See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/347443987

The Impact of Big Data Analytics on the E-commerce Industry

Preprint · December 2020

CITATIONS READS
0 5,644

1 author:

Shreyashi Shreyashi
Bournemouth University
1 PUBLICATION 0 CITATIONS

SEE PROFILE

All content following this page was uploaded by Shreyashi Shreyashi on 18 December 2020.

The user has requested enhancement of the downloaded file.

Abstract—One of the most daunting aspects of running an confirm the use of a mixed data strategy with both internal and
e-commerce organization is its absolute unpredictability. As external data figures in the project [4]. Big data analytics pro-
technology and customer behavior evolve continually, company vides value to e-commerce firms by combining the dynamics
owners need to be able to adapt. Moreover, rivalry in the market
is very strong and sellers face significant challenges such as of people, processes, and technology to transform data-related
abandonment of shopping carts and poor customer satisfaction. solutions for different business problems.
That’s why e-commerce firms are still searching for new ways to
keep their clients and improve their user experience. The cure B. Aims and objectives
is right at its fingertips: Big Data Analytics. The aim is to study the impact of Big Data Analytics
and how it has changed the shape of the advancement of
Keywords—Big Data Analytics; E-commerce; ML Algorithms; the e-commerce industry with time. The main objective is to
Customer behavior; Personalization. critically analyze the role of Data Analytics in modern-day
e-commerce and find out the strengths and weaknesses in the
I. I NTRODUCTION area.
A. Overview The paper will provide the outlook on the following key
Global e-commerce today is expanding rapidly. Several themes as it progresses:
trillion dollars are exchanged over the web annually. There has • Personlization: The action of customizing products to
been an increase in the global online population and the overall connect with the customers at an individual level.
internet population will face a rise of 66% by the year 2023 • Optimized pricing: Dynamic pricing by finding the most
having a total of 5.3 billion internet users up from the previous competitive pricing point for a commodity that maximizes
count of 3.9 billion in the year 2018 [1]. Another trend that was profitability.
noticed in the global internet usage segment was that mobile • Recommender Systems:Capable of forecasting the po-
has now become the most important channel for internet access tential choice of a collection of items for the customer,
globally. Nearly 91% of active internet users worldwide are and suggesting the top items by analyzing buyers’ pref-
mobile internet users. South Korea, UAE, and Denmark are erences.
the countries with the highest internet penetration worldwide • The concept of trust in e-commerce: Analytics helps in
while North Korea on the opposite end ranks last worldwide building a positive customer experience in the data-rich
with no online penetration among the general population [2]. environment through customer data management and data
With such an increase in the total number of users in the digital protection capability.
environment, e-commerce has noticed a massive growth. In • Enhanced Customer Service- feedback system: Main-
this paper, we shall discuss the impact of Data science in taining an overall positive purchase environment, gener-
the field of e-commerce. We shall analyze various research ating customer rating.
papers on some key areas of personalization, pricing dynamics, • Predict Trends and Demand Forecasts: Monitoring and
recommender systems, and the trust base of the consumers in analyzing future growth and opportunities for a product.
the e-commerce industry.
II. L ITERATURE R EVIEW
In the past few years, there has been a lot of research in
the field of Big data Analytics which has stirred up a lot of A. Personalization
interest from both academia and e-commerce industries. E- Personalization of products attracts more customers’ atten-
commerce firms have been working on ways to implement Big tion [5]. The problem of information overload makes it difficult
Data Analytics in this field as reports suggest a 5-6% increase for online customers to find out what they prefer. Data science
in productivity than their competitors. A recent study has algorithms make this possible. Customers feel valued and
observed that Big Data Analytics contributes to 10% or more honored when given special attention through personalization
of the growth for 56% of firms [3]. Survey results indicate that [5], [6]. The business impact of personalization is measured
88% of respondents from companies with big data initiatives through click-through rates, adoption conversion rates, sales
and revenues, user engagement and behavior, and effects of hesitation which also benefits the sellers in taking price-related
sales distribution [8]. Customers have heterogeneous needs decisions.
and making mass customization in e-commerce as a primary
style of manufacturing solves this problem. The user interface, C. Recommender system
content, and interaction process are the three dimensions that Recommender systems are widely used in e-commerce
influence the content and information displayed [6]. However, systems as they provide better matching mechanisms. These
novel quantitative measurements to understand the matrix is systems take care of price sensitivity and are socially aware
required. Data science focuses on the optimization of per- of the user’s background. Xu et al.’s [12] research show
sonalization applications by creating accurate algorithms and how social platforms affect the consumer’s purchase behavior.
prediction models in electronic commerce settings [6], [8]. Quicker purchase decisions and higher conversion rates are
Privacy concerns stop customers to share important data noticed when recommender marketing through social channels
with e-commerce websites. This makes personalization dif- is implemented. Wang et al. [13] proposed Economic Brand
ficult and hence customer satisfaction is also negatively af- Random Support Learning (EBRSL) that improves automated
fected. High privacy assurance along with high personaliza- brand marketing performance. ML is used for data mining and
tion will solve this problem [5]. Two traditional e-commerce Big data analytics helps to come up with accurate marketing
recommendation systems based on content-based filtering and strategies for the companies.
collaborating filtering are mostly used for personalization. Increasing the efficiency of categorization by applying big
With no historical data or preference records, these systems data analytics changes the way e-commerce platforms can help
failed to fulfill the customer requirement [6]. The associa- in creating a better understanding of customer needs in less
tive classification-based recommendation system proposed by time and with greater accuracy. The experiment performed
Zhang et al. solves this problem by improving efficiency and by Tan et al. provides a new paradigm with better predictive
effectiveness. To incorporate the proposed framework, a three- accuracy and surpass previous classification approaches. Using
tier architecture is implemented that facilitates connectivity Machine Translation (MT) for Product categorization can help
between the client-server, the application server, and the e-commerce organizations save a lot of manual effort that goes
database server. But the demerit of this is that it works only into updating product categories [14].
for mobile phone B2C e-commerce applications [6]. Large e-commerce companies operate their sites globally
in a variety of languages as per regions and countries and
B. Optimized pricing hence invest heavily in their MT capabilities. By using Tan et
Getting the best features at the lowest price helps the buyers al. [14] existing MT systems for product categorization, there
in making real purchase decisions [6]. Optimizing prices based will be a reduction in technical debt as companies shall have
on search patterns can play a vital role in purchase decisions. fewer algorithms, fewer systems to develop, fewer bugs to
The reputation system link to the pricing and plays an integral fix, and eventually lower maintenance cost. However, limiting
role when it comes to the consumer’s choice of the seller the experiment to two datasets questions the operational and
for purchasing a particular product online. Usually, consumers technical benefits. Being the first to demonstrate the viability
refrain from purchasing from sellers having a low reputation of this, many future works like evaluation and user interaction
score and therefore sellers usually try to maintain a good need to be done.
reputation score [6], [9], [10]. It is difficult for new sellers
to enter the market as buyers are not willing to purchase from D. The concept of Trust in e-commerce
sellers with low reputation scores. Hence, they try to increase Every organization aims to get closer to its customers.
their reputation scores using illegitimate practices reducing Understanding customers’ behavior with the expansion of
their ramp-up time. Using algorithms, ramp-up time for new analytical tools helps in increasing the revenue. Customer
sellers is accelerated through legitimate ways by providing engagement has a huge impact on sales and customer loyalty.
price discounts. The challenge is to figure out the appropriate Data science plays a crucial role in providing targeted promo-
discount as major discounts also lead to fewer profit margins tions through customer’s preferred channels and hence offering
[9], [10]. Through the algorithm, sellers provide the right a customer consistent and personalized experience [15].
discount rate using the transaction data, buyer’s long time and Trust is one of the most important concepts that should be
short time preferences over products. The experiments carried instilled into the consumers while using e-commerce systems.
out in this model have inferred an increase of 60% for short- Patricia et al. emphasize more on the theoretical part of this
term profits, a reduction in the ramp-up time by 40%. The process and how to proceed ahead to attain more qualitative
reduction in the ramp-up time also increases the sellers’ long- aspects of this topic. The major concern that has been brought
term profits by 20% [9]. ahead is the lack of trust from consumers [16]. Some surveys
E-commerce transactions are like transactions occurring indicate that 86% of Internet users have changed their online
between strangers because of the risks and complexity of using conduct, while 29% have decreased their online transactions
e-commerce. Trustworthiness in e-commerce is made strong due to identity fraud concerns [17]. The study observed online
by the presence of voice and interactivity [11]. The increasing user activity and while 75% of people felt that the Internet was
trust also helps consumers to provide their data without any a good place to perform transactions, only about 22%-55% of
people carried out transactions [18]. The key explanation for made it possible to understand every customer’s needs in
this disparity was the loss of customer trust in online vendors detail. Through advanced analytical techniques, data can be
[16]. mined and used to understand customer value [15]. Con-
Customers are happy when they receive personalized rec- sumers’ preferences are better satisfied while making online
ommendations, hence, they don’t hesitate to share more data. purchase decisions when social features are integrated with
Negative customer behavioral response is also observed for e-commerce platforms. However, the research findings from
disclosing personal information because of privacy concerns. this study have been mainly formulated from a singular social
E-commerce websites should provide privacy assurance to e-commerce platform hence might result in bias due to the
customers [5]. Online transactions require sensitive personal cultural background of the users and platform designs. Using
information. This places the consumers at risk of misalignment multiple social channels to accumulate the data, consumer
of funds (fraud), third-party vendors accessing credentials, behavior at large and for different channels can be studied.
and account details appropriating all finances attached to This would result in targeting specific social channels where
the account. Although websites ensure proper data security the conversion rate for a particular product is high [12].
practices certain segments of consumers are unaware of the To make the prediction more comprehensive and accurate,
same [5], [16]. It is inferred that data security is one of the various indices are selected through data mining that covers
key factors that must be taken into account while creating an dimensions such as order information, customer profile, and
e-commerce platform. While the organization adheres to the after-sales situation [20]. Exploring new types of data mining
data security policies, consumers too should be educated on methods is making predictions more accurate and faster than
the same to ensure the business is taken forward with honesty ever before. A model based on this was discussed by Xueling
on both fronts [16]. et al. [20]and achieved more than 85% accuracy in the forecast.
Not being widely used and tested still doubts this model
E. Enhanced Customer Service- a feedback system is working on real-time large data sets. Performance and
E-commerce communities have increased over the past two accuracy are improved and reliable conversion prediction can
decades but their process of generating feedback to build the be made by using machine learning algorithms on the previous
organization’s trust and reputation is still lacking. Organiza- state-of-the-art models even if less data is available for use.
tions have been investing a good amount of their revenue More efficient prediction models are required for detecting
share to devise ways to manage the feedback-based reputation customers with intentions to leave without purchasing. This
system better. The idea [19] is to create a robust reputation will help in converting more people and increase the profit of
system that leverages a reputation metric and attack prevention the website [21].
mechanisms. Ethical considerations must be taken care of the
transparency of algorithms in decision-making [8]. III. C ONCLUSION
Different threats such as ballot stuffing (where members E-commerce is based on a large array of knowledge that
rate themselves positively in fake transactions to inflate their would be difficult to process without the aid of big data.
reputation), bad-mouthing (where members rate others poorly From our study, we have presented how the personalization of
to deflate their reputation), and traitor attacks (where mem- products, optimized pricing, and recommender systems have
bers exploit their reputation by deceiving others until their enabled the sellers to connect to their respective buyers in
reputation dissolves) malign the reputation of the sellers. To a more familiar manner [5]- [12]. Effective implementation
avoid such issues and attacks, the paper proposes a centralized of these strategies has also enabled sellers to expand their
feedback-reputation system that not only reduces malicious customer base and establish foot into wider markets. Building
behavior but also reduces the usual feedback mechanism cost. trust is also a part of the process which also has brought up a
The paper focuses on generating proper ratings and feedback lot of challenges with itself the topmost being the protection
based on the buyers’ feedback experience. The data collected of personal data and preventing security breaches [15]- [19].
through the feedback shall process the rating mechanism of the Many phishing scams and risks come associated with data
seller. This system enables consumers to trust the product and breaches which impact the reputation of the organization as
the seller while deciding to purchase. Data analysis through well as put the consumers’ safety and resources at risk. Future
this process enables maintaining the integrity of e-commerce work is needed to overcome the limitations in this field and
organizations [19]. create a seamless customer experience. Following the recent
trends in the industry mobile shopping, voice commerce, and
F. Predict Trends and Demand Forecasts virtual shopping are on the horizon. These will enable new
New types of prediction models are being developed to purchase and marketing experiences for both buyers and sellers
predict customer churn in e-commerce platforms [20]. User and data analytics has got a big role to play in this area.
purchase predictions can be made even for small observations.
Clickstream data is generated during live shopping sessions. ACKNOWLEDGEMENT
Machine learning and deep learning helps in the prediction of The author wishes to thank Dr Philips Davies and Dr Festus
user intent from clickstream data of an e-commerce website Adedoyin in the Department of Computing, Bournemouth
[21]. Customer analytics by combining big data sources has University,UK for their continuous help and support.
 R EFERENCES [17] ”Leap of faith: Using the internet despite the dangers. Consumer
Reports Web Watch”, Advocacy.consumerreports.org, 2005.
[1] ”Cisco Annual Internet Report - Cisco Annual Internet Report [Online]. Available: https://advocacy.consumerreports.org/wp-
(2018–2023) White Paper”, Cisco, 2020. [Online]. Available: content/uploads/2013/05/leapoffaith.pdf.
https://www.cisco.com/c/en/us/solutions/collateral/executive- [18] D. Fallows, ”The Internet and Daily Life”, Pewinternet.org,
perspectives/annual-internet-report/white-paper-c11-741490.html. 2004. [Online]. Available: https://www.pewinternet.org/wp-
[Accessed: 17- Dec- 2020]. content/uploads/sites/9/media/Files/Reports/2004/PIPI nterneta ndD ailyL if e.pdf.pdf
[2] ”Internet users in the world 2020 — Statista”, Statista, 2020. [19] A. Panagopoulos, E. Koutrouli and A. Tsalgatidou, ”Modeling and Evaluating
[Online].Available: https://www.statista.com/statistics/617136/digital- a Robust Feedback-Based Reputation System for E-Commerce Platforms”,
population-worldwide/. [Accessed: 17- Dec- 2020]. ACM Transactions on the Web, vol. 11, no. 3, pp. 1-55, 2017. Available:
[3] L. Columbus, ”Making Analytics Accountable: 56% Of https://doi.org/10.1145/3057265. [Accessed 14 December 2020].
Executives Expect Analytics to Contribute To 10% Or [20] . X. Li and Z. Li, ”A Hybrid Prediction Model for E-Commerce
More Growth in 2014”, Forbes, 2014. [Online]. Available: Customer Churn Based on Logistic Regression and Extreme Gradient
https://www.forbes.com/sites/louiscolumbus/2014/12/10/making- Boosting Algorithm”, Ingénierie des systèmes d information, vol.
analytics-accountable-56-of-executives-expect-analytics-to-contribute- 24, no. 5, pp. 525-530, 2019. Available: https://search-ebscohost-
to-10-or-more-growth-in-2014/?sh=6c5c64bf3120. [Accessed: 17- Dec- com.libezproxy.bournemouth.ac.uk/login.aspx?direct=truedb=bsuAN=141446653site=eds-
2020]. livescope=site.
[4] ”Raising the Bar With Analytics”, MIT Sloan Management Review, [21] B. Requena, G. Cassani, J. Tagliabue, C. Greco and L. Lacasa, ”Shop-
2013. [Online]. Available: https://sloanreview.mit.edu/article/raising-the- per intent prediction from clickstream e-commerce data with minimal
bar-with-analytics/. [Accessed: 17- Dec- 2020]. browsing information”, Scientific Reports, vol. 10, no. 1, 2020. Available:
[5] C. Lee and D. Cranage, ”Personalisation–privacy paradox: The effects of https://doi.org/10.1038/s41598-020-73622-y. [Accessed 14 December 2020].
personalisation and privacy assurance on customer responses to travel
Web sites”, Tourism Management, vol. 32, no. 5, pp. 987-994, 2011.
Available: https://doi.org/10.1016/j.tourman.2010.08.011. [Accessed 14
December 2020].
[6] Y. Zhang and J. Jiao, ”An associative classification-based recommen-
dation system for personalization in B2C e-commerce applications”,
Expert Systems with Applications, vol. 33, no. 2, pp. 357-367, 2007.
Available: https://doi.org/10.1016/j.eswa.2006.05.005. [Accessed 14 De-
cember 2020].
[7] M. Zanker, L. Rook and D. Jannach, ”Measuring the impact of on-
line personalisation: Past, present and future”, International Journal
of Human-Computer Studies, vol. 131, pp. 160-168, 2019. Avail-
able: https://doi.org/10.1016/j.ijhcs.2019.06.006. [Accessed 14 Decem-
ber 2020].
[8] H. Xie, R. Ma and J. Lui, ”Enhancing Reputation via Price Dis-
counts in E-Commerce Systems”, ACM Transactions on Knowledge
Discovery from Data, vol. 12, no. 3, pp. 1-29, 2018. Available:
https://doi.org/10.1145/3154417. [Accessed 14 December 2020].
[9] H. Xie, R. Ma and J. Lui, ”Enhancing Reputation via Price Dis-
counts in E-Commerce Systems”, ACM Transactions on Knowledge
Discovery from Data, vol. 12, no. 3, pp. 1-29, 2018. Available:
https://doi.org/10.1145/3154417. [Accessed 14 December 2020].
[10] H. Xie, Y. Li and J. Lui, ”A Reinforcement Learning Approach to
Optimize Discount and Reputation Tradeoffs in E-commerce Systems”,
ACM Transactions on Internet Technology, vol. 20, no. 4, pp. 1-26,
2020. Available: https://doi.org/10.1145/3400024. [Accessed 14 Decem-
ber 2020].
[11] A. Basso, D. Goldberg, S. Greenspan and D. Weimer, ”First im-
pressions: emotional and cognitive factors underlying judgments of
trust e-commerce”, Proceedings of the 3rd ACM conference on
Electronic Commerce - EC ’01, pp. 137-143, 2001. Available:
https://doi.org/10.1145/501158.501173. [Accessed 14 December 2020].
[12] F. Xu, Z. Han, J. Piao and Y. Li, ””I Think You’ll Like It””, Proceedings
of the ACM on Human-Computer Interaction, vol. 3, no., pp. 1-23,
2019. Available: https://doi.org/10.1145/3359167. [Accessed 14 Decem-
ber 2020].
[13] Q. Wang, R. Cai and M. Zhao, ”E-commerce brand marketing based on
FPGA and machine learning”, Microprocessors and Microsystems, p.
103446, 2020. Available: https://doi.org/10.1016/j.micpro.2020.103446.
[Accessed 14 December 2020].
[14] L. Tan, M. Li and S. Kok, ”E-Commerce Product Categoriza-
tion via Machine Translation”, ACM Transactions on Management
Information Systems, vol. 11, no. 3, pp. 1-14, 2020. Available:
https://doi.org/10.1145/3382189.
[15] M. Hossain, S. Akter and V. Yanamandram, ”Revisiting customer
analytics capability for data-driven retailing”, Journal of Retail-
ing and Consumer Services, vol. 56, p. 102187, 2020. Available:
https://doi.org/10.1016/j.jretconser.2020.102187. [Accessed 14 Decem-
ber 2020].
[16] P. Beatty, I. Reay, S. Dick and J. Miller, ”Consumer trust in e-commerce
web sites: A meta-study”, ACM Computing Surveys, vol. 43, no.
3, pp. 1-46, 2011. Available: https://doi.org/10.1145/1922649.1922651.
[Accessed 14 December 2020].

View publication stats