MGMT90141

Business Analysis and Decision Making

Semester 1, 2022

Amazon Transshipment Optimization Problem

Instructor Prof. William Ho

Seminar Monday 3:15pm

Team Members and Student IDs Unnati Agarwal (1408690)

Divna Vinod Karkera (1410641)
Zenique Wee (1166489)
Jianing Sun (1298694)
Maria Gracia Ramon (1382906)

Word Count 2184

 Table of Contents

1. Introduction 2

2. Evaluation Criteria 4

3. Methodology 6

4. Literature Review

4.1 Strengths 7

4.2 Weaknesses 10

5. Discussions and Conclusions 12

References 13

Appendices 16

1
1. Introduction

A well-oiled supply chain is vital to any business, primarily an e-commerce business.

With the growing number of online purchases and consumer acclimatization to same-day

delivery, the need for an efficient inventory management system with optimized transshipment

routes has become imperative.

Amazon ranks as one of the world’s top companies by market value. As of January 27,

2023, Amazon has a market capitalization of $1 trillion with 15,41,000 employees worldwide

(Reiff, 2021). Founded by Jeffrey Bezos in 1994 from a garage in Washington, Amazon was

designed originally as an online marketplace for selling books. Quickly, the environment

expanded to offer a wider range of products, including video games, electronics, furniture,

food, and more (Ecommerce Platforms, 2018).

About 44% of customers start their search for products with Amazon, and 40% buy

products on the website at least once per month (Ecommerce Platforms, 2018). Amazon’s rise

has been relentless for more than a decade, with regulators and rivals alike failing to slow it.

But now Amazon has taken a hit from a pair of immovable forces: snarled supply chains and

worker shortages (Business, 2021).

The company’s third-quarter earnings and sales fell well short of Wall Street’s

projections, in terms of both revenue and profit. The company posted a net income of $2.9

billion in the third quarter (Q3) of its 2022 fiscal year (FY), a decrease from $3.2 billion in Q3

2021 (Reiff, 2021).

The biggest issue Amazon is facing is not the lack of consumer demand or the slowing

economy, as both remain strong, rather it’s the fact that Amazon is unable to deliver all of its

products to its customers quickly enough, thereby staring down the barrel of a supply chain

nightmare. In addition to this, the inflation on raw materials, labour, trucking costs and the

staffing shortages at some warehouses forced Amazon to reroute products to other facilities

2
that were fully staffed but less convenient, resulting in less optimal placement, which gave rise

to longer and more expensive transportation routes with high fulfilment costs (Business, 2021).

The third-party sellers contribute 55% of all sales made on Amazon (Bauer, 2021).

There are two ways in which the sellers can sell through Amazon: i) Fulfillment by Amazon

(FBA), and ii) Fulfillment by Merchant (FBM). With Amazon providing the benefit of same

day or two-day window delivery, it is essential that it recognises its most efficient mode of

transport and maximizes delivery speed. Amazon has also introduced facilities such as Amazon

Trucking, Amazon Warehousing and Amazon Prime Air to better serve their customers and

provide fast delivery services. In future, they plan to introduce Drone delivery services in order

to expand their delivery network. For this assignment particularly, we focus on the Fulfillment

by Amazon (FBA) method, as represented in Figure 1 below.

Figure 1: Amazon Supply Chain Diagram

(How the Amazon Supply Chain Works: Strategy & Management, 2022)

3
 Therefore, Amazon’s optimization problem is defined as: To determine the most

efficient transshipment process of products from the suppliers, to the warehouses and finally

to the consumers, so as to minimize the cost and keep up with the consumer demand, whilst

simultaneously maximizing the customer satisfaction.

2. Evaluation Criteria

Objective function: Minimizing Total Distribution Cost

Decision variables Description

Shipment quantity The quantities good x shipped between

warehouse i and distribution centre j

Unit cost of inventory The average cost c of each item that exists in the

warehouse i

Shipment time The time t it takes to get from warehouse i to

distribution centre j

Shipment route Route of product x shipped from Warehouse i to

its destination d

Distribution location The location of distribution centre j

4
Mode of transportation Using transportation mode s to deliver goods

Placing order time The time t it takes to place the order for

customers

Inventory at each warehouse The quantities that each good (i.e product x, y, z)

stored in the Warehouse i

Constraint Description

Capacity Constraint The maximum number of products that

warehouses stored should be at least meeting

customer’s demand

Inventory Budget Constraint The cost of containing and producing inventory

should not exceeds the value of the goods

Supply Capacity The Maximum amount that sellers provided to

the warehouse

Demand Constraint The required amount of inventory should not be

less than quantities demanded

Shipment Budget Constraint The amount of money that is available to

transfer goods

5
 Time constraint The schedule for completion of the delivery

Production Capacity The Maximum amount of goods can produce

Non - negative Constraint The quantities of goods produced, stored and

shipped should not be negative

Capacity constraint of vehicle The maximum amount of vehicle available for

delivery during certain time period

Distribution centre capacity The maximum quantities that distribution centre

could store

Maximum length of route The maximum length of route that the vehicle

can travel for delivery

3. Methodology

A number of journal articles have been reviewed in order to learn more about the

transshipment issue and supply-chain management in Amazon's supply chain. The information

gathered from these journal articles will assist in enhancing how we tackle the optimization

problem. The purpose of the literature review is to identify concepts and findings from the

journal articles that have examined and optimized related transshipment issues that might be

relevant to the Amazon enterprise.

Through reliable databases such as Google Scholar, ScienceDirect, and mainly Scopus,

we were able to find the relevant journal articles. Keywords such as ‘transshipment’,

6
‘transshipment optimization’, ‘supply chain optimization’, and ‘transshipment linear

programming model’ were entered into the search engine to obtain the relevant 10 articles that

could be applied to our current research. The articles were then filtered with the time period of

the past 10 years in order to maintain the relevancy of the data.

4. Literature Review

4.1 Strengths

Strength 1: Factors being relevant to the optimization problem faced by Amazon

The articles reviewed could be applied to the optimization problem that Amazon

currently faces. For instance, Odeyinka et al. (2022) emphasised on the importance of having

an efficient supply chain management system. This article is relevant to the issue of the snarled

supply chain that Amazon is currently dealing with. Furthermore, Alnaggar et al. (2020)

tackled the difficulties faced in distribution planning in a transshipment model through an

outsourced third-party logistics service (3PL) that utilises consolidation facilities to maximize

savings and takes into account the consumer’s logistical needs. This is applicable to Amazon’s

current optimization problem of meeting high customer demands.

Additionally, Lefever et al. (2018) and Peres (2017) both examined the Inventory-

Routing Problem with Transshipment (IRPT) by establishing the framework for decision-

making. It considers demand uncertainty from consumers, inventory level, transportation cost,

and transshipment costs. Furthermore, it ensures that transshipments between suppliers,

warehouses and consumers are feasible while minimizing the total cost. Amazon is currently

facing a problem of longer and more costly transportation routes with highly dynamic customer

demand, due to suboptimal routing of products to faraway warehouses, on the grounds of

staffing shortages (Business, 2021).

7
 Moreover, Fallah-Tafti et al. (2022) focussed on rapid transit networks such as

subways, metro, or railway systems. As Amazon offers a one-day delivery window for their

customers, this would be relevant towards Amazon’s aim of achieving efficient modes of

transport, while maximizing delivery speed in order to meet customers demands.

Therefore, the factors taken into consideration by the above articles, are relevant to

Amazon’s optimization problem of obtaining transshipment of products from suppliers to the

warehouse, and then to consumers, within a stipulated time period, as a means to maximize

customer satisfaction whilst minimizing cost.

Strength 2: Applicable to realistic scenarios

Nasiri et al. (2015) considered several real-world restrictions such as incorporating

demand uncertainty and variability in their model which is applicable to real-world situations

wherein consumer demand is always changing. Odeyi and Senicheva (2018) employed forward

approximate dynamic programming that is effective in dealing with problem instances that

could arise in realistic contexts like overstocking and/or understocking of inventory due to

dynamic changes in the customer demand. Alnaggar et al. (2020) created a model whereby it

focused on assessing unpredictability such as uncertainty encountered for the transportation of

products, by analysing the problem and considering realistic possibilities that may arise. For

example, arranging transportation between suppliers and customers when customer demand is

speculative, their model selects the optimal transportation option for its customer by

determining delivery schedule and arranges for pickup at the supplier’s once shipments are

produced. Hence, the application to realistic scenarios would make the proposed model more

feasible for Amazon.

Strength 3: Utilised varied mathematical modelling problems and approaches

8
 Findings from studies and realistic approaches from the articles revealed that there are

several ways to solve the transshipment problem by applying different methods and

mathematical modelling. The mathematical modelling used by Liu et al. (2022) provided

solutions to the problem of ensuring customer satisfaction through prompt product delivery, a

problem that is extremely relevant to Amazon, hence was rightly used as an example in their

study as well.

Moreover, the dynamic programming approach applied by Odeyi and Senicheva (2018)

proposed lateral transshipment such as the transportation of products between locations of the

same echelon. This showed how to prevent shortage and overstocking of inventory in facilities,

due to the uncertain demand of customers, resulting in optimized transshipment routes and an

efficient inventory management system which could potentially solve one of Amazon’s gravest

problems.

Strength 4: Provided related and additional constraints

Most of the journal articles provided constraints that can be used to solve Amazon’s

optimization problem. To illustrate this point, Fallah-Tafti et al. (2022) research showed that

companies should also consider the maximum length of the route to maintain efficient

transshipment of the products, rather than only depending on the capacity constraint of the

vehicle transporting the products. Additionally, Liu et al. (2022) has also taken into account

the constraint related to the length of route in the model that is applicable to e-retailers such as

Amazon. Therefore, Amazon’s planned model should include the maximum length of the route

as a constraint.

4.2. Weaknesses:

9
Weakness 1: Unrealistic assumptions and overgeneralized models

A generic study of the variables has been done, by assuming that they will remain

constant. This is unrealistic, which leads to oversimplification of the model, thereby failing to

account for all key elements such as external uncertainty, player cooperation, and transit costs,

leading to erroneous outcomes. An example of this is the assumption made by (Liu et al., 2022)

wherein the author assumed factors such as vehicle capacity, number of orders, and delivery

route to be constant.

Moreover, uncertainty is not taken into consideration by Crainic et al. (2021) since each

product's demand is assumed to be predictable and to have a known priority, while the

supplier's lead time for shipment is believed to be static. Additionally, even Fallah-Tafti et al.

(2022) assumed that variables such as travel time and demand will be constant.

Weakness 2: Incomplete Scope

Some of the articles do not cover key components of the transshipment issue, resulting

in unforeseen challenges or gaps in the conclusion. Odeyinka et al. (2022) excluded the external

obstacles related to transportation, which might invariably delay the supply chain and affect

consumer demand. It only focused on a specific model for a 3-tier supply chain with horizontal

cooperation.

Moreover, the mathematical model developed by Liu et al. (2022) is essentially based

on pre-established contexts that prevent it from being applicable to Amazon beyond the specific

scenarios explored in it since the model is specifically developed for optimizing vehicle routing

within a single city taking into consideration parameters such as client location, vehicle

capacity, and delivery time windows. However, Liu et al. (2022) has also suggested that since

the model may not be immediately relevant to other cities or areas, some modifications must

10
be made for it to be applicable to all places including additional constraints or variables such

as local rules or traffic patterns.

Weakness 3: Limited Validation and Comparison

The inability to compare the suggested methodology to current models restricts our

capacity to analyse the relative strengths and weaknesses of the recommended approach,

thereby causing uncertainty with regards to how it will perform under other scenarios, leading

to a lack of trust in the validity and generalizability of the results. To assure the correctness and

dependability of a model, this must be rigorously validated on several examples and compared

to a wide range of current techniques and benchmarks. For instance, Fallah-Tafti et al. (2022)

compared their suggested algorithms solely to two additional meta-heuristic algorithms,

therefore it is not clear whether other techniques were explored or how these algorithms

performed in relation to other methods. In this same article the inadequate external verification

and case study analysis undermines the trust in the findings and conclusions.

This weakness is validated by Odeyinka et al. (2022) suggestion that the benchmark

applied might not fully reflect all conceivable circumstances, and hence the model’s

generalizability must be confirmed by additional testing.

Weakness 4: Implementation and Practicality Challenges

Implementation problems refer to hurdles that might develop while attempting to apply

a certain approach in real-world settings. Several articles include incomplete mathematical

models which require more accurate answers. For instance, Nasiri et al. (2015) single-echelon

distribution model may be insufficient in order to suit the demands of Amazon. This company

operates on a global level and, to manage its supply chain, implements a multi-echelon

distribution model with numerous tiers of intermediaries such as fulfillment and sorting centers,

11
and delivery stations. Therefore, it seems improbable that a single-echelon model would be

enough to simulate Amazon’s distribution network given the scope of their operations.

Moreover, Fallah-Tafti et al. (2022) includes very little debate regarding implementation

concerns since they prioritize the mathematical framework of the model rather than the real-

world issues, such as lack of emphasis on practical factors, namely traffic congestion and the

effect of transportation alternatives. Finally, Liu et al. (2022) article can be considered

extremely cumbersome, since an entirely different algorithm was explained prior to

incorporating MOGWO and NSGS-II extensive techniques to arrive at solutions, making the

model unnecessarily tedious and impractical.

5. Discussions and Conclusions

To conclude, we would like to emphasize on the growing dependability of e-commerce

and online shopping, owing to technological advancement. This makes it imperative for e-

commerce companies to have an extremely efficient transshipment framework, without

compromising on customer satisfaction while also achieving the objective of minimizing

transportation costs.

In this assignment, we have evaluated ten journal articles with each article providing us

with different insights and perspectives. Firstly, from the decision-making variables and

constraints such as shipment route, inventory storage, distribution center capacity, etc. we can

deduce that it makes the mathematical model more exhaustive to the issues faced by Amazon

today. However, the articles also rendered some weaknesses such as limited validation and

overgeneralised models based on unrealistic assumptions.

Our suggestion to revamp these drawbacks would be to come up with more

comprehensive mathematical models that specifically takes into account the e-commerce

12
industries and online retailers. Secondly, to overcome the uncertainty faced in real-life

distribution networks and change in consumer behavior post technological advancements, we

suggest enabling and practicing lateral transshipment in order to keep up with the customer

demand. Lastly, in addition to the vehicle capacity constraint, Amazon’s model must also

incorporate the maximum length of route as one of their key constraints.

References

Ecommerce Platforms. (2018). What is Amazon? Ecommerce Platforms; EP.

https://ecommerce-platforms.com/glossary/amazon

How The Amazon Supply Chain Works: Strategy & Management. (2022, October 21).

https://profitwhales.com/archives/articles/amazon-supply-chain-strategy

Banker, S. (n.d.). Amazon Supply Chain Innovation Continues. Forbes. Retrieved

March 26, 2023, from https://www.forbes.com/sites/stevebanker/2021/04/01/amazon-

supply-chain-innovation-continues/?sh=62de04c777e6

Odeyinka, O. F., Okandeji, A. A., & Ogunwolu, F. O. (2022). Mathematical modeling

of inventory cost in a 3-tier supply chain with horizontal cooperation. Scientific

African, 16, e01164. https://doi.org/10.1016/j.sciaf.2022.e01164

Odeyi Meissner, J., & Senicheva, O. V. (2018). Approximate dynamic programming

for lateral transshipment problems in multi-location inventory systems. European

Journal of Operational Research, 265(1), 49–64.

https://doi.org/10.1016/j.ejor.2017.06.049

Alnaggar, A., Gzara, F., & Bookbinder, J. H. (2020). Distribution planning with random

demand and recourse in a transshipment network. EURO Journal on Transportation

and Logistics, 9(1), 100007. https://doi.org/10.1016/j.ejtl.2020.100007

13
Liu, P., Hendalianpour, A., Feylizadeh, M., & Pedrycz, W. (2022). Mathematical

modeling of Vehicle Routing Problem in Omni-Channel retailing. Applied Soft

Computing, 131, 109791. https://doi.org/10.1016/j.asoc.2022.109791

Lefever, W., Aghezzaf, E.-H., Hadj-Hamou, K., & Penz, B. (2018). Analysis of an

improved branch-and-cut formulation for the Inventory-Routing Problem with

Transshipment. Computers & Operations Research, 98, 137–148.

https://doi.org/10.1016/j.cor.2018.05.023

Crainic, T. G., Giusti, R., Manerba, D., & Tadei, R. (2021). The Synchronized

Location-Transshipment Problem. Transportation Research Procedia, 52, 43–50.

https://doi.org/10.1016/j.trpro.2021.01.007.

Nasiri, G. R., Ghaffari, N., & Davoudpour, H. (2015). Location-inventory and

shipment decisions in an integrated distribution system: an efficient heuristic solution.

European J. Of Industrial Engineering, 9(5), 613.

https://doi.org/10.1504/ejie.2015.071779

Fallah-Tafti, M., Honarvar, M., Tavakkoli-Moghaddam, R., & Sadegheih, A. (2022).

Mathematical modeling of a bi-objective hub location-routing problem for rapid transit

networks. RAIRO - Operations Research, 56(5), 3733–3763.

https://doi.org/10.1051/ro/2022170

Peres, I. T., Repolho, H. M., Martinelli, R., & Monteiro, N. J. (2017). Optimization in

inventory-routing problem with planned transshipment: A case study in the retail

industry. International Journal of Production Economics, 193, 748-756.

https://doi.org/10.1016/j.ijpe.2017.09.002

Rautela, H., Janjevic, M., & Winkenbach, M. (2021). Investigating the financial

impact of collection-and-delivery points in last-mile E-commerce distribution.

14
Research in Transportation Business & Management, 100681.

https://doi.org/10.1016/j.rtbm.2021.100681

Reiff, N. (2021, February 5). How Amazon Makes Money: Cloud Services Takes Off.

Investopedia. https://www.investopedia.com/how-amazon-makes-money-4587523

Bauer, M. (2021, February 23). How Amazon Supply Chain Works. Teikametrics.

https://www.teikametrics.com/blog/how-amazon-supply-chain-works/

Business, N. M., CNN. (2021, October 29). Supply chain nightmares are doing what

regulators and rivals can’t: Slow Amazon down. CNN.

https://edition.cnn.com/2021/10/29/business/amazon-supply-chain-holiday-

shopping/index.html

15
Appendices

Appendix 1. Summary, strengths, and weaknesses of each journal article

Journal Article 1 Odeyinka, O. F., Okandeji, A. A., & Ogunwolu, F. O. (2022).

Mathematical modeling of inventory cost in a 3-tier supply chain with

horizontal cooperation. Scientific African, 16, e01164.

https://doi.org/10.1016/j.sciaf.2022.e01164

Model Linear mixed integer constrained single objective mathematical

program

Summary This research paper highlights the by focussing on horizontal

cooperation, which is a type of collaboration that exists between

companies operating at the same level of the market.

The main objective of this research paper is to minimize the inventory

cost across a 3-tier supply chain (supplier, manufacturer, and

distributor) with horizontal cooperation at the post-production

inventory end of the manufacturing tier. The paper presents a test

problem and compares the results of solving the mathematical model

with and without horizontal cooperation scenarios to show the benefits

of cooperation among supply chain players.

The deterministic decision variables were quantity of materials to be

supplied, which included, quantity of product types to be produced,

quantity of product types not to be produced but to be invented on

16
 behalf of another factory for horizontal cooperation, and quantity of

products to be distributed.

The parameters used in the Supply tier entailed storage capacity

available to the supplier, base level materials and inventory cost per

unit of base level of materials. Whereas, the Manufacturing/

Production tier parameters were material storage capacity and product

storage capacity.

Strengths ● The paper considers horizontal cooperation among supply

chain players at the post-production inventory end of the

manufacturing tier, which is a realistic and relevant scenario in

many supply chains

● The paper provides a test problem and presents the results of

solving the mathematical model with and without horizontal

cooperation scenarios, which shows the benefits of cooperation

among supply chain players

● The presence of charts, and diagrams explaining in detail the 3-

tier supply chain model made understanding the mathematical

model a lot easier

Weaknesses ● The paper only considers horizontal cooperation at the post-

production inventory end of the manufacturing tier, and other

forms of cooperation or coordination among supply chain

players are not explored

● The paper does not consider the impact of external factors such

as demand uncertainty, lead time variability, or transportation

17
 costs, which could affect the inventory cost and the

performance of the supply chain

● The test problem used to validate the model may not be

representative of all possible scenarios, and the results may

vary for different supply chain configurations or product types

Journal Article 2 Odeyi Meissner, J., & Senicheva, O. V. (2018). Approximate dynamic

programming for lateral transshipment problems in multi-location

inventory systems. European Journal of Operational Research, 265(1),

49–64.

https://doi.org/10.1016/j.ejor.2017.06.049

Model Linear mixed integer constrained single objective mathematical

program

Summary This paper examines how companies can optimize their inventory

allocation across multiple locations to avoid excess stock in some

locations and stockouts in others, occurring due to random fluctuations

in consumer demand.

To alleviate the mismatch between an actual consumer demand and an

available stock in numerous locations under the incapacity to replenish

from a central warehouse, the authors propose lateral transshipment,

e.g., the movement of stock between locations of the same echelon.

The paper proposes a dynamic programming approach to find an

optimal policy that will maximize the expected sum of profits over all

18
 time periods. However, due to the large size of the state and decision

spaces, the paper uses forward approximate dynamic programming to

find a near-optimal transshipment policy. The paper conducts an

extensive numerical study to compare the proposed algorithm to state-

of-the-art methods in the literature.

The decision variables taken into consideration were transshipment

from location i to location j at the beginning of the period t, vector of

transshipments from location i to time t, and transshipment decision in

the system at time t.

The parameters were initial order-up-to level of the inventory of the

location i, initial state of the system at time t, holding cost per unit in

location i, cost of transshipment per unit per distance, and distance

between location i and location j.

Strengths ● Numerous articles consider the transshipment as a single time

event per order cycle, however this journal article formulates a

model of a period review where each demand triggers a

transshipment decision

● The paper uses forward approximate dynamic programming to

find a near-optimal transshipment policy, which is

computationally efficient and can handle real-world sized

problem instances

● The paper conducts an extensive numerical study to compare

the proposed algorithm to state-of-the-art methods in the

19
 literature, which shows that the proposed algorithm performs

very well when used in real-world scenarios

● The authors have backed their problem statement of over

stocking and/or understocking of products due to demand

mismatch, by multiple factual case studies gathered over a

comprehensive time period

Weaknesses ● The paper assumes that the demand and supply processes are

stationary and independent across locations, which may not

hold in all real-world scenarios. For example, demand may be

correlated across locations due to external factors such as

weather conditions or marketing campaigns.

● The paper does not consider the cost of transshipment

explicitly, which may be significant in some real-world

scenarios. For example, the cost of transportation, handling,

and storage may vary across locations and impact the

profitability of the network.

Journal Article 3 Alnaggar, A., Gzara, F., & Bookbinder, J. H. (2020). Distribution

planning with random demand and recourse in a transshipment

network. EURO Journal on Transportation and Logistics, 9(1),

100007.

https://doi.org/10.1016/j.ejtl.2020.100007

Model Sample Average Approximation (SAA) and Stochastic Distribution

Planning with Consolidation - path based formulation (SDPC-PF)

20
Summary This paper talks about the problems in distribution planning in a

transshipment model by introducing a stochastic customer demand. In

this model, the authors study the problem in view of an outsourced

third-party logistics provider (3PL) who uses consolidation premises

to maximize savings and takes care of all the logistics needs of the

customers. A spot market carrier has also been introduced to meet the

excess customer demands.

The two main objectives of the programming are to a) minimize the

transportation cost, and b) minimize the inventory holding cost and

cost of shipping products via spot market carrier.

The decision-making variables are expressed in binary variables -

inbound transportation is reserved for supplier, outbound

transportation is reserved for customer, shipment is transported from

supplier to CC, shipment transported from CC to customer, shipment

transported from CC to customer by a spot market carrier, shipment

transported from CC to customer by spot market carrier.

The parameters used in the model are set of customers, set of suppliers,

demand realizations, capacity of inbound & outbound transportation,

dispatch time, inbound & outbound spot market carrier cost, holding

cost, inbound & outbound transportation cost of transportation.

Strengths ● The authors have bridged a gap between the haphazardness of

the distribution network and the demand uncertainty.

21
 ● Generally, the uncertainty faced in real-life while distributing

the products is not considered and is overlooked; however this

article specifically focuses on creating a model that assesses

unpredictability.

● Multiple possibilities and relevant situations were looked at

while evaluating the problem to exactly achieve economies of

scale making it relevant to global distribution channels with

different types of transports.

Weaknesses ● A lot of variables affecting the problem are kept fixed in the

entire model, for example: transportation route, capacity of the

vehicle.

● Numerous assumptions are made throughout the calculations

(for example: the 3PL does not operate its own vehicle) to

achieve a certain solution.

● The article first uses the sample average method to get a

solution and then applies it to the linear programming model

for the final solution, making it a lengthy process.

Journal Article 4 Liu, P., Hendalianpour, A., Feylizadeh, M., & Pedrycz, W. (2022).

Mathematical modeling of Vehicle Routing Problem in Omni-Channel

retailing. Applied Soft Computing, 131, 109791.

https://doi.org/10.1016/j.asoc.2022.109791

Model Mixed-Integer Programming with Binary Variables

22
Summary This article uses mathematical modeling to provide solutions to one of

the biggest challenges faced by e-retailers today, that is: customer

satisfaction through fast delivery of products. The authors have taken

4 different points of delivery, and two algorithms into consideration.

Through linear programming they have derived a pareto optimal point

wherein the costs of delivery are low and the customer satisfaction is

higher.

There are two main objective functions in this article:

i) to minimize the cost associated with the distribution network, and

ii) to maximize the customer benefits by providing the fastest and

shortest delivery time via the four distribution networks (that is: retail

store, Points of Delivery (POD), Intermediate Depots (IDs), and City

Distribution Center (CDC))

The parameters for this particular model are capacity of vehicle, POD,

CDC, ID and retail stores, the demand of the customer, inventory of

the product in all the four distribution system, the distance from one

distribution point to another for all the four systems, the shipment,

shipment cost, maximum length of route, maximum acceptable delay,

time needed to prepare the product, and customer benefits.

The decision variables for this model are binary and are as follows:

delivery made from one point to another via a given vehicle and time

23
 period, delivery starting time, total product shipped, and time of

product delivery.

Strengths ● The article uses simple language along with a proper

explanation of every algorithm, key terms, and objective.

● Continuous examples of large stores such as Amazon and

Walmart are used throughout the article to make the model

more relevant.

● This journal considers both the methods of product shipment,

that is a) physical, and b) online sales

● The authors have used a good mix of graphs, examples, and

tables in the model. They have incorporated 2 algorithms,

compared and contrasted them to find the best optimal point.

Weaknesses ● The set up laid in the model considers a distribution channel of

only one city, hence it may not be applicable to all the places.

● An entirely different algorithm Grey Wolf Optimizer was

explained prior to incorporating MOGWO and NSGS-II,

making the article very descriptive.

Journal Article 5 Lefever, W., Aghezzaf, E.-H., Hadj-Hamou, K., & Penz, B. (2018).

Analysis of an improved branch-and-cut formulation for the Inventory-

Routing Problem with Transshipment. Computers & Operations

Research, 98, 137–148. https://doi.org/10.1016/j.cor.2018.05.023

Model Branch-and-Cut (B&C)

Summary This article investigated the B&C formulation of the Inventory-

Routing Problem with Transshipment (IRPT) whereby transshipments

24
 between retailers or between the supplier and a retailer are also

feasible, in addition to the supplier delivering products to the retailers.

The purpose is to identify the quantities supplied to all retailers by the

supplier and subcontractor during the various times of the planning

period while minimizing the total cost.

Strengths ● This article showed that IRPT allows enterprises to build more

flexible models. Such that although retailers with an unclear

demand profile may be excluded from the supplier’s fixed

schedule, they are still able to receive service through

transshipments from other retailers or from the warehouse

when their demand has been made clear.

Weaknesses ● Not much studies have been conducted on IRPT, hence, it may

be unreliable and not applicable to the problems faced in

Amazon.

Journal Article 6 Crainic, T. G., Giusti, R., Manerba, D., & Tadei, R. (2021). The

Synchronized Location-Transshipment Problem. Transportation

Research Procedia, 52, 43–50.

https://doi.org/10.1016/j.trpro.2021.01.007.

Model Mixed integer linear programming (MILP)

Summary In order to manage a number of possible situations based on various

sizes of the transshipment facilities and locations, this article suggested

a formulation for the issue based on mixed-integer linear programming

25
 and conducted an economic analysis to investigate the Synchronized

Location-Transshipment Problem.

This paper presented the need for the cross-docking paradigm to

minimize leftovers and ensure the precise fulfilment of the demand

needed for the just-in-time paradigm.

Additionally, it showed that the trade-off between the various test

cases' objective functions demonstrates that purchasing smaller

facilities as compared to hiring larger ones may be more practical.

Strengths ● Provided a realistic context in terms of synchromodal

operations which showed that in regards to transportation costs,

having additional facilities can help to reduce costs.

Weaknesses ● This article did not take into account the demand uncertainty.

It would be relevant for real-world applications where demand

is uncertain and changeable.

● Assumed that demand is deterministic, and as demand from

consumers is uncertain. We are not able to relate this to real life

situations.

Journal Article 7 Nasiri, G. R., Ghaffari, N., & Davoudpour, H. (2015). Location-

inventory and shipment decisions in an integrated distribution system:

an efficient heuristic solution. European J. Of Industrial Engineering,

9(5), 613. https://doi.org/10.1504/ejie.2015.071779

26
Model Mixed-integer linear programming (MILP)

Summary The article offers an effective heuristic method in order to make

location-inventory and shipment decisions in an integrated distribution

system. The suggested strategy is founded on a mixed-integer linear

programming (MILP) model that takes into account a number of real-

world restrictions such restricted capacity, time limitations, and

numerous means of transportation. The authors also present a greedy

approach that can quickly locate solutions that are close to optimal. The

results demonstrate that the suggested strategy surpasses other current

approaches in terms of solution quality and computational efficiency

when evaluated on a set of examples and compared with them. The

essay offers a workable solution for a distribution system's location-

inventory and shipment selections overall.

Strengths ● The use of a heuristic approach or heuristic algorithm: the

article suggests this algorithm in order to optimize location

inventory and shipment decisions, which is an effective

solution method to handle large-scale problems which are

difficult to solve using exact methods.

● The integrated approach to location inventory and shipment

decisions also provides a comprehensive solution to the

problem since it can capture the interactions between these two

decisions.

● Incorporation of uncertainty and demand variability: this is

relevant for real-world applications where demand is uncertain

and variable.

27
Weaknesses ● Complexity of the multi-echelon distribution network: the

model assumes a single-echelon distribution system, which

may not be applicable to Amazon’s operations.

● The need for more precise solutions: the heuristic algorithm

may also not be precise enough to meet the requirements of

Amazon’s operations.

● Generalizability: The article’s model is based on assumptions

that might not hold for Amazon’s specific context, such as the

assumptions about demand patterns and the cost structure of the

distribution system. This limits the generalizability of the

model to other contexts beyond the one studied in the article.

Journal Article 8 Fallah-Tafti, M., Honarvar, M., Tavakkoli-Moghaddam, R., &

Sadegheih, A. (2022). Mathematical modeling of a bi-objective hub

location-routing problem for rapid transit networks. RAIRO -

Operations Research, 56(5), 3733–3763.

https://doi.org/10.1051/ro/2022170

Model Mixed-integer linear programming (MILP)

Summary The article discusses the bi-objective hub location-routing problem

(HLRP) for rapid transit networks, which entails figuring out where

transit hubs should be situated and the routes that should lead therein

to best serve clients while reducing costs and travel time. The authors

present a multi-objective hybrid genetic algorithm (MOHGA) as a

solution to the problem, which is formulated as a mixed-integer linear

28
 programming (MILP) model, which includes variables such as the

number and location of hubs, transit line allocation to hubs, and transit

line frequency and capacity. They also take into account other

constraints such as capacity limits, travel time, and demand

satisfaction. The results demonstrate that the suggested approach

surpasses other current approaches in terms of solution quality and

computing time in a real-world case study of a rapid transit network in

Tehran, Iran.

Strengths ● Bi-objective model: it can be adapted to Amazon’s

transportation network to minimize both cost and time, as well

as maximizing the service provided to customers.

● Two-stage solution approach: this approach is efficient since it

can be applied to multiple types of rapid transit network, which

makes it applicable in different regions.

● The local search algorithm: it can help optimize the route and

hub location decisions, which can lead to significant cost

savings.

Weaknesses ● Assumptions and simplifications: the suggested model, like

any mathematical model, is based on certain assumptions and

simplifications. For example, it assumes that transit service

demand is evenly spread across the network and that journey

time on each transit line is constant. The validity of these

assumptions must be assessed in the context of the particular

transit network.

29
● Limited discussion of implementation challenges: The paper

focuses primarily on the mathematical formulation of the

model and does not explore the practical issues that transit

operators and policymakers may face when implementing the

model to real-world settings. A more in-depth study of these

problems would offer a better understanding of the suggested

model's viability.

● Limited comparison with other models: The suggested model

is not compared to other current models for hub location and

routing in rapid transit networks in the article. This sort of

analysis would provide light on the relative strengths and

limitations of different models, allowing researchers and

practitioners to select the best model for their unique

requirements.

● Assumes a single type of transportation mode (rapid transit):

this might not reflect the complex transportation network of

Amazon, which includes various modes of transportation.

● Limited external validation: the proposed model’s

effectiveness may vary when applied to different locations or

contexts. Therefore, to assess the model's generalizability, it

could benefit from being tested on further rapid transit

networks.

30
Journal Article 9 Peres, I. T., Repolho, H. M., Martinelli, R., & Monteiro, N. J. (2017).

Optimization in inventory-routing problem with planned

transshipment: A case study in the retail industry. International

Journal of Production Economics, 193, 748-756.

https://doi.org/10.1016/j.ijpe.2017.09.002

Model Mixed-integer linear programming (MILP)

Summary In this paper, a multi-period, multi-product Inventory-Routing

Problem with planned Transshipment (IRPT) is introduced. The use of

the optimization method in the actual retail business environment is

examined in this study using the case study methodology. The result

shows that using the IRPT approach can reduce total cost by 9% and

70% in the inventory cost.

Strengths
● The model provides an in-depth framework for decision-

making and takes into account a number of variables, including

demand, total amount of shipping product, inventory level,

transportation cost, and transshipment cost.

● Control variables are used to compare the expenses of old and

innovative forms of transportation.

Weaknesses ● The features included in the new formulations developed in this

paper are not based on specific case studies.

● The optimization model relies on certain assumptions, such as

fixed car and constant transportation costs, which may not hold

true in all real-world scenarios.

31
Journal Article 10 Rautela, H., Janjevic, M., & Winkenbach, M. (2021). Investigating the

financial impact of collection-and-delivery points in last-mile E-

commerce distribution. Research in Transportation Business &

Management, 100681. https://doi.org/10.1016/j.rtbm.2021.100681

Model Mixed integer linear programming

Summary The paper investigates the fiscal impact of collection- and- delivery

points (CDPs) in last-mile e-commerce distribution and develops an

optimization model with the ideal of minimizing the total cost of the

network encompassing the three possible routes for the entire megacity

area. The results show the use of CDPs is efficient in cost-minimizing.

Strengths ● The paper compares different scenarios which includes with

and without CDPs that gives us an immediate view of impact

on logistics cost.

● The case study methodology employed in the study allows for

a thorough analysis of certain events and applied in the

findings.

Weaknesses ● Cost savings are the primary emphasis of the study's

investigation into the financial impact of CDPs, which may not

sufficiently take into consideration the broader economic and

social benefits of CDPs.

● Further research is required to establish the generalizability and

reliability of the study's conclusions, which are currently

limited to the particular setting and assumptions employed in

the case study.

32
Appendix 2. Meetings notes

Meeting 1

Time: 12pm, Tuesday 14th of March 2023

Attendees: Unnati Agarwal, Divna Vinod Karkera, Zenique Wee, Jianing Sun

Content:

● Studied the flow of the past sample assignments and detail review and discussion of the

same

● Selected an optimization problem and finalized a company

● Allotment of task: Each one of the group members to search for journal articles in

relation to the problem

To do list: Look for as many relevant journal articles, Create an excel file to jot down the links

and title of the articles

Next meeting: 6:15 pm Monday 20th of March 2023

Meeting 2

Time: 6:15 pm Monday 20th of March 2023

Attendees: All

Content:

● Emailed Prof. William all our articles and got his confirmation of the journals

● Altered our optimization problem along with the company

● Finalised all our journal articles and consolidated them into an excel sheet

● Allotment of task: Each one of the group members to read 2 articles each, summarise

them, and identify strengths and weaknesses in a word document

33
To do list: Summarise and identify strengths and weaknesses of the article in a word document

Next meeting: 12pm Tuesday 21st of March 2023

Meeting 3

Time: 12pm Tuesday 21st of March 2023

Attendees: Unnati Agarwal, Divna Vinod Karkera, Zenique Wee, Jianing Sun

Content:

● Discussed the strength and weakness of each article, identified the points that are

common in every article

● Put together the first page of the assignment, and formatted the tables

● Task Allotment:

i) Introduction (Divna)

ii) Constraints (Jianing)

iii) Methodology and Strengths (Zenique)

iv) Weaknesses (Maria)

v) Discussion & Conclusion and General Housekeeping (Unnati)

To do list: Work on each of our individual tasks

Next Meeting: 11:30am Saturday 25th of March 2023

Meeting 4

Time: 11:30am Saturday 25th of March 2023

Attendees: All

Content:

● Finalised our strengths, weaknesses, methodology and evaluation criteria of the

assignment

34
 ● Unnati to write the final conclusion based on all the strengths and weaknesses

To do list: Write the final conclusion and discussion, vet the content and check the word count

35