains and an empirically grounded evaluation to
gain managerial insights useful for the practice of retail supply chain management.
In addressing issues related to the order fulfillment process, we have followed
the deterministic demand approach, similar to previous studies in this area. The
use of deterministic demand in order fulfillment research is also consistent with
practice in retail supply chain. Retailers use estimates of mean total demand to
develop strategic level plans for order fulfillment activities. These retailers manage
demand variability at the operational level through a distributed order management
system that handles this issue by leveraging inventory stocked at multiple locations
in the retailer’s distribution network (Chao & Norton, 2016; Gibson, Defee, &
Ishfaq, 2015). Some papers discussed above (see Table 1) have used a stochastic
demand approach. Stochastic demand elements in these papers address supply and
inventory replenishment issues. Our article is focused not on these issues, but rather
on a different area in the retail supply chain, that is, options used to fill and deliver
customers’ online orders. We have discussed such papers in the literature review
to account for elements in these studies that relates to order fulfillment, which we
note is also based on mean demand estimates (i.e., deterministic demand), same
as in our article.
Order fulfillment
options: √
Vendors √ –
√ – – – – – –
Distribution – – – – – –
centers √ √ √ √ √
DTC fulfillment – – –
centers √ √ √ √
Retail stores – – – –
Considerations for: √ √ √ √
Inventory costs √ √ –
√ – √ –
√ –
√
Transportation – –
costs √ √
Order fulfillment – – – – – –
costs √
Cost of goods √ – –
√ –
√ – – –
√ –
√
Deterministic – – –
demand √
Cost differentiation – – – – – – –
by echelon √ √ √ √
Multiple planning – – – –
periods √ √
Inventory limits by – – – – – –
echelon √ √ √
Fixed operational – – – – –
costs
7
8 Evaluation of Order Fulfillment Options in Retail Supply Chains
is, integrated fulfillment. The other option is for the retailer to operate separate
direct-to-customer (DTC) fulfillment centers, represented by set F, which serve
online demand exclusively, that is, dedicated fulfillment. However, in this case
there are additional costs related to operating these facilities. We use parameter
fj ∈F to represent the additional financial burden of operating separate fulfillment
centers specifically for online orders. In the former case (integrated fulfillment),
online orders are filled using existing DCs resulting in no additional fixed costs.
In that sense, we use fj to signify the additional financial burden of operating
separate fulfillment facilities over existing DCs. Another fulfillment option used
by retailers is to fill online orders directly from vendors, represented by set V (i.e.,
vendor fulfillment). This option can be used by the retailer to sell an assortment of
products in the online channel that it does not hold in stock. The fourth option is
to fulfill online orders from retail stores (i.e., store fulfillment), where orders are
shipped directly from retail stores to online customers.
The retailer fills online demand from inventory held at any of the above-
mentioned stocking locations. The fulfillment and order delivery costs vary de-
pending on the fulfillment node used. Based on these considerations, the best order
fulfillment node can be identified. These choices are represented by decision vari-
km
ables Sωt to identify the number of orders of product k in market m using inventory
located at node ω ∈ , during selling season t.
The framework discussed above provides a useful mechanism to evaluate
trade-offs in the order fulfillment process within retail distribution. For example,
the fulfillment of online orders from retail stores may have the lowest last-mile
delivery cost because of the stores’ proximity to local customers. However, if
orders are filled from retail stores, order picking and packing costs would be quite
different compared to filling orders from DCs. Unlike a DC, the display shelves
and inventory placement in a retail store are set according to marketing and sales
considerations. These considerations may limit retailers’ ability to efficiently pick
items for order fulfillment, resulting in higher costs. Furthermore, the high real
estate cost of urban and suburban retail outlets and significant inventory carrying
costs to hold stock for the purpose of filling online orders raise concerns.
A retailer may decide that concerns about high order fulfillment and inventory
holding costs at retail stores warrant the use of existing DCs as order fulfillment
nodes for the online channel. However, note that retailers’ DCs are organized to
efficiently stock or flow-through pallet-size or full-case loads—a configuration
not efficient for filling online orders. This dichotomy raises concerns about the
suitability of using store-facing DCs for filling online orders.
The above-mentioned dilemma can be resolved by using dedicated, direct-
to-customer (DTC) fulfillment centers which are configured exclusively for filling
online orders. However, this option results in higher operating costs, as well as
additional system-level inventory under the portfolio effect theory (Croxton &
Zinn, 2005). Hence, to achieve economies of scale, retailers favor using a few
centrally located, large DTC fulfillment centers. The framework captures these
trade-offs through inventory, transportation, warehousing operations, and delivery
processes. The integrated decision as to which order fulfillment node is most
suitable for a given scenario and how to incorporate inventory to support order
fulfillment choices are included in the modeling framework. We use this framework
10 Evaluation of Order Fulfillment Options in Retail Supply Chains
Model Formulation:
Parameters:
System variables:
Decision variables:
km
Sωt ≥ 0, Orders filled from ω for product k in market m during period t.
k
(Q) min km
Sωt γω + dωk + δωkm
ω∈,t∈T,k∈K,m∈M
+ fj (κj )Yj + cek Xetk + hkω Iωt
k
(1)
j ∈F e∈EO ,t∈T,k∈K ω∈,t∈T,k∈K
subject to: km
Sωt ≤ ptkm (πtkm ) ; ∀k ∈ K, m ∈ M, t ∈ T (2)
ω∈
Ishfaq and Raja 11
Xetk = t +
Sjkm Xetk + Ijkt − Ijk(t−1) ; ∀j ∈ D, k ∈ K, t ∈ T (3)
e∈Hj m∈M e∈Tj
Xetk = t + Ij t − Ij (t−1)
Sjkm k k
; ∀j ∈ R ∪ F, k ∈ K, t ∈ T (4)
e∈Hj m∈M
α k Iωt
k
≤ κω ; ∀ω ∈ \ V, t ∈ T (5)
k∈K
Xjkt ≤ MYj ; ∀j ∈ F (6)
k∈K,t∈T
km
Sωt , Xetk , Iωt
k
≥ 0, Yj ∈ {0, 1}.
The objective of the model is to find the least cost fulfillment option. The first
term of the objective function (1) computes the total cost of order fulfillment, order
delivery, and cost of goods for different fulfillment choices represented by decision
km
variables Sωt . The second term is related to retailer’s decision to use dedicated
DTC fulfillment center(s), which would add operating costs for such facilities.
The third term computes the total outbound transportation costs from DCs to retail
stores. The last term computes the total inventory holding cost of the system. The
order fulfillment choices are made while considering a number of requirements
related to supply, inventory and product sales. In constraints (2), the model ensures
online demand is fully filled from available inventory. The constraints in (3) and
(4) describe inbound/outbound product flows. The availability of products at DC
j is covered by supply from the vendors (Xetk ) and the carryover inventory from
k
the previous period (Ij,t−1 ). The inbound supply variables (Xetk ) are defined over
a subset of edges e ∈ Tj ⊆ EI associated with a fulfillment point j, that consists
of all edges which terminate at j. The outbound flows in (3) include replenishment
shipments from DCs to retail stores and orders sold/shipped to customers. Note
that outbound product flows Xetk are defined over a subset e ∈ Hj ⊆ EO associated
with node j that comprises all edges originating at j. Similar to DCs, product
flow balance requirements are also described for retail stores and DTC fulfillment
facilities through constraints (4). Note that in (4), all outbound product flows
are related to orders delivered to different markets. Constraints in (5) restrict
inventory due to limited storage space. The constraints in (6) ensure that only
operational DTC fulfillment centers are considered for the dedicated fulfillment
option.
The model discussed above incorporates different fulfillment paths to serve
online customers and fill orders using a mix of fulfillment options based on the
relevant logistics costs and order delivery requirements. The model incorporates
the underlying differentiation of fulfillment options by considering: (i) order ful-
filments costs that differ across facilities due to labor wages, facility layouts, and
variations in pick-pack-ship operations; (ii) online orders have different (last-mile)
12 Evaluation of Order Fulfillment Options in Retail Supply Chains
delivery requirements; (iii) inventory costs vary across echelons (i.e., store vs.
DCs); and (iv) there are delivery considerations, such as pick-from-store and free
shipping. Our model captures these issues through model parameters, grounded in
empirical data and the model/analysis setting developed through discussions with
the retail supply chain executives consulted for this research.
The modeling framework and resulting mathematical representation of the
retail distribution system discussed above was used with data collected from a
large store-based retail firm in the United States. These data (described in the next
section) allowed us to create a base-case setting representing the legacy distribution
network used by the retailer. Using these data within a factorial design research
study, we evaluated order fulfillment options through the analysis of different
scenarios and the evaluation of trade-offs among order fulfillment, order delivery,
inventory, and transportation.
RESEARCH STUDY
Research Site
For this research study, we collected data from a large retail firm operating more
than 5,000 stores in the United States. The retailer sells a wide assortment of prod-
ucts, in-store and online, including home and office supplies, health and beauty
products, and pet supplies. The retailer offers national brands from leading man-
ufacturers (such as Proctor and Gamble, Nestle, and Unilever) and private label
selection of comparable quality. The retailer competes with other retail firms by
offering highly competitive prices on private label and national brands sold in its
conveniently located small-box stores. Most of the merchandise flows through the
retailer’s distribution network of 12 DCs, and is delivered to stores by third-party
transportation carriers.
This study focuses on the retailer’s southeastern U.S. market, which com-
prises of seven states: Alabama (AL), Florida (FL), Georgia (GA), Mississippi
(MS), North Carolina (NC), South Carolina (SC), and Tennessee (TN). The south-
eastern region was selected due to its significance for the retailer. The retailer has
its corporate headquarters in this region where it began its operations four decades
ago. The retailer has a deep understanding of the local markets and enjoys brand
recognition among customers in the region. It also has strong relationships with
suppliers and transportation carriers in this area. The southeastern U.S. market
accounts for 51% of the retailer’s total sales. To select appropriate markets in the
region, historical sales data were extracted for all cities served by the retailer. The
markets were rank ordered according to aggregate annual sales and the top 10%
of markets in each of the seven states were selected. This process identified the 35
markets (M) included in the research study.
set according to zip codes of origins (DCs and DTCs) and destinations (cities).
These costs were calculated from zone-based Commercial-plus Parcel shipping
rates charged by the U.S. Postal Service (USPS, 2015) for package deliveries.
Another fulfillment option available to retailers is retail stores, which is
useful for orders to be delivered in locations close to retailers’ stores. However,
this fulfillment option requires pushing inventory to stores where real estate costs
and employee wages are higher compared to the off-site locations of DCs. The store
fulfillment cost was set at $10.00 per order, 30% higher than DC-based fulfillment
(Rueter, 2013).
The retailer may use dedicated direct-to-customer fulfillment centers. This
option allows retailers to leverage economies-of-scale in handling large volumes
of small-sized orders. The cost attributes related to the DTC facilities in AL and
GA are evaluated in the model to decide whether to utilize these facilities, that
is, Yj ∈ {0, 1}. A decision to operate a dedicated DTC facility (Yj = 1) incurs
additional operating costs (fj ) for the retailer. The unit warehousing cost was set
at $10.00 per sq-ft-per year which is the average cost of warehousing operations in
the southeastern U.S. region (Rogers, 2010). To estimate order fulfillment charges
at a dedicated DTC, cost estimates from Fulfilment by Amazon (FBA) services
were used. Amazon operates dedicated e-commerce order fulfillment centers with a
level of efficiency that is an industry benchmark. The fulfillment services provided
by Amazon include order picking, packing and shipping, as well as managing
inventory for their clients. The FBA fulfillment charge for online orders is $5.25
per order, which includes $4.50 for order handling (including standard shipping
for items weighing up to 15 lbs.) and $0.75 for pick and pack charges (Amazon,
2015). A 12-month planning period (|T| = 12) is used in the study.
Research Design
The data collected from the focal firm and secondary sources were evaluated using
a factorial experiment design approach. The key factors and the factor levels used
in the study were selected in consultation with supply chain executives to ensure
validity. This approach allowed us to incorporate sufficient variation in the case
data to generalize the findings of this study. The variations in factor levels are also
useful for developing managerial insights by exploring how changes in parameters’
values affect order fulfillment choices.
The factorial design is based on following key factors: (i) fulfillment cost dωk ,
(ii) delivery cost δωkm , (iii) DTC operational cost fj , and (iv) inventory allocation
κω . Three different levels are used for each factor (see Table 2). The factor levels
represent associated parameter settings, for example, the store order fulfillment
cost factor is set at following three levels: L1 (0.90dR = $9.00), L2 (0.80dR =
$8.00), and L3 (0.70dR = $7.00). Note that (1.00) levels in Table 2 represent data
collected from the retail firm (referred to as base-case). In the factorial design,
factors related to DC fulfillment were fixed at the base-case level. This setting would
help us analyze corresponding changes in the use of order fulfillment options due
to changes in other factors.
The delivery cost factor levels were used as multipliers with the associated
parameter (e.g., level L2 of store order delivery costs scales δR values at 80%
Ishfaq and Raja 15
Fulfillment cost:
Store dR $9.00 $8.00 $7.00
(0.90)dR (0.80)dR (0.70)dR
DTC dF $6.65 $5.95 $5.25
(0.95)dD (0.85)dD (0.75)dD
Vendor dV $5.00 $4.00 $3.00
(1.00)dV (0.80)dV (0.60)dV
Delivery cost:
Store δR (1.00)δR (0.80)δR (0.60)δR
DTC δF (1.00)δF (0.90)δF (0.80)δF
Vendor δV (1.00)δV (0.80)δV (0.60)δV
DTC unit operational cost fF $15.00 $10.00 $5.00
(1.50)fF (1.00)fF (0.50)fF
Retail inventory allocation κR $270 $240 $210
(0.90)κR (0.80)κR (0.70)κR
of base-case data values). Note that factors in the study are set in the context of
general merchandise retail that do not include large and bulky items. Under this
setting, we did not find fulfillment-related costs to vary much across product types.
Thus, factor levels were based on cost averages across sample products. Response
variables in the analysis represent a proportion of online demand filled by each
fulfillment option, that is, there are four response variables for each factor-level
combination presented in Table 2.
Each combination of the factor levels is equivalent to a specific business
scenario based on the underlying model parameter values. With eight factors at
three levels each, a total of 6,561 scenarios were evaluated. The resulting scenario
data and parameter settings were recorded in separate data files. The respective
data files were linked to the model developed earlier and solved using commercial
solver CPELX through AMPL modeling interface. The optimal solution output
yields best order fulfillment options based on the least cost-to-serve approach
currently practiced by store-based retailers (Gibson et al., 2015). The values of the
response variables were recorded from the output of each test scenario.
1 for each data observation. For these type of data, traditional estimation meth-
ods (linear regression, general linear method, etc.) are not appropriate (Ramalho,
Ramalho, & Murteira, 2011).
To analyze data with a fractional response variable, the fractional multino-
mial logit model (FMlogit) technique is used (Papke & Wooldridge, 1996). This
technique combines multivariate fractional logit and multinomial logit such that
the model output represents the expected values of the proportions for different
response variables, all of which sum to 1. In our case, proportions represent use of
different fulfillment options used to fill online orders. FMlogit model measures the
changes in multiple response variables simultaneously as a measure of the effects
of independent variables. Consider a random sample of i = 1, ..., N observations
of orders, each with M outcomes of fulfillment choices. In our study, M = 4, which
corresponds to each of the order fulfillment types. Given sik represents the kth out-
come for observation i, and xi , i = 1, ..., N , is a vector of exogenous covariates.
The FMlogit model assumes that M conditional means have a multinomial logit
functional form in linear indices as:
exp(xβk )
E[sk |x] = Gk (x; β) = M .
m=1 exp(xβm )
We can define a multinomial logit quasi-likelihood function L(β) that uses the
observed shares sik [0, 1] in place of a binary indicator that would otherwise be
used in multinomial functions:
N
M
L(β) = Gm (xi ; β)sim .
i=1 m=1
data and predicted values from the model was recorded (Prob > χ 2 = .000). The
beta coefficients predicted from the estimation of the fractional multinomial logit
model were used to obtain average marginal effects for every independent variable
that are also reported in Table 2. Average marginal effects that are statistically
different from zero at 1% and 5% levels are indicated with two and one asterisks,
respectively.
Each coefficient in Table 4 represents the marginal change in fulfillment
options as a result of a unit change in the independent variables. For example,
the marginal effect for dF under DTC[p] is −0.0915, which suggests that a unit
increase in DTC fulfillment cost is associated with a decrease of 9.15% in the use
of the DTC fulfillment option. Because all fulfillment options must sum to 1, this
shift would mean that some orders will be filled by other fulfillment options. The
results show that a big share of these orders will be filled using DC fulfillment
(coefficient = 0.0852). Similarly, a unit increase in the vendor fulfillment fee dV
would result in a reduction of 6.29% in use of the vendor fulfillment option V[p].
These orders are shifted to DTC for fulfillment (coefficient= +0.0697).
The results show that orders that are filled through DCs and DTCs would
shift between these fulfillment options under a marginal effect due to changes
in relevant (fulfillment and delivery) costs. This effect is seen for dF , fF , and δF
variables. For example, a unit change in fF would shift an equal proportion of
orders from DTC fulfillment (−7.31%) to DC fulfillment (+6.88%). It is also
interesting to note that while both fulfillment costs and order delivery costs show
significant effect on the choice of order fulfillment option, the scale of effect for
fulfillment costs is much higher than that for order delivery costs. For example, a
unit increase in delivery costs δF would reduce DTC fulfillment by 5.69% whereas
unit change in fulfillment costs would reduce DTC use by 9.15%. We explore these
dynamics in more depth later.
The marginal effects for the order delivery costs are more prominent on
vendor fulfillment option than the store fulfillment options. A unit increase in
delivery costs reduces use of store fulfillment by 1.23% and vendor fulfillment by
5.02%. The small scale of the effect on store fulfillment can be explained by the
small use of this fulfillment option for most combinations of factor levels in the
factorial design. We investigate this aspect in more detail later.
The discussion above is based on a full-model analysis using the fractional
multinomial logit model approach. This analysis identified the effects of different
factors on the response variables: DTC[p], DC[p], R[p], and V[p]. In the next step of
the analysis, we focus specifically on two key elements: order fulfillment and order
delivery, in terms of how changes in these factors would affect the retailer’s use
Ind. variable: dF
Coefficient −0.0915** 0.0852** 0.0001** 0.0062**
Std. error 0.0134 0.0124 0.0000 0.0018
z-value −68.16 68.34 17.13 33.50
Ind. variable: dV
Coefficient 0.0697** −0.0068 0.0001 −0.0629**
Std. error 0.0066 0.0061 0 0.0015
z-value 10.44 −1.11 −2.61 −39.7
Ind. variable: dR
Coefficient −0.009 0.0155** −0.0024** −0.0041
Std. error 0.0062 0.0059 0.0001 0.0007
z-value −1.43 2.60 −18.39 −5.38
Ind. variable: fF
Coefficient −0.0731** 0.0688** 0.0001** 0.0043**
Std. error 0.0009 0.0009 0.0000 0.0001
z-value −73.59 74.72 17.29 30.96
Ind. variable: δF
Coefficient −0.0569** 0.0535** 0.0000** 0.0033**
Std. error 0.0856 0.0800 0.0004 0.0115
z-value −66.53 66.92 16.31 29.10
Ind. variable: δV
Coefficient 0.0603** −0.0100** −0.0001 −0.0502**
Std. error 0.0365 0.0326 0.0001 0.0105
z-value 16.52 −3.08 −0.76 −47.52
Ind. variable: δR
Coefficient −0.0459 0.0787** −0.0123** −0.0205**
Std. error 0.0314 0.0297 0.0006 0.0037
z-value −1.46 2.64 −18.17 −5.45
Ind. variable: κR
Coefficient 0.0162 −0.0187 0.0015** 0.0011
Std. error 0.0302 0.0286 0.0001 0.0034
z-value 0.53 −0.65 14.86 0.31
Goodness-of-fit Wald χ 2 = 23,869
statistics
Log pseudolikelihood = −3,064.97
Prob > χ 2 = .000
Note: Statistical significance: **p < .01; *p < .05.
dF ( ddDF ) F D F D F D
1.00 δD 100% − − −
0.95 δD 64% 36% − −
0.90 δD 64% 36% − −
0.85 δD − 100% − −
0.80 δD − 100% − −
Special cases:
δ√F = 0 δD √=0 =0
δR √ =0
δV √
√ 75% 4% − 21%
− − − − − 100% −
and facility operating costs. For example, to compensate for 50% higher DTC
operating costs (fF = $10.00 level vs. fF =$15.00 level), full utilization of DTC
fulfillment (100% orders) would require a DC order fulfillment cost deferential of
more than +30% (i.e., dF ≤$5.00; ddDF ≤ 0.70).
The choice of a fulfillment option not only depends on the efficiency of the
order fulfillment process, as discussed above, but also on the corresponding order
delivery process. We test how the order fulfillment choice may change if the retailer
negotiates a different order delivery rate. For this analysis, results are compiled
for different discount levels in the DC order delivery cost factors (see Table 6).
Each discount level reduces base delivery costs (1.00 δD ) by the corresponding
multiplier, for example, = 0.80δD corresponds to a 20% discount in DC order
delivery costs. The results show that the best fulfillment option changed from DTC
to DC fulfillment for = 0.95. At this point, the proportion of orders filled from
DTC reduced to 64%, with DCs filling the remainder (34%) of online orders. For
≤ 0.85, all orders were filled using DCs.
The delivery process for online orders depends on customers’ preferences.
In some cases, customers would pay delivery charges which would offset retailer’s
order delivery costs. In other cases, the retailer would incur these costs when
customer orders qualify for free shipping. In a related case, customers may opt to
pick their online orders from local stores to avoid the order delivery charges. A
case in which the retailer does not incur order delivery costs is represented by the
setting δF = δD = δR = δV = 0 (see Table 6). In this scenario, a large proportion
(75%) of online orders were filled from DTC fulfillment centers, and about 20%
of the orders were filled by vendors.
Another order delivery option considered in this study is the pick-from-
store arrangement in which last-mile delivery to customers’ homes is not needed.
Customers can pick up their orders from stores without paying any order delivery
charges. In this case the retailer would incur only shipping costs (costs may vary
depending on whether orders are shipped from DCs to stores separately or through
routine store deliveries). If the retailer fills orders from the local store inventory,
Ishfaq and Raja 21
there is no additional shipping/delivery cost for the retailer. The results for this
setting (see last row, Table 6) show a strong use of the store fulfillment option.
However, due to the higher cost structure of urban/suburban stores (real estate and
operational costs), the retailer’s fulfillment costs were significantly higher (43%)
than the base-case. We address store- and vendor-related issues in more detail in
later sections.
A related issue in order fulfillment is whether changes in online demand
may effect the retailer’s mix of fulfillment options. We conducted a post hoc test
to evaluate this effect by considering demand at different levels using multiplier
p = {1.0, 0.8, 0.6, 0.4, 0.2} to represent the proportion of online demand for
the retailer. Note that p = 1.0 represents demand in the base case. Because,
previous results had identified that order fulfillment and warehousing cost factors
affect fulfillment choices, we included these factors in the post hoc evaluation as
well.
The results (see Table 7) show that the DTC fulfillment option is greatly
affected by the level of online demand as it requires sufficient demand volume to
justify use of DTC facilities. For example, under the base-level demand scenario
for the (dF = $5.25, fF = $15.00) setting, 75% of online demand was filled using
the DTC option (the number in parenthesis identifies the number of DTCs used
for fulfillment). However, when demand falls below p= 0.6 level (40% less than
the base case demand), the DTC fulfillment option was no longer used. Below this
threshold, DC fulfillment was used to process all orders. It is interesting to note
that a very large reduction in demand (40%) would be needed to overcome the cost
advantage of the DTC option at dF = $5.25 level (which represents a 20% lower
unit fulfillment cost rate compared to the DC option).
The results also show that a shift from DTC to DC fulfillment occurred
step-wise depending on changes in the demand level. For example, in the case of
(dF = $5.95, fF = $5.00) under the base case demand setting, two DTC facilities
handled 100% of online orders. However, when demand level was set at p= 0.8
level (demand reduced by 20%), only one facility was used to fill 77% of demand
and the remaining online orders were filled by two DCs operated by the retailer.
The next shift in the mix of fulfillment options was recorded for levels below p=
0.4 (demand reduced by 60%) when the DTC facilities were closed and all orders
were filled by the retailer’s DCs.
$5.25 0.75 1.0 100% (2) − 100% (2) − 75% 25% (4)
(1)
0.8 100% (2) − 100% (2) − 75% 25% (4)
(1)
0.6 100% (1) − 100% (1) − 75% 25% (4)
(1)
0.4 100% (1) − 100% (1) − − 100% (6)
0.2 100% (1) − − 100% (6) − 100% (6)
$5.95 0.85 1.0 100% (2) − 65% (1) 35% (4) − 100% (6)
0.8 77% (1) 23% (2) 65% (1) 35% (4) − 100% (6)
0.6 77% (1) 23% (2) − 100% (6) − 100% (6)
0.4 77% (1) 23% (2) − 100% (6) − 100% (6)
0.2 − 100% (6) − 100% (6) − 100% (6)
$6.65 0.95 1.0 − 100% (6) − 100% (6) − 100% (6)
0.8 − 100% (6) − 100% (6) − 100% (6)
0.6 − 100% (6) − 100% (6) − 100% (6)
0.4 − 100% (6) − 100% (6) − 100% (6)
0.2 − 100% (6) − 100% (6) − 100% (6)
Evaluation of Order Fulfillment Options in Retail Supply Chains
Ishfaq and Raja 23
and rows represent the cost differential between store and DC options. Note that
both factor levels are normalized over the same (1.00 dD = $7.00) level.
The results in Table 8 show that for dR = $9.00 level, there is no use of the
store fulfillment option. It is interesting to note that the store fulfillment option
was selected only for the highest dF level (dF = $6.65 ). At this level, the store
fulfillment option was only used for the store order fulfillment cost dR = $7.00
level. In this case, 10% of the orders were filled from stores. A deeper review of
the results indicated that this situation was a result of orders that were filled from
stores located within USPS zones 1 and zone 2 of the retailer’s DCs, and thus were
marginally equivalent in logistics costs to fulfillment by DCs.
To further evaluate the shift from DC fulfillment to store fulfillment, we com-
piled results for an additional level of the order fulfillment cost factor: dR =$6.00
(see last row of Table 8). Note that we added this factor level for post hoc analysis
in this section only. The results show that the store fulfillment option becomes
the dominant option at this level, which corresponds to 15% reduction in store
fulfillment costs compared to the base case (i.e., ddRD = 0.85). This result reveals
that retailers would need to improve store fulfilment processes to reduce fulfillment
costs.
option: order delivery and retail inventory allocation for online orders. Typically,
retailers hold retail inventory to serve in-store customers, but some portion of this
inventory may be allocated for filling online orders.
To evaluate the effect of order delivery costs (δR ), retail inventory allocation
(κR ), and store fulfillment costs (dR ) on store fulfillment, results were compiled in
Table 9. The results show that store fulfillment is useful when the retailer’s order
delivery cost (δR ) is less than $3.65 for USPS zones 1 and 2 deliveries, and its
order fulfillment cost is around the dR = $8.00 level. At these levels, 9% of orders
were filled from stores. If the retailer can negotiate a lower unit order delivery rate
of δR = $2.75 for USPS zones 1 and 2 (a 40% discount from USPS published
rates), store fulfillment would increase significantly. The executives at the focal
retail firm confirmed that this level of shipping rate discount is negotiable with
package carriers, given the retailer’s current volume of online orders and projected
growth in the online sales channel.
The results show that low store inventory allocations (κR < 300) result in
fewer orders filled from stores. For example, in the case of dR = $7.00 and δR =
$2.75, 100% orders were filled by stores when κR ≥ 300. For inventory allocation
levels between 270 and 210, the proportion of store filled orders was reduced
from 95% to 75%. This result is important for retailers in establishing inventory
allocation targets if they want to utilize stores in their supply chain strategy. It is
also interesting to note that the use of retail stores was high for dR + δR ≤ $10.00.
This means that retailers can evaluate their order fulfillment and order delivery
costs collectively to set operational targets such that these costs (in total) are less
than the target threshold value ($10.00 for the case study data) when considering
the store fulfillment option.
Under the pick-from-store option, customers place orders online but are
willing to collect their orders from a local store to avoid order delivery charges.
In this setting, retailers do not incur order delivery costs (δR = $0.00). Note that
this case also applies to when customers pay for home delivery and the orders are
filled from the local stores, which would also result in δR = $0.00. The results of
the pick-from-store case are presented in the last column of Table 9. The results
show that the use of store fulfillment increased significantly under the pick-from-
store option (100% orders were filled from store), irrespective of the unit store
fulfillment cost level. A significant use of store fulfillment is seen even for the
base-case dR = $10.00, which confirms the $10.00 threshold for store fulfillment
discussed above. The only time the use of store fulfillment option reduced was
when store inventory allocation was restricted to a level less than the base-case.
be optimized for a retailer’s customer base, forcing orders to be shipped over long
distances and incurring higher order delivery charges.
The results are compiled (see Table 10) for different levels of the order
fulfillment fee (dV ) charged by vendors to fill online orders, and order delivery
costs (δV ). The results show that the Vendor fulfillment option is useful for the
retailer when total fulfillment and delivery costs (dV + δV ) are less than $9.10 per
order. For order delivery costs of δV ≤$5.10, a significant use of Vendor fulfillment
was seen in the results. When the combined fulfillment fee and delivery costs were
set at $8.10 (dV = $3.00, δV =$5.10) per order, 74% of demand was filled by
vendors.
Retailers with sufficient shipment volume often negotiate significant dis-
counts on package delivery charges that can be as high as 60% (Stevens, 2016).
To test the effect of a lower order delivery cost level, we compiled results for an
additional factor level of vendor delivery costs (δV = $3.42, i.e., 60% discount
on base case rates), for post hoc analysis. At this lower delivery cost level, re-
sults show a higher use of Vendor fulfillment. For example, 87% of orders were
filled through vendor fulfillment for combined fulfillment and delivery costs of
dV + δV = $6.42. These results show that in order to achieve a high use of Vendor
fulfillment, retailers would need a lower fulfillment fee charged by the vendors
(around $3.00 per order) and to negotiate a significant discount on order delivery
charges with package carriers.
The focal firm in our study can leverage its existing store-facing distribution
operations to process online orders. Such an arrangement would require the retailer
to redesign its DC processes to handle two different types of orders (online orders
and store replenishment orders), resulting in suboptimal order fulfillment costs.
Table 11 presents operational and logistics costs computed for this scenario for
the focal firm (see column #2, referred to as the base case). We compare these
costs with the case of DTC fulfillment. A key operational benefit of using DTC
facilities focusing exclusively on processing online orders is lower order fulfillment
costs. However, operating additional facilities adds to total costs for the retailer.
To control such costs and to ensure economies-of-scale, retailers use sufficiently
large fulfillment centers. Results for the DTC fulfillment option (see column #3)
show that while warehousing and order delivery costs were higher for the retailer,
potential savings in fulfillment costs made it a better (lower cost) option. Overall,
total costs for DTC fulfillment were lower (−7.5%) for the retailer as compared to
the base case.
Next, we explored the use of store fulfillment for the focal firm. In evaluating
this option, we noted that current levels of fulfillment costs at the firm do not
justify the use of stores. Hence, even when {R} was included in the {DC, V,
DTC} scenario in Table 11, stores were not used for order fulfillment. To further
explore the use of store fulfillment, we evaluated two additional scenarios. In store
option 1, we used the insights as discussed earlier (i.e., dR + δR ≤ $10.00) and set
the unit store fulfillment cost at $5.00 per order and local delivery rate at $4.00 per
order. The results (see column #4) show this setting is favorable for using stores
for order fulfillment. Under this option, shipping and inventory costs were quite
high for the retailer. However, lower last-mile delivery costs provided a significant
cost differential from the base case resulting in a major reduction (−18.55%) in
total costs for the retailer.
A related setting in store fulfillment is that retailer’s online customers pick
their orders from local store (known as BOPIS—Buy Online Pick In Store). In this
setting, retailer’s order delivery costs would be further reduced. We considered this
option as a special case of the previous scenario where all online orders will be
picked up by customers in local stores. The purpose of this setting is to evaluate
the effect of a higher store fulfillment cost rate that reflects the current challenge
for retailers to execute store fulfillment at requisite low costs. For this scenario,
dR =$10.00 was used to represent a fulfillment cost rate which is double that of
28 Evaluation of Order Fulfillment Options in Retail Supply Chains
the previous scenario. The results confirm (see column #5) high fulfillment costs
for the retailer, however, this option still costs the retailer less than the current DC
fulfillment option (−10.36%) due to savings in order delivery costs. While it is
unrealistic to assume that all online orders will be picked up from local stores,
this scenario demonstrates that last-mile delivery cost is a significant expense
in order fulfillment. It also highlights the benefit of the BOPIS arrangement for
retailers in using store fulfillment even when the fulfillment cost rate at stores
remains high.
In light of the analysis above, we note that retailers’ choices of order ful-
fillment options depend greatly on the underlying operational and logistics costs.
Under prevalent conditions, the focal firm in our study can see cost reductions by
using DTC fulfillment, even when the expense of operating additional facilities is
included. A bigger cost advantage is possible when retailers use their stores for
order fulfillment, mainly due to the reduction in last-mile delivery costs. We note
that these results reflect cost benefits for the retailer, however, managing separate
and independent fulfillment options in the physical and virtual channels would
increase administrative overheads. Retailers would need to consider this aspect
while evaluating their specific circumstances. We also note that realizing requisite
low store fulfillment cost rates will be challenging for retailers that struggle with
store execution in the physical channel. In such cases, retailers may be better off
not complicating store operations and employ the DTC fulfillment option.
DISCUSSION
Implications for Research and Practice
As retailers seek to employ suitable order fulfillment options to manage online
orders, the difference in operational requirements between filling online orders and
managing store replenishment orders presents a major challenge. This challenge
arises when retailers try to integrate the online order fulfillment process within
their existing store-based distribution processes. For order fulfillment, store-based
retailers have four options available: existing store-facing DCs, retail stores, new
direct-to-customer fulfillment centers, and fulfillment through vendors. The results
of this study show that economies-of-scale inherent in large-size distribution and
fulfillment centers provide a strong rationale for use of such facilities over store
and vendor fulfillment methods.
Using the warehousing cost measure, our study revealed the additional ex-
pense of operating direct-to-customer facilities must be offset by improvements
in order fulfillment efficiency compared to DC fulfillment. Under the current lo-
gistics cost structure of the focal retailer, our study found that proposed DTCs
offer a marginal advantage over DC fulfillment. Our results also show that DTC
fulfillment is not a good option if its relative operational benefit in unit order ful-
fillment costs is less than 15% of DC fulfillment costs. Hence, retailers looking
to use direct-to-customer fulfillment facilities need to employ logistics processes
that provide sufficient operational cost differential compared to retailers’ existing
DCs. Otherwise, retailers are better off using existing store-facing DCs for filling
online orders.
Ishfaq and Raja 29
An important outcome of our study is that retailers can assess the benefit of
facility improvements (such as installing automated material handling equipment,
and acquiring new technology for order picking and employee training), and ensure
such an investment provides requisite operational efficiency gains that result in
lower order fulfillment costs. For example, our results show that the use of DTC
did not change, even with 50% higher warehousing costs/investments, if such
an increase would help the retailer reduce the order fulfillment costs by 30%
or more.
Our article shows that retailers’ order delivery costs directly impact their
order fulfillment choice in that negotiating lower delivery rates alters the use-
fulness of order fulfillment options. For example, a 15% discount in package
delivery charges resulted in a major shift toward order fulfillment through ex-
isting DCs for the focal firm in our study. This finding offers retailers the op-
tion to avoid major capital investments required for new fulfillment facilities
(DTCs), and instead leverage existing distribution facilities (DC) to fill online
orders.
The high cost of last-mile order delivery has been a major concern for store-
based retailers. To lower this expense, retailers have considered using their retail
stores to fill online orders. Our results have shown store fulfillment can be a viable
option for retailers with competitive order fulfillment costs. The analysis, based
on economic estimates of the labor wages of sales associates employed in retail
stores and order delivery charges for the last mile, shows that retailers need to keep
the combined order fulfillment and delivery charges under $10 per order. Given
relatively high wage rates for sales associates in retail stores as compared to DC
employees, the best opportunity for retailers to achieve this goal is to reduce order
delivery charges for local (USPS zones 1 and 2) order deliveries. For example,
a 40% rate reduction compared to the undiscounted USPS parcel shipping rates
used in this study significantly increased the use of store fulfillment over DC
fulfillment.
While a higher wage rate of store employees is a concern in store fulfillment,
there are other aspects that would compensate for this downside. In consultations
for this research, retailers identified that a major motivation for store fulfillment
is the ability to sell items online at full price that are otherwise slated for price
markdown or clearance. In this setting, the consideration of lost revenue due to a
price markdown would offset the higher order fulfillment cost in stores.
In a store retail setting, many retailers leverage their relationship with vendors
engaged in the retailers’ store business. By hosting the products of these vendors
on the retailers’ Web sites, customers are offered access to a much broader product
assortment online than what is stocked in stores. This setting is beneficial for
a retailer in attracting more customers to its brand. The retailer pays an order
processing fee to the vendor to fill online orders and contracts out order deliveries
to package carriers to ship orders to customers directly from vendor facilities. Our
analysis has shown that the current order processing fee structure used by vendors
does not offer a profitable option for retailers. We observed in the study that the
threshold for the order processing fee and delivery costs for the vendor fulfillment
option is around $9.10. At this cost point, vendor fulfillment was used for 18%
of all online orders. Further reductions in the order processing fee and delivery
30 Evaluation of Order Fulfillment Options in Retail Supply Chains
costs led to a much higher use of this option. Our results show that nearly 75%
use of vendor fulfillment is possible when order processing and delivery costs are
about $8.10 per order. The results also show that retailers are better off negotiating
a lower order delivery rate than attempting to reduce the order processing fee
charged by vendors, as the former would have a higher marginal impact on the use
of the vendor fulfillment option.
is that the same customer may decide to have the order delivered to their home
by paying the shipping costs in one order and select the pick-from-store option
for another order which incurs no last-mile order delivery cost for the retailer.
An operational-level study can address the case of multiple items per order where
pooling benefits can be realized by shipping multiple items together in the same
box. However, issues related to split shipment of orders due to multiple fulfillment
locations would need to be considered in this case. This setting requires complex
product flow and inventory placement decisions in addition to deciding the mix of
order fulfillment options.
CONCLUSIONS
This research presented a detailed framework of the order fulfillment process in the
retail supply chain. This framework considers multiple fulfillment options avail-
able to store-based retailers that include existing network of stores and distribution
(DC) facilities, new direct-to-customer (DTC) fulfillment centers, and collaborat-
ing with vendors to fill online orders directly from vendor facilities. In an empirical
study, these fulfillment options were evaluated to provide managerial insights re-
garding each fulfillment option and identify operational and cost thresholds where
a particular fulfillment option would be preferred.
Using data from a large U.S. retailer and a factorial design approach, this
study found that large distribution facilities (DCs and DTCs) provide economies-
of-scale more suitable for order fulfillment compared to other options. However,
retailers would need to carefully consider the corresponding increase in operating
costs in their DC/DTC networks. The study identified an efficiency threshold,
reflected by the 10% differential in order fulfillment costs, for the DTC option
to justify operating dedicated facilities to fill online orders. Conversely, retail
firms that can operate DTC facilities with low operating costs would not need to
achieve a prohibitively high order fulfillment efficiency goal to justify the use of
DTC facilities. The study also indicated that the usefulness of an order fulfillment
option depends greatly on the corresponding order delivery process. For example,
in our study, a 15% discount in order delivery charges allowed the retailer to use
existing DCs and avoid the need to operate dedicated (DTC) fulfillment facilities.
This study finds that stores are costly order fulfillment nodes for retailers
compared to the distribution facilities. Retailers would need to improve store ful-
filment (through improved store processes and better training of store associates)
to bring the combined fulfillment and order delivery costs under $10.00 per order.
Above this threshold, store fulfillment is not a very competitive option compared
to DCs and DTCs. The results have also highlighted that it is important for re-
tailers to allow sufficient store inventory for online orders to maximize use of
store fulfillment, especially for the pick-from-store delivery option. The study also
found that a lower order processing/fulfillment fee, charged by vendors, would
be needed (reduced to around $3.00 per order) to make vendor fulfillment a cost-
competitive fulfillment option for retailers. Another approach that would favor
vendor fulfillment is to negotiate discounted delivery rates with package carriers
so that retailers’ total cost per order is below the $6.40 threshold.
32 Evaluation of Order Fulfillment Options in Retail Supply Chains
Overall, this study presented a detailed analysis of different settings for the
order fulfillment and order delivery process and demonstrated how changes in the
underlying cost structure can impact the usefulness of one fulfillment option over
another. The study concluded that the scale economies achieved in DC and DTC
fulfillment facilities outperform the store and vendor fulfillment options. However,
with some changes in the order fulfillment and delivery processes, retailers can
leverage their stores more effectively for order fulfillment.
REFERENCES
Acimovic, J., & Graves, S. C. (2015). Making better fulfillment decisions on the
fly in an online retail environment. Manufacturing & Service Operations
Management, 17(1), 34–51.
Agatz, N., Fleischmann, M., & Nunen, J. V. (2008). E-fulfillment and multi-channel
distribution: A review. European Journal of Operational Research, 187(2),
339–356.
Aksen, D., & Altinkemer, K. (2008). A location-routing problem for the conver-
sion to the click-and-mortar retailing: The static case. European Journal of
operational Research, 186(2), 554–575.
Alptekinoğlu, A., & Tang, C. (2005). A model for analyzing multi-channel distribu-
tion systems. European Journal of Operational Research, 163(3), 802–824.
Amazon. (2015). Fulfillment by Amazon, accessed May 10, 2017. Avail-
able at http://www.amazon.com/gp/help/customer/display.html/?nodeId=
201119430
Ayanso, A., Diaby, M., & Nair, S. K. (2006). Inventory rationing via drop-shipping
in internet retailing: A sensitivity analysis. European Journal of Operational
Research, 171(1), 135–152.
Barry, J. (2012). Controlling and reducing your fulfillment costs, accessed
May 10, 2017. Available at http://www.fcbco.com/articles-and-whitepapers/
articles/bid/129456/
Bell, D., Gallino, S., & Moreno, A. (2014). How to win in an omnichannel world.
MIT Sloan Management Review, 56(1), 45.
Bretthauer, K., Mahar, S., & Venakataramanan, M. (2010). Inventory and dis-
tribution strategies for retail/e-tail organizations. Computers & Industrial
Engineering, 58(1), 119–132.
Buis, M.L. (2008). FMLOGIT: Stata Module Fitting a Fractional Multinomial
Logit Model by Quasi Maximum Likelihood, Statistical Software Compo-
nents.
Chao, L., & Norton, S. (2016). Retailers bet big on retooling their supply chains
for e-commerce. Wall Street Journal. Accessed May 10, 2017. Available
at https://www.wsj.com/articles/retailers-bet-big-on-retooling-their-supply-
chains-for-e-commerce-1453977002.
Croxton, K., & Zinn, W. (2005). Inventory considerations in network design.
Journal of Business Logistics, 26(1), 149–168.
Ishfaq and Raja 33
Rafay Ishfaq is the HCoB Endowed Research Fellow and associate professor
of supply chain management in the Raymond J. Harbert College of Business at
Auburn University. He received his PhD degree in operations management from
The University of Alabama. Combining mathematical modeling and field stud-
ies, his research addresses issues related to strategic planning of operations and
efficient use of resources. His research work in supply chain management, informa-
tion systems, health care, and energy is published in scholarly journals including
IIE Transactions, European Journal of Operational Research, OMEGA, Supply
Chain Quarterly, International Journal of Physical Distribution and Logistics
Management, Journal of Operational Research Society, Computers and OR, and
International Journal of Logistics Management. His research has been recognized
by the Association of European Operational Research Societies with the Best
Ishfaq and Raja 35
EJOR Research Paper Award. His research has also appeared in Science Direct’s
Top-25 Hottest Articles list in multiple years.