Journal Pre-proof

Designing a sustainable multi-channel supply chain distribution network: A case study

Arezoo Vafaei, Saeed Yaghoubi, Javad Tajik, Farnaz Barzinpour

PII: S0959-6526(19)34498-1
DOI: https://doi.org/10.1016/j.jclepro.2019.119628
Reference: JCLP 119628

To appear in: Journal of Cleaner Production

Received Date: 17 August 2019

Revised Date: 10 November 2019
Accepted Date: 8 December 2019

Please cite this article as: Vafaei A, Yaghoubi S, Tajik J, Barzinpour F, Designing a sustainable multi-
channel supply chain distribution network: A case study, Journal of Cleaner Production (2020), doi:
https://doi.org/10.1016/j.jclepro.2019.119628.

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2019 Published by Elsevier Ltd.

 "CRediT Author Statement"

Arezoo Vafaei: Software, Validation, Investigation, Formal analysis, Resources, Data Curation,
Writing - Original Draft Preparation.
Saeed Yaghoubi: Conceptualization, Methodology, Formal analysis, Writing- Reviewing and
Editing, Visualization, Supervision, Project administration.
Javad Tajik: Software, Validation, Formal analysis, Data Curation, Writing- Reviewing and Editing.
Farnaz Barzinpour: Formal analysis, Methodology, Visualization.
 Designing a sustainable multi-channel supply chain distribution
network: A case study

Arezoo Vafaeia, Saeed Yaghoubia,1, Javad Tajika, Farnaz Barzinpoura

a
School of Industrial Engineering, Iran University of Science and Technology, Tehran, Iran

1
Corresponding author: Tel:+982173225053; Fax:+982173021537, E-mail address: [email protected]

1
 Designing a sustainable multi-channel supply chain distribution network: A
case study

Abstract
The existence of numerous social, economic, and environmental problems and the need to improve
the long-term performance in organizations have led the supply chain designs toward sustainable
designs. On the other hand, due to the emersion of new technologies and the creation of new
business models, the increase of competition level among retailers and the raise of customers
expectations, the use of a multi-channel distribution system is nowadays growing. Therefore, the
appropriate distribution channel according to the product type and the number of vehicles for the
transportation of the products is a significant and important issue in retailing systems. In this
regard, in this paper, a mixed-integer programming model for the sustainable distribution network
design, considering multi-product, multi-echelon and multi-transportation mode especially third
party logistics (3PL), is presented. The economic, environmental, and social objectives are
mathematically formulated, where these objectives include: 1) minimizing transportation costs,
purchasing vehicles and building warehouses as the first objective, 2) minimizing the amount of
carbon dioxide released by transport vehicles and building warehouses, 3) maximizing the number
of job opportunities. After linearization of the proposed model, it is solved using the goal
programming technique by GAMS software. Finally, to evaluate the applicability of the proposed
mathematical model, Digikala company is used as a case study.
Keywords: Sustainable distribution network design, Retailing system, Multiple channels, Goal
programming, Third party logistics (3PL).

1. Introduction

Due to the increase in production rates and geographical distances between suppliers and
customers, distribution operations have an important place in the supply chain performance.
Therefore, the supplier would inevitably bring his products to the customer at long distances,
through a person called the wholesaler or retailer. A proper facility location, selecting appropriate
distribution channels, and choosing a suitable type of vehicle will have social, environmental, and
economic impacts on the region of the construction site of the factory, in addition to the economic
impact on the performance of the industrial unit. Studies related to finding the optimal location of
warehouses, choosing the appropriate distribution channels, reducing the adverse environmental
impacts and the social problems caused by the lack of job positions, not only have academic value
but also have numerous economic benefits.
To get the customer satisfaction, cost reduction, and adverse environmental impacts between the
supplier and the customer, the present study intends to design a sustainable distribution network by
employing a mixed-integer programming model with a multi-channel, multi-objective, single-
period, multi-product, and multi-vehicle, multi-3PL problem and with the free shipping. Besides,
in the proposed model, it is considered that if the customer’s order is higher than a specific price,
free shipping is suggested by the retailer companies.

1
 One of the major challenges in the distribution network design is choosing the optimal
distribution channel among multi-channel distribution networks. In these distribution channels,
customers' orders can be received from any facility (suppliers, retailer warehouses). Furthermore,
sustainability objectives are used such as reducing costs, reducing environmental impacts, and
increasing the number of job opportunities. In order to improve the distribution network's
performance, it can be referred to the research carried out by Cintron et al. (2010). As shown in
Fig. 1, in the distribution channel 1 and 2, the distribution of the product is carried out by the
company and in the distribution channel 3 and 4, the company assigns the product to a company of
the third-party logistics.

Fig.1. Different distribution channels in the research of Cintron et al. (2010)

According to the above by considering the distribution network costs and environmental and
social impacts, the present paper is seeking an optimal method for distributing the product from the
supplier to the customer. In the multi-objective, single-period, multi-product, and multi-vehicle,
multi-3pl model delivering products to customers is made up of several distribution channels.
Also, with the investigation of different distribution channels and the choice of the appropriate
vehicle types, the best distribution channel of each product is selected for each customer and it is
presented as a mixed-integer programming model. Therefore, the objectives of this paper are
minimizing the cost of establishing warehouse, the cost of buying the vehicles, the cost of
transporting products, minimizing the amount of carbon dioxide released by the establishing
warehouse and transporting different vehicles, and maximizing the number of job opportunities.
The model of this paper is solved by CPLEX solver of the GAMS® optimization software, and in
order to validate the proposed model, the data of Digikala company is used as a case study.
In the supply chain structure of the proposed model, the products are sent to the customer in two
ways: a. direct: from suppliers to customers and b. indirect: collecting from different suppliers,
transferring to the retailer's warehouses, and then transferring them to customers. Furthermore, the
product distribution is done in three ways: a. supplier distribution team, b. retailer distribution
team, and c. third-party logistics.
Designing an effective distribution network can significantly reduce the cost of transportation
and increase the customer satisfaction. As a result, the design of the distribution network can be
considered as a key factor for the company's profitability. At the macro level, due to the
importance and significance of the distribution sector in the country's gross domestic product and

2
the macro objectives set in the macro plans in this regard, improving the efficiency of networks
and distribution infrastructure is one of the important points for reorganizing of the supply chain in
the country. The effects of this change will be on improving the country's economic indicators,
especially GDP, the transition of goods, increasing the share of services in the economy, and
strengthening the country's economic status in the region. As a result, one of the most important
issues faced by managers is deciding the distribution system of the goods. This decision has direct
effects on other marketing decisions such as pricing, advertising, packaging, etc.
Our model has some advantages in comparison with other studies, five of which will be deeply
explained in what follows:
• Multi-channel distribution networks are considered in the modeling. Therefore, the best
distribution channel for a specific type of product is selected and this leads to the reduction
of transportation cost.
• Considering different vehicles for using the retailer's distribution team as well as selecting
the optimal type of vehicle.
• Modeling in order to consider free shipping.
• Studying and modeling the different aspects of sustainability (economic, environmental,
and social) in the supply chain, in accordance with the real-world needs, which have a
practical aspect.
• Finally, the data of Digikala company is used as a case study to demonstrate the
applicability of the model.
Here are the questions of research of the paper:
• What types of distribution channels are typically best suited for specific types of products?
• What is the optimal allocation of product from each warehouse to the customers in the
supply chain?
• What will be the configuration of the supply chain network to compromise between
economic, environmental, and social objectives?
• What are the environmental and social impacts of the supply chain and how are they
calculated?
The rest of the paper is investigated as follows: The literature review is examined in section 2.
The problem definition and mathematical model is expressed in section 3. The case study and
sensivity analysis are mentioned in section 4 & 5, respectively. The conclution and future research
of the paper are finally indicated in section 6.

2. Literature Review

In this section, the topics we are discussing fall into three areas: network design, distribution
systems, and sustainability.
Geoffrion and Graves (1974) were among the first authors to consider middle distribution
equipment in addition to the manufacturer's equipment. They proposed a mixed-integer linear
programming model for the design of a multi-product distribution system with capacity
constraints. In a study by Jayaraman (1998), the relationship between inventory management,
facility locating, and transportation policy designation has been considered. So that it analyzes the

3
interdependence between the three domains and proposes an integrated model of mixed-integer
programming for the design of the distribution network.
Jayaraman and Ross (2003), have investigated PLOT design system (production, logistics,
outsourcing, transportation). The system explores a range of distribution network design issues that
include several product types, a central production facility, multiple distribution centers, temporary
storage warehouses, and retailer (customer), which each of the units demands to different products.
To solve this problem, a simulated annealing algorithm has been used.
Alptekinoglu and Tang (2005) have studied a kind of retailer network, which consists of
suppliers, warehouses, and stores based on the customer’s profiles. Distribution is carried out
through two direct (from warehouses) and indirect (through stores) channels. Using the
decomposition method, the problem is subdivided into problems which they are solved using a
heuristic method. A research conducted by Amiri (2006) has addressed the issue of designing a
distribution network in a supply chain including the location of facilities (factories and distribution
depots) and yet determined the best strategy for distributing the product from the factory to the
warehouse and from the warehouse to the customer. For this purpose, a composite integer
programming model is presented in which the facility has multiple capacity levels and the optimal
capacity of the facility is determined by the model. A heuristic method has been proposed to solve
the problem.
Ko et al. (2006) also have proposed a hybrid optimization/simulation methodology in order to
design a distribution network for third-party logistics companies. In this study, a genetic algorithm
is used to solve the model. On the other hand, Selim and Ozkarahan (2008) have designed a
distribution network by means of fuzzy goal programming. The purpose of the model is to select
the optimum number, location, and capacity levels of the factories and warehouses for delivering
the products to the retailers with the lowest cost and maximum satisfaction level. In a study by Lee
et al. (2007), for the design of a forward and reverse integrated distribution network in third-party
logistics companies, an integer linear programming model is presented. Due to the complexity of
the problem and a large number of variables and model constraints, heuristic solving methods
including genetic algorithm and two greedy algorithms are proposed.
Several studies have been carried out on the status and function of the network and the
distribution channels of products. Hulthén (2007) has concluded that the factors such as the
creation of networks and large distribution chains, the market analysis and optimal market
segmentation could greatly reduce the cost of physical distribution of goods. In the paper by
Lorentz et al. (2007), it is shown that the factors such as the use of modern technologies (including
e-commerce), the creation of large distribution chains (such as retailers’ stores and proprietary
stores), and advertising tailored to each market segments are the placeholders to improve the
efficiency of distribution networks in these areas.
Ross and Jayaraman (2008) have presented a new heuristic solution to locate temporary storage
and distribution centers in the design of the supply chain network. They have proposed two
heuristic methods in their study using a simulated annealing algorithm and Tabu Search.
Akkerman et al. (2009) have presented a mixed-integer Programming model for the production
and distribution of persistent ready-made meals. Cintron et al. (2010) have presented a multi-
criteria mixed-integer linear programming model for designing a distribution network. The
purpose of this model is to take tactical decisions, or specifically, to design the flow of products

4
from the manufacturing plants to customers. For validate of the model, data from a consumer
goods company is applied, which is used from two distribution centers located in the same area.
Grant and Banomyong (2010) have conducted a study on the design of the distribution chain for
consumer goods in Thailand and Japan. In their study, they have concluded that the skill and
expertise of human resources, the availability of appropriate infrastructure for the distribution and
transportation of goods, and the use of new technologies, had a great effect on improving
efficiency and reducing the distribution costs of consumer goods are in these countries. Millet
(2011) has examined the criteria for achieving a sustainable supply chain that simultaneously
incorporates economic, social and environmental considerations. Sadjady and Davoudpour (2012)
have proposed a mixed-integer linear programming model for the design of a dual-level supply
chain network with multi-product delivery. Shu et al. (2013) have proposed a model for designing
a logistical distribution network, including a supplier, a set of potential warehouses, and a set of
crude vendors offer two different product categories.
In a study by Mangiaracina et al. (2015), 126 articles from 1970 to 2013 have been studied.
Bortolini et al. (2016) have presented a multi-objective linear programming model for a
sustainable food distribution network. These objectives include minimizing operational costs,
carbon effects, and delivery times. Zhang et al. (2016) have presented a mixed-integer linear
programming model for designing a sustainable supply chain network with multiple channels. This
model has economic, social, and environmental objectives. The first objective is to reduce the
costs of setting up facilities and the cost of transporting products. The second objective is to
maximize customer service and the third objective is to measure the environmental impact of the
network in question, which is based on the operation of the facility and the transportation
provided. Janatyan et al. (2018) have presented a multi-objective model to design a pharmaceutical
distribution network according to the main concepts of sustainability. Yadav et al. (2018) have
introduced an MCDSCN MILP model to integrate online giants with local distribution network
retailers. Ouhimmou et al. (2019) have addressed the problem of designing a robust distribution
network under demand uncertainty based on a real industrial case study in pulp and paper.
Akgün and Erdal (2019) have studied the strategic-level ammunition distribution network
design problem of the GCG (General Command of the Gendarmerie) where the purpose is to
determine the number and locations of depots and the service assignments considering several
factors. Samani et al. (2019) recently proposed a multi-objective mathematical model by
incorporating both quantitative and qualitative factors for distribution of blood supply chain,
considering a multilateral perspective in an uncertain environment.
The different features of mathematical models in the literature are summarized in Table 1, in the
design of supply chain distribution network from 2010 to 2019 and are compared to the proposed
model in this paper. The review of the literature (Table 1) shows that the number of articles which
have been considered transport mode selection, distribution channel selection and the three aspects
of sustainability (economic, environmental, and social) at the same time in their mathematical
model, is very limited. Also, there are no articles for free shipping model.

5
 Table 1. Comparison of literature of supply chain distribution network design with our model
Features

Transport mode selection

sustainability Objectives

Free Shipping model

Distribution channel
Multi-objective

selection
Author

Minimizing CO2
Minimizing cost

Maximizing the
Environmental

number of job
opportunities
Economic

emission
Social
Cintron et al. (2010) ✓ ✓ ✓ ✓
Shu et al. (2013) ✓ ✓
Azad & Davoudpour ✓ ✓ ✓
(2013)
Hlyal et al. (2015) ✓ ✓
Bortilini et al. (2016) ✓ ✓ ✓ ✓ ✓ ✓ ✓
Deqqaq & Abouabdellah ✓ ✓ ✓ ✓
(2016)
Zhang et al. (2016) ✓ ✓ ✓ ✓ ✓ ✓
Halat & Hafezalkotob ✓ ✓
(2017)
Barzinpour & Taki ✓ ✓ ✓
(2018)
Janatyan et al. (2018) ✓ ✓ ✓ ✓ ✓ ✓
Yadav et al. (2018) ✓ ✓ ✓ ✓ ✓
Validi et al. (2018) ✓ ✓ ✓ ✓ ✓ ✓
Guerrero Campanur et ✓ ✓
al. (2018)
Mogale et al. (2019) ✓ ✓ ✓ ✓ ✓
Our study ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

3. Problem definition and mathematical formulation

According to the introduction, reducing the costs of the distribution network, and the
environmental and social impacts of the supplier and the customer is very important. Therefore, it
is necessary to examine the distribution of the product from the supplier to the customer using a
multi-objective, single-period, multi-product, and multimodality with multiple modes of
transportation. Also, with the investigation of different distribution channels and the choice of the
appropriate vehicle types, the best distribution channel of each product is selected for each
customer. Therefore, designing a sustainable distribution network model with multiple channels is
a fundamental issue of the present study, which has objectives such as minimizing transportation
costs, purchasing vehicles, and building warehouses as the first objective, minimizing the amount
of carbon dioxide released by transport vehicles and building warehouses as the second objective,
and maximizing the number of job opportunities as the third objective. To achieve these
objectives, mathematical modeling has been performed and then the obtained model is linearized
and solved using a goal programming method. As shown in Fig. 2, in the supply chain structure of

6
the proposed model, the products are sent in two ways to the customer: a. direct: from suppliers to
the customer and b. indirect: collecting from different suppliers and transferring to the warehouse
of the retailer and then transferring it to the customer. The product distribution is also done in three
ways: a. Supplier distribution team, b. Retailer distribution team, and c. Third party logistics.
In this section, there is a sustainable design of single-periodic, multi-product, multi-echelon,
multi-transportation mode, and considering multiple channels. Hence, a mixed-integer
programming model has been formulated with multi objectives. In this issue, we identify the build
of the warehouse in the potential locations, the possibility of assigning the product distribution to
the third-party logistics company, or the retailer distribution team, and the type and number of
vehicles for the transportation of the product. In Fig. 2, this concept is presented. In this sense,
manufacturer/suppliers and warehouses are used as direct and indirect facilities for customers. The
objectives of the proposed model are as follows:
• The first objective is to reduce the cost of the retailer company, which is including the cost of
building warehouses, purchasing the vehicles, and transporting products.
• The second objective is to minimize the amount of carbon dioxide released by transporting
different vehicles and building warehouses.
• The third objective is to increase employment by taking into account the number of job
opportunities created by the construction of warehouses and transport of vehicles.

Fig. 2. Illustration of the proposed supply chain

7
3.1. Assumptions
The underlying assumptions of the problem are as follows:
• Demand and customer location are definitive.
• The customer is the final consumer.
• There is no capacity restriction for the warehouses.
• The location of the suppliers is predefined.
• Products are defined and each supplier supplies one type of product.
• Locations and the customer demand are preset.
• The potential locations of the warehouses are determined.
• Any facility can obtain its products only from upstream facilities except for customers.
• Customers can buy their products from just one distribution channel.
• The customer demand should be met.
3.2. Notations

i
Sets:

j
The set of suppliers (i = 1,2, ..., I)

s
The set of customers (j = 1,2, ..., J)

b
The set of supplier distribution team (s = 1,2, ..., S)

o
The set of potential locations for warehouse (b = 1,2, ..., B)
The set of third-party logistics companies from the supplier to the warehouse (o = 1,2, ..., O)
e
The set of third-party logistics companies from the supplier/warehouse to the customer (e =

r
1,2, ..., E)

t
The set of retailer distribution team from supplier to the customer/warehouse (r = 1,2, ..., R)
The set of retailer distribution team from the warehouse to the customer (t = 1,2, ..., T)
Parameters:
M A large real number
bmax Maximum number of warehouses allowed to build
The demand amount of customer j from supplier i

′
Capacity of retailer distribution team r vehicle for supplier product i
Capacity of retailer distribution team t vehicle for supplier product i
Product price i
The price level for free shipping

′
Cost of purchasing vehicle retailer distribution team r
Cost of purchasing vehicle retailer distribution team t
Fixed cost of establishing a warehouse b
Shipping cost of product flow from supplier i to customer j by 3pl e for free shipping if
Ⅱ
customer j purchases more than µ
Shipping cost of product flow from supplier i to customer j by retailer distribution team t for
Ⅰ
free shipping if customer j purchases more than µ
The cost of transporting product flow from warehouse b to customer j by the retail
Ⅲ
distribution team t for free shipping if customer j purchases more than µ
The cost of transporting product flow from warehouse b to customer j by 3pl e for free
Ⅳ
shipping if customer j purchases more than µ
Ⅰ Environmental impacts of the supplier distribution team s
Ⅱ Environmental impacts of establishing potential warehouse b
Ⅲ Environmental impacts of 3pl o

!"
Ⅳ Environmental impacts of 3pl e

!"′
Carbon di-oxide emission factor of vehicle retailer distribution team r
Carbon di-oxide emission factor of vehicle retailer distribution team t

8
 #$ Ⅰ
#$ Ⅱ
The number of job opportunities created by the supplier distribution team s

#$ Ⅲ
The number of fixed job opportunities created by establishing potential warehouse b

#$ Ⅳ
The number of job opportunities created by 3pl o

#$ Ⅴ
The number of job opportunities created by 3pl e

#$ Ⅵ
The number of job opportunities created by vehicle retailer distribution team r

%′
The number of job opportunities created by vehicle retailer distribution team t

%
The distance between supplier i and warehouse b

%′′
The distance between supplier i and customer j
The distance between warehouse b and customer j
Continues Variables:
&Ⅵ
The amount of product shipped from supplier i to warehouse b by supplier distribution team

&Ⅶ
s

&Ⅷ' (
The amount of product shipped from supplier i to warehouse b by retailer distribution team r
The amount of product shipped from supplier i to warehouse b by 3pl o
&Ⅲ
The amount of product shipped from warehouse b to customer j by retailer distribution team

&Ⅳ
t

&Ⅰ
The amount of product shipped from warehouse b to customer j by 3pl e

&Ⅱ
The amount of product shipped from supplier i to customer j by retailer distribution team t

&Ⅴ
The amount of product shipped from supplier i to customer j by 3pl e

)!Ⅰ
The amount of product shipped between supplier i, warehouse b and customer j
The number of vehicles needed to transfer the product from supplier i to customer j by the

)!Ⅶ
retailer distribution team t
The number of vehicles needed to deliver the product from supplier i to the warehouse b by

)!Ⅲ
the retailer distribution team r
The number of vehicles needed to transfer the product from stock b to customer j by the
retailer distribution team t
Binary Variables:

*
1 if the warehouse b is to be established, and 0 otherwise
1 if the order price of customer j is less than , and 0 if the order price of customer j is more

+
than

,
1 if the supplier i is allocated to the customer j by 3pl e, and 0 otherwise
1 if the supplier i is allocated to the customer j by the retailer distribution team t, and 0

-Ⅲ
otherwise
1 if the warehouse b is allocated to the customer j by the retailer distribution team t, and 0

-Ⅳ
otherwise

.Ⅵ
1 if the warehouse b is allocated to the customer j by 3pl e, and 0 otherwise
1 if the supplier i is allocated to the warehouse b by the supplier distribution team s, and 0

.Ⅶ
otherwise
1 if the supplier i is allocated to the warehouse b by the retailer distribution team r, and 0

.Ⅷ
otherwise
1 if the supplier i is allocated to the warehouse b by 3pl o, and 0 otherwise

3.3. Objective functions

The three-objective mathematical model presented in this model is in the form of mixed-integer
programming, which is presented in the following:

9
!'/ 1 = 2 × +222 ′ × )!Ⅰ +222 ′ × )!Ⅲ

+ 222 × )!Ⅶ + 365 × (1 − * )

× ;2 2 2 × &Ⅱ + 222 × &Ⅰ

(1)
Ⅱ Ⅰ

+ 222 Ⅲ × &Ⅲ + 222 Ⅳ × &Ⅳ <

!'/ 2 = 2 Ⅱ × + 365

× ;2 2 2 Ⅳ × % × &Ⅱ + 2 2 2 !"′ × % × &Ⅰ

+ 2 2 2 !"′ × %′′ × &Ⅲ + 222 Ⅳ × %′′ × &Ⅳ (2)

+ 222 Ⅰ × %′ × &Ⅵ + 2 2 2 !" × %′ × &Ⅶ

+ 222 Ⅲ × %′ × &Ⅷ <

!>? 3 = 2 #$ Ⅱ × + 2 2 2 #$ Ⅵ × )!Ⅰ + 2 2 2 #$ Ⅵ × )!Ⅲ

+ 2 2 2 #$ Ⅴ × )!Ⅶ + 365

× ;2 2 2 #$ Ⅳ × &Ⅱ + 2 2 2 #$ Ⅳ × &Ⅳ
(3)

+ 2 2 2 #$ Ⅲ × &Ⅷ + 2 2 2 #$ Ⅰ × &Ⅵ <

The objective function (1) tries to minimize the total cost of the retailer company. The first part
is related to the fixed cost of establishing the warehouse. The second, third and fourth parts belong
to the fixed cost of purchasing vehicles and the fifth part is free shipping costs if the customer’s
purchase is more than a certain amount. The objective function (2) minimizes the environmental
impacts of the network. The first part refers to the amount of carbon dioxide emissions from the
establishing of the warehouse. The second part is related to the amount of carbon dioxide
emissions from vehicle transportation by the supplier distribution team, the retail distribution team,
and the third-party logistics company. The objective function (3) is to maximize the number of job
opportunities. The first part shows the number of job opportunities created by the establishment of
the warehouse. The second, third and fourth parts relate to the number of job opportunities
resulting from the selection of the retailer distribution team and the fifth part refers to the number
of job opportunities arising from the transportation of the product by the third-party logistics
company and the supplier distribution team.

10
3.4. Constraints

Flow and establishing of the warehouse constraints:

&Ⅲ ≤ ×! ∀ ,,
∀ , ,C
(4)
&Ⅳ ≤ ×!
∀', ,D
(5)
&Ⅵ ≤ ×!
∀', ,E
(6)
&Ⅶ ≤ ×!
∀', ,(
(7)
&Ⅷ ≤ ×! (8)
2 = F>? (9)

Single allocation constraint:

2+ + 2, + 2 -Ⅲ + 2 -Ⅳ =1 ∀', , (10)

Fulfillment of demand constraints:

2 &Ⅱ + 2 &Ⅰ + 2 &Ⅴ ≥ ∀', (11)

2 &Ⅱ × + 2 &Ⅰ × + 2 &Ⅴ × ≤ × * + ! × (1 − * ) ∀', (12)

Capacity vehicles constraints:

&Ⅰ &Ⅰ
≤ )!Ⅰ ≤ +1 ∀', ,
′ ′
(13)

&Ⅶ &Ⅶ
≤ )!Ⅶ ≤ +1 ∀', ,E (14)

&Ⅲ &Ⅲ
≤ )!Ⅲ ≤ +1 ∀', , ,
′ ′
(15)

∀', ,C
Flow and allocation constraints:
&Ⅱ ≤ + × !
∀', ,
(16)
&Ⅰ ≤ , × !
∀ ,,
(17)
&Ⅲ ≤ - Ⅲ × !
∀ , ,C
(18)
&Ⅳ ≤ -Ⅳ × !
∀', ,D
(19)
&Ⅵ ≤ .Ⅵ × !
∀', ,E
(20)
&Ⅶ ≤ . Ⅶ × !
∀', ,(
(21)
&Ⅷ ≤ . Ⅷ × ! (22)

∀ ,,
Allocation and establishing of the warehouse constraints:
-Ⅲ ≤
∀ , ,C
(23)
-Ⅳ ≤
∀', ,D
(24)
.Ⅵ ≤
∀', ,E
(25)
.Ⅶ ≤
∀', ,(
(26)
.Ⅷ ≤ (27)
Flow balancing constraints:
2 &Ⅵ + 2 &Ⅶ + 2 &Ⅷ = 2 &Ⅲ + 2 &Ⅳ ∀', , (28)

11
 H∑ &Ⅵ +∑ &Ⅶ +∑ &Ⅷ +∑ &Ⅲ +∑ &Ⅳ J
&Ⅴ = ∀', ,
2
(29)

&Ⅵ , &Ⅶ , &Ⅷ , &Ⅲ , &Ⅳ , &Ⅰ , &Ⅱ ,

&Ⅴ , )!Ⅰ , )!Ⅶ , )!Ⅲ ≥ 0 & '/
(30)

, * , + , , , - Ⅲ , - Ⅳ , .Ⅵ , .Ⅶ , .Ⅷ ∈ N0,1O (31)

Constraints (4-8) states that the product flow occurs between two tiers by the distribution teams
when the warehouse has been established. The constraint (9) specifies the maximum number of
permitted warehouses. The constraint (10) ensures that a distribution team is assigned to each
customer from each tier. The constraint (11) represents the satisfaction of customer demand. The
constraint (12) is for free shipping by the retailer distribution team when the order price of
customer j is more than . Constraints (13), (14), and (15) indicate the number of vehicles required
by the retailer distribution team, taking into account the capacity of the vehicle. Constraints (16-
22) suggest that the product flow occurs between two tiers when the shipment is allocated to
distributed teams. Constraints (23-27) states that the product is allocated between the two tiers
once the warehouse is established. Constraint (28) indicates flow balance. The constraint (29)
shows the amount of product flow from the supplier to the customer by the warehouse. Constraints
(30) and (31) show the range of variables.

3.5. Linearization of the proposed model

In the fifth part of the first objective function, two decision variables are multiplied, which leads
to the non-linearization of the model. For linearization of the model, the variables are defined as
(32-35):
&1 = H1 − * J × &Ⅱ ∀', ,C
&2 = H1 − * J × &Ⅰ ∀', ,
(32)

&3 = H1 − * J × &Ⅲ ∀ ,,
(33)

&4 = H1 − * J × &Ⅳ ∀ , ,C
(34)
(35)

The variables &1 , &2 , &3 , and &4 are positive integers, if * = 0, then they are equal to
&Ⅱ , &Ⅰ , &Ⅲ , and &Ⅳ respectively, otherwise equal to zero. By applying the
above changes, the fifth part of the first objective function becomes the relation (36):

222 Ⅱ × &1 + 222 Ⅰ × &2 + 222 Ⅲ × &3

+ 222 × &4
(36)
Ⅳ

And the constraints (37-48) will be added to the model:

&1 ≤ Q! × H1 − * JR + &Ⅱ ∀', ,C (37)
&1 ≥ &Ⅱ + ! × QH1 − * J − 1R ∀', ,C (38)
&1 ≤ ! × H1 − * J ∀', ,C (39)
&2 ≤ Q! × H1 − * JR + &Ⅰ ∀', , (40)

12
&2 ≥ &Ⅰ + ! × QH1 − * J − 1R ∀', , (41)
&2 ≤ ! × H1 − * J ∀', , (42)
&3 ≤ Q! × H1 − * JR + &Ⅲ ∀ ,, (43)
&3 ≥ &Ⅲ + ! × QH1 − * J − 1R ∀ ,, (44)
&3 ≤ ! × H1 − * J ∀ ,, (45)
&4 ≤ Q! × H1 − * JR + &Ⅳ ∀ , ,C (46)
&4 ≥ &Ⅳ + ! × QH1 − * J − 1R ∀ , ,C (47)
&4 ≤ ! × H1 − * J ∀ , ,C (48)

3.6. Multi objective solution approach: Goal programming

Goal programming is an optimization technique to solve problems with multiple objectives.

This technique seeks to find optimal solutions, which is predefined goal values exist for one or

sum of the positive and negative deviations (%ST (E %SU ) from each of the goals. For each goal,
more objectives and minimize deviations from these goal values. This methodology minimizes the

weights are given on a priority basis (VS ), indicating the importance of the objective function to
other objectives. As stated above, the general model of goal programming will be as follows.
W

!'/ 2 VS (%ST + %SU ) (49)

SXY
ZS + %SU − %ST = ZS∗ ℎ = 1, … , ^
%ST . %SU = 0 ℎ = 1, … , ^
(50)
(51)

To formulate a goal programming model, we must define the values of the goals for the
objective functions. To obtain the values of the goals, the mathematical model of the problem is
considered with each of the objective functions alone and the optimal value of each of the
objective function is obtained.
The objective function of the goal programming model is defined as (52).
!'/` = Va . %aT + V . % T + V . % T (52)

In fact, in this model, we seek to minimize the total of these deviations.

%bT : Positive deviation (undesirable) of the economic objective function
%T : Positive deviation (undesirable) of the environmental objective function
%dT : Positive deviation (undesirable) of the social objective function
%ST . %SU = 0, ℎ = e, C, D (53)

4. Case study: Digikala company

The expanding use of the Internet has led business activists to use this to introduce and sell
goods. Setting up Internet stores is the most important way to use the Internet. Digikala is an
online retailer platform. The company was originally dedicated only to the sale of digital and
electronic products, but today there are various product groups such as cosmetics, clothing, home

13
appliances, kitchens, books and stationery, toys, art and culture, sports, entertainment, travel,
supermarket, and a variety of car accessories and tools from a variety of brands. The products
provided on the company's website are made in two ways by purchasing from reputable companies
and official and legal importers and their offerings or the cooperation of some vendors with the
company to supply their products online for sale. In Fig. 3, an overview of the supply chain of the
Digikala company is provided.
Digikala company has no sales and after-sales services at any point in the country other than the
central unit located in Tehran, and all services are provided throughout the country only at this
center. Since Digikala is an online store and one of its goals is to reduce unnecessary travel, it is
not possible to deliver the order in the Digikala site. In Fig. 4, Digikala's network supply chain is
seen for four of its products.

Fig. 3. Digikala company supply chain

4.1. Data collection

The method for extracting the required data for the input parameters of the problem is presented
in Table 2 As shown in Fig. 4, the network under study consists of three echelons. The first
echelon is identified with suppliers. At this echelon, the retailer is contracted with the four
different suppliers as shown in Table 3.
Table 2. How to extract the data input parameters of the problem
Row Input parameters of the problem Extracted data
1 Dimensions of the vehicle chamber www.almascantin.ir
• Supplier product price
• Purchase rate for free shipping
• Headquarter experts and after-sales service of
• Product shipping Costs by third-party logistics
2 Digikala company
company
• www.digikala.com
• Product dimensions
• Customer satisfaction of the product
Coefficient = Total population of Iran × Customer
satisfaction of the product
3 Product demand
Demand of the province = Provincial population ×
Coefficient
Distance traveled by vehicle × Vehicle fuel
4 Product shipping cost by retailer distribution team
consumption × Price per liter of gasoline / diesel

14
 Average emissions of carbon dioxide per kilometer by
5 Ameri and Zahed (2013)
vehicles
The amount of carbon dioxide emissions in retailer
6 Rai et al. (2011)
company distribution warehouses
The number of job opportunities created by
7 Hezarkhani (2017)
establishing potential warehouse
8 Determine the distance between network components www.bahesab.ir/map/distance

Furthermore, in the second echelon, a warehouse was constructed in Tehran and 3 candidate
places are considered for establishing of the warehouse where the location of the warehouses is
shown in Table 4.
Table 3. Indicator number, location and supplier product i
Supplier indicator Supplier
Supplier location i
number i product i
1 Amin Hozur Crossing, Tehran Province Steamer
2 4 Baghe Bala street, Isfahan Province Suitcase
3 Basij Square, Zanjan Province Copper dish
4 Alborz Industrial Town, Qazvin Province Washing machine

Table 4. Indicator number and candidate points for establishing warehouse b

Warehouse
Warehouse location b
indicator number b
1 Shad Abad Street, Tehran province
2 Shokuhie Industrial Estate, Qom province
3 Shiraz Industrial Estate, Fars province
4 Fanavarihaye Bartar Industrial Estate, Razavi Khorasan province
Finally, the third echelon is related to customer, which is considered by 31 provinces as
customer. Table 5 indicates the indicator number and province name for the customer.

Table 5. Indicator number and province name for customer j

Customer indicator Customer indicator
Customer j Customer j
number i number i
1 East Azerbaijan 17 Fars
2 West Azerbaijan 18 Qazvin
3 Ardabil 19 Qom
4 Isfahan 20 Kurdestan
5 Alborz 21 Kerman
6 Ilam 22 Kermanshah
7 Bushehr 23 Kohgiluyeh and Boyer-Ahmad
8 Tehran 24 Golestan
9 Chahar Mahaal and Bakhtiari 25 Gilan
10 South Khorasan 26 Lorestan
11 Razavi Khorasan 27 Mazandaran
12 North Khorasan 28 Markazi
13 Khuzestan 29 Hormozgan
14 Zanjan 30 Hamadan
15 Semnan 31 Yazd
16 Sistan and Baluchistan

15
 Fig. 4. Digikala's supply chain for four products including steam, suitcase, copper dish and washing machine

4.2. Optimal amount of objective functions

To formulate the goal programming model, one must define the values of the goals for the
objective functions. To obtain the goal values, the mathematical model of the problem with each
objective function is considered separately, and the optimal value of each function is obtained.
The optimal value of each of the objective function is presented in the Table 6.

16
 Table 6. Optimal amount of objective functions
Objective Optimal amount
Economic ($) 20417400000
Environmental (gram/kilometer) 23030400000
Social (person) 1128999000

5. Sensitivity Analysis

In order to solve the model, using the data of Digikala, four constraints (54), (55), (56), and (57)
are added to the model.
&Ⅱ =0 ∀',j,e, i=4, e=1 (54)

The equation (54) prevents the flow of the product (washing machine) from the supplier i=4
(Qazvin) to the customer j, by the 3pl company e=1 (express mail). It is due to the point that the
Digikala company has a contract for the transportation of heavy products with shipping.
&Ⅰ =0 ∀',j,t, i=4, t=3 (55)

The equation (55) prevents the flow of the product (washing machine) from the supplier i=4
(Qazvin) to the customer j, by the retailer distribution team t=3 (motorcycle), because the Digikala
company has a contract for the transportation of heavy products with shipping and this product is
not portable with the motorcycle.

&Ⅴ = f2 &Ⅵ + 2 &Ⅶ + 2 &Ⅷ + 2 &Ⅲ + 2 &Ⅳ h=0

Xg XY
(56)
∀', , , i=4

The equation (56) prevents the flow of the product (washing machine) from the supplier i=4
(Qazvin) to the warehouse b and the customer j, by the 3pl company e=1 (express mail) and the
retailer distribution team t=3 (motorcycle). the Digikala company has a contract for the
transportation of heavy products with shipping and the product of the washing machine is not
portable with a motorcycle.
=1 ∀ , =1 (57)

Due to the fact that Digikala company has a warehouse in the Tehran province, therefore Y = 1.
Then, the decision making is done just for the construction of potential warehouses in Qom, Fars,
and Khorasan Razavi.

5.1. Sensitivity analysis of parameter of cost of shipping

In this section, changes to the objective function of the goal programming model and the
percentage of deviations are measured relative to the parameter of the cost of shipping. The results
of this measure are presented in Table 7 and Fig. 5.

17
 Table 7. Effect of parameter of cost of shipping on objective function of goal programming model
Objective function of Economic Environmental Social objective
Vehicle
goal programming objective function objective function function
purchase cost
model Percentage deviation
50% reduction 3.319 0.537 1.781 1.000
40% reduction 3.497 0.716 1.781 1.000
30% reduction 3.676 0.894 1.781 1.000
20% reduction 3.854 1.073 1.781 1.000
10% reduction 4.033 1.251 1.781 1.000
- 49611700000 64052000000 286
Normal status
4.211 1.430 1.781 1.000
10% increase 4.390 1.608 1.781 1.000
- 36732400000 80765000000 60
20% increase
4.540 1.033 2.507 1.000
- 28238600000 92553100000 126
30% increase
4.632 0.613 3.019 1.000
- 25326200000 96885400000 161
40% increase
4.700 0.494 3.207 1.000
- 23157100000 100311000000 193
50% increase
4.757 0.401 3.356 1.000

5
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
-50% -40% -30% -20% -10% 0% 10% 20% 30% 40% 50%

Goal programming model Economic objective function

Environmental objective function Social objective function

Fig.5. Effect of parameter of cost of shipping on objective function

Fig. 5 shows variations of the objective function of the goal programming model and the
percentage of the deviation relative to the transportation parameter. By reducing the transportation
cost parameter, only the value of the objective function of the goal programming model and
deviations percentage of the economic objective function is reduced. These changes do not occur
on the deviation percentage of environmental and social objective functions, which seems to be
due to the lack of change in the model's response and model's decision making. By increasing the
transportation cost parameter by 10%, the value of the objective function of the goal programming
model and the deviation percentage of the economic objective function increase. But this change is
not reflected in the deviation percentage of environmental and social functions. It is also noted that
with an increase of 20 to 50 percent of the transportation cost parameter, the value of the objective
function of the goal programming model and the deviation percentage of the environmental
objective function will be significantly increased. However, the deviation percentage of the
economic objective function decreases and there is no change in the deviation percentage of social
function.

18
5.2. Effect of the costs on the objective function of the goal programming model

In this section, the effect of the parameter of the costs of purchasing a vehicle, transporting and
constructing a warehouse are measured on the objective function of the ideal model.

5
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
-50% -40% -30% -20% -10% 0% 10% 20% 30% 40% 50%

Vehicle purchase cost Shipping cost

Warehouse construction cost

Fig. 6. The effect of the cost of vehicle purchase, transportation and warehouse construction on the objective function

In Fig. 6, it can be seen that with the increase of the three parameters of the costs of purchasing
a vehicle, transport, and warehouse construction, the trend of changes in the value of the objective
function of the goal programming model is also incremental. Changes to the parameter of the cost
of buying a vehicle make a slight effect on the objective function of the goal programming model
and there is less sensitivity to this parameter. It is also seen that the model is more sensitive to
changes in the cost of warehouse construction.

5.3. Sensitivity analysis of the parameter of the number of warehouses

In this section, the sensitivity of the objective function of the goal programming model and the
percentage of deviations is measured relative to the parameter of the number of warehouses. The
results of this measure are presented in Table 8 and Fig. 7.
Table 8. The effect of the parameter of the number of warehouses on objective function of goal programming model
Objective Economic Environmental Social
Number of Provinces selected by
function goal objective objective objective
warehouses model for warehouse
programming function function function
constructed construction
model Percentage deviation
- 46411700000 64052000000 209
1 Tehran
4.054 1.273 1.781 1.000
- 49611700000 64052000000 286
2 Tehran - Fars
4.211 1.430 1.781 1.000
- 53111700000 64052000000 370
3 Tehran - Qom - Fars
4.382 1.601 1.781 1.000
- 59111700000 64052000000 514 Tehran - Qom - Fars -
4
4.676 1.895 1.781 1.000 Razavi Khorasan

19
 5
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
1 2 3 4

Goal programming model Economic objective function

Environmental objective function Social objective function

Fig. 7. The effect of the parameter of the number of warehouses on the objective function

As it is shown in Fig. 7, if the number of constructed warehouses increases, the value of the
objective function of the goal programming model and the percentage of the deviation of the
economic objective function is increased and social and environmental objective functions are
constant. However, these changes do not occur in the deviation percentage of environmental
objective functions, which seems to be due to the lack of change in the model's response and
model's decision making.

5.4. Sensitivity analysis of the parameter of demand

In this section, sensitivity of the objective function of the goal programming model and
percentage of deviations is measured relative to the demand parameter . The results of this measure
are presented in Table 9 and Fig. 8.
Fig.8 shows variations of the objective function of the goal programming model and the
percentage of the deviation relative to the parameter of demand. By increasing the parameter of
demand, the value of the objective function of the goal programming model and deviation
percentage of the economic and environmental objective functions are increased.

0
-50% -40% -30% -20% -10% 0% 10% 20% 30% 40% 50%

Goal programming model Economic objective function

Environmental objective function Social objective function

Fig.8. The effect of the parameter of demand on the objective function

20
 Table 9. The effect of the parameter of demand on the objective function of goal programming model
Objective function Economic Environmental Social objective
Demand goal programming objective function objective function function
model Percentage deviation
- 30405900000 32026000000 242
50% reduction
1.880 0.489 0.391 1.000
- 34247000000 38431200000 251
40% reduction
2.417 0.677 0.740 1.000
- 38088200000 44836400000 260
30% reduction
2.895 0.865 1.030 1.000
- 41929400000 51241600000 269
20% reduction
3.279 1.054 1.225 1.000
- 45770500000 57646800000 277
10% reduction
3.745 1.242 1.503 1.000
- 49611700000 64052000000 286
Normal status
4.211 1.430 1.781 1.000
- 53452900000 70457200000 295
10% increase
4.677 1.618 2.059 1.000
- 57294100000 76862400000 304
20% increase
5.144 1.806 2.337 1.000
- 61135200000 83267600000 313
30% increase
5.610 1.994 2.616 1.000
- 64976400000 89672800000 322
40% increase
6.076 2.182 2.894 1.000
- 68817600000 96078000000 331
50% increase
6.542 2.371 3.172 1.000

In Fig. 8, Increasing demand will require more transportation to meet demand, leading to higher
costs and higher gas emissions. The slope of the graphs also shows that environmental issues need
more management and control. As a result, vehicles with less CO2 emissions can be used.
5.5. Sensitivity Analysis after Weighting the Objective Functions

In the previous section, the objective function of the goal programming model is presented in
the form (64):
!'/` = %aT + % T + % T (64)

different values, then the objective functions of the model can be represented as (65). If wb =
In this case, the objective functions have the same value. However, if the functions have

wj = wd , then the relations (64) and (65) are identical.

!'/` = Va . %aT + V . %T + V . % T (65)

In Table 10, different weights are considered to sensitivity analysis of the objective functions.
The numerical values of 0.8, 0.75, and 0.6 represent significant importance of objective functions,
the numerical values of 0.3 and 0.2 show moderate importance of objective functions and the
numerical value of 0.1 represents the low importance of objective functions. The analysis of Fig.9
and Table 10 indicates that with the increase in the weight of the economic objective function and
reduction in weight of the social objective function, the amount of objective function of the goal
programming model, the percentage of the deviation of the economic objective function and the

21
value of the social objective function decrease and the percentage of the deviation of the
environmental objective function increases.

4
3.528
3.5

2.5

2 1.781
1.43
1.5 1.277
1.177
1 1
1

0.5
0.03
0
Goal programming model Economic objective function Environmental objective Social objective function
function
Case2: Wc=0.1, We=0.3, Ws=0.6 Case7: Wc=0.6, We=0.3, Ws=0.1

Fig.9. Comparing the modes 2 and 7 in weighting the objective functions and its effect on the percentage of deviations and the
objective function of the goal programming model

Fig. 9 is presented for comparing the mode no. 2 with the mode no. 7 on weighting objective
functions. Fig. 10 is presented for comparing the mode 3 with the mode 5 on weighting objective
functions.

2
1.781 1.781
1.8
1.598
1.6 1.512
1.43 1.43
1.4
1.2
1 1
1
0.8
0.6
0.4
0.2
0
Goal programming model Economic objective function Environmental objective Social objective function
function

Case3: Wc=0.1, We=0.6, Ws=0.3 Case5: Wc=0.3, We=0.6, Ws=0.1

Fig. 10. Comparing the modes 3 and 5 in weighting the objective functions and its effect on the percentage of deviations and
the objective function of the goal programming model

Comparison modes 3 and 5 in Fig.10 and Table 10 indicate that with the increase in weight of
the economic objective function and reduction in the weight of the social objective function, the
amount of objective function of the goal programming model is increased. However, it seems that
because the model response is not changed, changes are not reflected in the deviation percentage
of economic, environmental, and social objective functions.
Fig. 11 is presented for comparison modes 4 and 6 on weighting objective functions.

22
 Table 10. Weighing the objective functions and its effect on the percentage of deviations and the objective function of the goal
programming model
Economic Environmental
Weight of Objective function Social objective
objective objective
Status objective goal programming function
function function
functions model
Va = 0.75
Percentage deviation

V = 0.05
- 20736900000 105140000000 266

V = 0.20
1
0.390 0.016 3.565 1.000
Va = 0.1
V = 0.3
- 49611700000 64052000000 286

V = 0.6
2
1.277 1.430 1.781 1.000
Va = 0.1
V = 0.6
- 49611700000 64052000000 286

V = 0.3
3
1.512 1.430 1.781 1.000
Va = 0.3
V = 0.1
- 20805400000 104873000000 256

V = 0.6
4
0.961 0.019 3.554 1.000
Va = 0.3
V = 0.6
- 49611700000 64052000000 286

V = 0.1
5
1.598 1.430 1.781 1.000
Va = 0.6
V = 0.1
- 20739300000 105114000000 262

V = 0.3
6
0.666 0.016 3.564 1.000
Va = 0.6
V = 0.3
- 21036800000 104282000000 244

V = 0.1
7
1.177 0.030 3.528 1.000
Va = 0.8
V = 0.1
- 20739300000 105114000000 262

V = 0.1
8
0.469 0.016 3.564 1.000
Va = 0.1
V = 0.8
- 49611700000 64052000000 286

V = 0.1
9
1.668 1.430 1.781 1.000
Va = 0.1
V = 0.1
- 49611700000 64052000000 286

V = 0.8
10
1.121 1.430 1.781 1.000

4
3.554 3.564
3.5

2.5

1.5
0.961 1 1
1 0.666
0.5
0.019 0.016
0
Goal programming model Economic objective function Environmental objective function Social objective function

Case4: Wc=0.3, We=0.1, Ws=0.6 Case6: Wc=0.6, We=0.1, Ws=0.3

Fig. 11. Comparing the modes 4 and 6 in weighting the objective functions and its effect on the percentage of deviations and
the objective function of the goal programming model

23
 Comparison modes 4 and 6 in Fig.11 and Table 10 indicate that with the increase in the weight
of the economic objective function and the reduction in the weight of the social objective function,
the amount of objective function of the goal programming model, the percentage of the deviation
of the economic objective function is decreased, the value of the social objective function is
constant and percentage of the deviation of the environmental objective function increases.

6. Conclusions and future research

In this research, a mix-integer programming model was developed for optimizing economic,
environmental and social objectives, which these goals include minimizing transportation costs,
purchasing a vehicle and building a warehouse, minimizing the amount of carbon dioxide released
by transport vehicles and building warehouses and maximizing the number of job opportunities.
The multi-objective model was solved using the goal programming technique and GAMS
software. In the following, sensitivity analysis was performed on the main parameters of the model
(number of warehouses, demand, vehicle purchase costs, transportation and warehouse
construction) and the change in the objective function and the percentage of the deviations of the
economic, environmental, and social objective functions were measured relative to those
parameters. Finally, each of the objective functions was weighed according to a consultation with
the logistics director of Digikala company, as well as the importance of economic, environmental
and social considerations, and the results and sensitivity analysis were presented in the form of
tables and charts. Computational results are reported based on real-world data analyses. The used
instance was consisted of four suppliers, three warehouses, and 31 provinces, which meant that the
integer linear programming could be solved optimally using CPLEX in a short CPU time. Solving
the proposed model results in the selection of optimal distribution channel of Digikala products
(including steamer, suitcase, copper dish and washing machine), the number of vehicles required,
and the number of job opportunities created by the product transportation. The consequences of
solving the model are concluded as:
• If the cost of shipping increases by 10% in the next year, it will not have much effect on the
total cost of the system, but if this cost increases by 50% in the next five years, the cost the
system will dramatically increase. Thus, managing this cost is very significant.
• The variation in the cost of a vehicle's purchase cost has little effect on the objective function
of the goal programming model, and there is less sensitivity to this parameter. This can
encourage management to invest in the purchase of a vehicle. As a result, when demand
increases, more vehicles are purchased to accelerate the delivery and customer satisfaction.
• If the Digikala company cares for economic considerations, the distribution of steamer product
for 29 provinces by the third-party logistics company (shipping and express mail) and for the
provinces of Tehran and Alborz by the distribution company t (pride vans), the distribution of
suitcase product For 31 provinces, by the third-party logistics company e (shipping and express
mail), copper dish product distribution for 30 provinces is provided by the third-party logistics
company (shipping and express mail) and Qazvin province by the retailer distribution team t
(pride van) and the distribution of the washing machine product for 31 provinces is done by the
retailer's distribution team t (pride van).
• If the Digikala company is concerned with environmental or social considerations, distribution
of products for 31 provinces will be done by the retailer distribution team t (pride van).

24
• Digikala company main motivation for using this model was to reduce system costs. The
distribution costs are reduced when the product reaches the customer directly from the supplier
by third-party logistics and the retailer distribution team.
In the future works, it can be occupying to model the uncertainty of real cases. Another extension
could be the inclusion of more real-life constraints like capacity limitations of warehouses and
transportation time. In this study, we examined non-deteriorating items. Formulating a model that
considers fresh food or deteriorating items would also be beneficial.

References

Akgün, İ., & Erdal, H. (2019). Solving an ammunition distribution network design problem using multi-objective
mathematical modeling, combined AHP-TOPSIS, and GIS. Computers & Industrial Engineering, 129, 512-528.
Akkerman, R., Wang, Y., & Grunow, M. (2009, July). MILP approaches to sustainable production and distribution of
meal elements. In 2009 International Conference on Computers & Industrial Engineering (pp. 973-978). IEEE.
Alptekinoğlu, A., & Tang, C. S. (2005). A model for analyzing multi-channel distribution systems. European Journal
of Operational Research, 163(3), 802-824.
Ameri, M., & Zahed, F. (2013). Estimating Global Warming External Costs Due to Road Transportation in Iran (Case
study: Expressways). Journal of Environmental Studies, 39(3), 45.
Amiri, A. (2006). Designing a distribution network in a supply chain system: Formulation and efficient solution
procedure. European journal of operational research, 171(2), 567-576.
Azad, N., & Davoudpour, H. (2013). Designing a stochastic distribution network model under risk. The International
Journal of Advanced Manufacturing Technology, 64(1-4), 23-40.
Barzinpour, F., & Taki, P. (2018). A dual-channel network design model in a green supply chain considering pricing
and transportation mode choice. Journal of Intelligent Manufacturing, 29(7), 1465-1483.
Bortolini, M., Faccio, M., Ferrari, E., Gamberi, M., & Pilati, F. (2016). Fresh food sustainable distribution: cost,
delivery time and carbon footprint three-objective optimization. Journal of Food Engineering, 174, 56-67.
Cintron, A., Ravindran, A. R., & Ventura, J. A. (2010). Multi-criteria mathematical model for designing the
distribution network of a consumer goods company. Computers & Industrial Engineering, 58(4), 584-593.
Deqqaq, H., & Abouabdellah, A. (2016). Design of a green distribution network with multiple transportation modes.
ARPN Journal Engineering Apply Science, 11(21), 12354-61.
Geoffrion, A. M., & Graves, G. W. (1974). Multicommodity distribution system design by Benders decomposition.
Management science, 20(5), 822-844.
Grant, D. B., & Banomyong, R. (2010). Design of closed-loop supply chain and product recovery management for
fast-moving consumer goods: The case of a single-use camera. Asia Pacific Journal of Marketing and Logistics,
22(2), 232-246.
Guerrero Campanur, A., Olivares-Benitez, E., Miranda, P. A., Perez-Loaiza, R. E., & Ablanedo-Rosas, J. H. (2018).
Design of a Logistics Nonlinear System for a Complex, Multiechelon, Supply Chain Network with Uncertain
Demands. Complexity, 2018.
Halat, K., & Hafezalkotob, A. (2017). Distribution Network Design with Distributors Advance Payment and Fuzzy
Commitment. Journal of Advanced Manufacturing Systems, 16(04), 339-356.
Hlyal, M., Ait Bassou, A., Soulhi, A., El Alami, J., & El Alami, N. (2015). Designing a distribution network using a
two level capacity location allocation problem: Formulation and efficient genetic algorithm resolution with an
application to a Moroccan retail company. Journal of Theoretical and Applied Information Technology, 72(2),
294-306.
Hezarkhani, B. (2017). Design of warehouse taking into account order picking processes, Master's thesis, University
of tehran.
Hulthén, K. (2007). Economizing in differentiated distribution networks: a transaction approach. European Business
Review, 19(6), 508-523.
Janatyan, N., Zandieh, M., Alem-Tabriz, A., & Rabieh, M. (2018). Designing Sustainable Distribution Network in
Pharmaceutical Supply Chain: A Case Study. International Journal of Supply and Operations Management, 5(2),
122-133.

25
Jayaraman, V. (1998). Transportation, facility location and inventory issues in distribution network design: An
investigation. International Journal of Operations & Production Management, 18(5), 471-494.
Jayaraman, V., & Ross, A. (2003). A simulated annealing methodology to distribution network design and
management. European Journal of Operational Research, 144(3), 629-645.
Ko, H. J., Ko, C. S., & Kim, T. (2006). A hybrid optimization/simulation approach for a distribution network design of
3PLS. Computers & Industrial Engineering, 50(4), 440-449.
Lee, D. H., Bian, W., & Dong, M. (2007). Multiproduct distribution network design of third-party logistics providers
with reverse logistics operations. Transportation Research Record, 2008(1), 26-33.
Lorentz, H., Yew Wong, C., & Hilmola, O. P. (2007). Emerging distribution systems in Central and Eastern Europe:
Implications from two case studies. International Journal of Physical Distribution & Logistics Management,
37(8), 670-697.
Mangiaracina, R., Song, G., & Perego, A. (2015). Distribution network design: a literature review and a research
agenda. International Journal of Physical Distribution & Logistics Management, 45(5), 506-531.
Millet, D. (2011). Designing a sustainable reverse logistics channel: the 18 generic structures framework. Journal of
Cleaner Production, 19(6-7), 588-597.
Mogale, D. G., Cheikhrouhou, N., & Tiwari, M. K. (2019). Modelling of sustainable food grain supply chain
distribution system: a bi-objective approach. International Journal of Production Research, 1-24.
Ouhimmou, M., Nourelfath, M., Bouchard, M., & Bricha, N. (2019). Design of robust distribution network under
demand uncertainty: a case study in the pulp and paper. International Journal of Production Economics, 218, 96-
105.
Rai, D., Sodagar, B., Fieldson, R., & Hu, X. (2011). Assessment of CO2 emissions reduction in a distribution
warehouse. Energy, 36(4), 2271-2277.
Ross, A., & Jayaraman, V. (2008). An evaluation of new heuristics for the location of cross-docks distribution centers
in supply chain network design. Computers & Industrial Engineering, 55(1), 64-79.
Sadjady, H., & Davoudpour, H. (2012). Two-echelon, multi-commodity supply chain network design with mode
selection, lead-times and inventory costs. Computers & Operations Research, 39(7), 1345-1354.
Samani, M. R. G., Hosseini-Motlagh, S. M., & Ghannadpour, S. F. (2019). A multilateral perspective towards blood
network design in an uncertain environment: Methodology and implementation. Computers & Industrial
Engineering, 130, 450-471.
Selim, H., & Ozkarahan, I. (2008). A supply chain distribution network design model: an interactive fuzzy goal
programming-based solution approach. The International Journal of Advanced Manufacturing Technology, 36(3-
4), 401-418.
Shu, J., Wang, G., & Zhang, K. (2013). Logistics distribution network design with two commodity categories. Journal
of the Operational Research Society, 64(9), 1400-1408.
Validi, S., Bhattacharya, A., & Byrne, P. J. (2018). Sustainable distribution system design: a two-phase DoE-guided
meta-heuristic solution approach for a three-echelon bi-objective AHP-integrated location-routing model. Annals
of Operations Research, 1-32.
Yadav, V. S., Tripathi, S., & Singh, A. R. (2018). Bi-objective optimization for sustainable supply chain network
design in omnichannel. Journal of Manufacturing Technology Management.
Zhang, S., Lee, C. K. M., Wu, K., & Choy, K. L. (2016). Multi-objective optimization for sustainable supply chain
network design considering multiple distribution channels. Expert Systems with Applications, 65, 87-99.

26
Declaration of interests

☒ The authors declare that they have no known competing financial interests or personal relationships
that could have appeared to influence the work reported in this paper.

☐The authors declare the following financial interests/personal relationships which may be considered
as potential competing interests: