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

net/publication/362857897

The Role of Big Data Analysis and Machine Learning in Marketing

Communications: A Case Study of Amazon

Chapter · June 2022

DOI: 10.4018/978-1-6684-4045-2.ch011

CITATIONS READS

0 1,972

2 authors:

Ali Aycı Shivam Tyagi

19 PUBLICATIONS 61 CITATIONS
Amazon.com
8 PUBLICATIONS 35 CITATIONS
SEE PROFILE
SEE PROFILE

All content following this page was uploaded by Ali Aycı on 12 January 2023.

The user has requested enhancement of the downloaded file.

Machine Learning for
Societal Improvement,
Modernization, and
Progress

Vishnu S. Pendyala
San Jose State University, USA

A volume in the Advances in Human and Social

Aspects of Technology (AHSAT) Book Series
Published in the United States of America by
IGI Global
Engineering Science Reference (an imprint of IGI Global)
701 E. Chocolate Avenue
Hershey PA, USA 17033
Tel: 717-533-8845
Fax: 717-533-8661
E-mail: [email protected]
Web site: http://www.igi-global.com

Copyright © 2022 by IGI Global. All rights reserved. No part of this publication may be reproduced, stored or distributed in
any form or by any means, electronic or mechanical, including photocopying, without written permission from the publisher.
Product or company names used in this set are for identification purposes only. Inclusion of the names of the products or
companies does not indicate a claim of ownership by IGI Global of the trademark or registered trademark.
Library of Congress Cataloging-in-Publication Data

Names: Pendyala, Vishnu, 1968- editor.

Title: Machine learning for societal improvement, modernization, and
progress / Vishnu Pendyala, editor.
Description: Hershey, PA : Engineering Science Reference, [2023] | Includes
bibliographical references and index. | Summary: “’Machine learning for
societal improvement, modernization, and progress’ portrays the
application of Machine Learning that is playing a prominent role in
improving the quality of life and the progress of civilization. It
focuses on not just Machine Learning, but its application to specific
domains that are resulting in substantial global progress”-- Provided by
publisher.
Identifiers: LCCN 2022008798 (print) | LCCN 2022008799 (ebook) | ISBN
9781668440452 (h/c) | ISBN 9781668440469 (s/c) | ISBN 9781668440476
(ebook)
Subjects: LCSH: Machine learning.
Classification: LCC Q325.5 .M32225 2023 (print) | LCC Q325.5 (ebook) |
DDC 006.3/1--dc23/eng20220503
LC record available at https://lccn.loc.gov/2022008798
LC ebook record available at https://lccn.loc.gov/2022008799

This book is published in the IGI Global book series Advances in Human and Social Aspects of Technology (AHSAT)
(ISSN: 2328-1316; eISSN: 2328-1324)

British Cataloguing in Publication Data

A Cataloguing in Publication record for this book is available from the British Library.

All work contributed to this book is new, previously-unpublished material. The views expressed in this book are those of the
authors, but not necessarily of the publisher.

For electronic access to this publication, please contact: [email protected].

 225

Chapter 11
The Role of Big Data Analysis
and Machine Learning in
Marketing Communications:
A Case Study of Amazon

Ali Ayci
Ankara Yıldırım Beyazıt Universtiy, Turkey

Shivam Tyagi
Amazon, USA

ABSTRACT
Technological and commercial developments in the 21st century have increased the demands and expec-
tations of consumers. Big data analysis and machine learning can play a critical role in establishing
healthy communication. The insights obtained from big data analysis and ML models can help in making
marketing communication more personalized and useful to customers. In this chapter, the authors present
an exploratory analysis of the role of big data and ML in the marketing communication of enterprises.
For this, Amazon, which provides big data analysis infrastructures to other businesses which are big
data users, is being researched as a case study. This study is one of the first studies in the literature to
examine the significance of big data and ML in marketing communication and investigate the obstacles
encountered when using them. The insights presented in this chapter will help professionals in the
marketing communication domain to better understand how they can utilize big data and ML models to
make a significant impact on their customers and sales.

INTRODUCTION

Experts estimate that consumers receive an average of between 4,000 and 10,000 marketing messages
per day (Simpson, 2017). While on one hand consumers are exposed to thousands of marketing mes-
sages every day, each consumer expects a personal approach from the companies. At the same time,

DOI: 10.4018/978-1-6684-4045-2.ch011

Copyright © 2022, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 
The Role of Big Data Analysis and Machine Learning in Marketing Communications

considering that the target audience of the enterprises is large, how is it possible for the enterprises to
communicate with their target audiences in a healthy and effective manner in line with their goals, with
personal judgments, past estimates, and unfounded assumptions? In the big data era we live in, computer
scientists, physicists, economists, mathematicians, political scientists, informatics, sociologists and other
scientists strive to access the enormous amount of information produced about people, objects and their
interactions.
The interaction of billions of people using computers, GPS devices, cell phones and medical devices
creates an ever-increasing flood of data, and many of these interactions occur through the use of mo-
bile devices in the developing world, used by people whose needs and habits have not been adequately
understood until now (WorldEconomicForum, 2012). Scientists across disciplines, businesses, public
and private institutions are exploring the potential benefits of analyzing genetic sequences, social
media interactions, health records, phone records, government records, and other digital traces left by
humans (Boyd and Crawford, 2012, p. 662; Saggi and Jain, 2018, p. 758). Although there is no official
definition (De Mauro, Greco and Grimaldi, 2016, p. 122), Big data is a resource that generates new
ideas by quickly obtaining and analyzing different sources such as humans and machines (EU, 2021)
and ultimately big data is the information asset used to be turned into value (De Mauro at all. 2016, p.
122). Big data is data management that cannot be solved with traditional databases due to the increasing
volume, velocity and variety (Amazon, 2021a). Big data can be divided into structured and unstructured
big data. Structured data comes in a defined format and offers clear answers. In user-generated content,
structured data typically has ratings and data from responses such as totally agree, strongly disagree
continues to be useful to marketers.
However, there are also negative aspects of structured big data. For example, the target audience may
be uncomfortable with being forced to answer, or people may not be able to accurately and fully express
the answer they would like to give (Bendle and Wang, 2016, p. 118). Unlike structured data, unstructured
data is not ready and comes in a form that must be processed to be usable. Such unstructured data is often
rich and exciting but can be extremely difficult to use. Managers trying to guess what consumers are
thinking by browsing these reviews will often create overly biased estimates. When the volume of reviews
is very high, reading all reviews can present significant practical difficulties (Bendle and Wang, 2016,
p. 118). With the use of big data analysis in businesses, the benefits of creating value for the company,
reducing costs, producing innovative solutions, launching new products or services, creating customer
value, helping marketing decisions, personalization in the marketing mix can be achieved (Benoit,
Lessmann and Verbeke, 2020, p. 235). Understanding the wishes and needs of the target audience is
one of the areas where businesses can benefit from big data analysis. On the one hand, businesses use
communication tools (Social media, websites, consumer forums, etc.) as large data collection areas, on
the other hand, they give their messages regarding their marketing strategies with these marketing com-
munication tools. In other words, big data is both a field where data is collected and a field where data
is created. With this feature, big data analysis is becoming an increasingly important channel in terms
of marketing communication tools for enterprises.
The findings show that research in big data applications for marketing is at the embryonic stage and
therefore more direct efforts towards business need to be developed for big data to evolve in the field
of marketing (Erevelles, Fukawa and Swayne, 2016, p. 897). Research is needed to clearly identify the
pros and cons of enterprises’ big data investments (Amado, at all, 2018, p. 6). When evaluating big data
investments, first of all, the benefits of big data for businesses should be revealed. In this study, first
of all, a general perspective is presented in order to reveal how and for what purpose big data can be

226

The Role of Big Data Analysis and Machine Learning in Marketing Communications

used in marketing communications. Until recently, big data investments had to be met with business
resources, and academic studies were based on high investment values ​​and limited opportunities for
businesses. However, today, some leading companies providing big data analysis services are bringing
their services and capabilities with other businesses at affordable costs. In this context, Amazon and
Amazon Personalize (AP) are researched as a case study in order to investigate the service scopes and
usability of outsourcing organizations for the use of big data in marketing communications. technically
and practically. The reason why Amazon is included in the research is that it is one of the businesses that
use big data and big data analysis most effectively, and more importantly, it provides big data analysis
infrastructure services to other businesses, especially SMEs. It makes big data applications more acces-
sible for companies. Because Amazon is a business that markets its services online, the business’ website
is used as the data source for the exploratory research within the scope of this article. Finding detailed
technical information about the services offered on the Amazon web page also provides convenience in
accessing information. This study is one of the few studies in the international literature that deals with
marketing communication and big data analysis. New and up-to-date studies are needed because con-
sumer expectations, which is one of the most important areas of use of big data, are constantly changing
and technology and actors in big data analysis are constantly developing.

BIGDATA, MACHINE LEARNING AND MARKETING

COMMUNICATION FROM MARKETING ASPECTS

Marketing Communications and Big Data Needs

Traditionally, marketing has always been a field in which data plays an important role. Since the 1980s,
with the development and cheapening of computing technologies, more business processes have been
digitized and the data produced and captured by the devices have been made available to marketing
researchers and practitioners. Customer choice modelling, market basket analysis, estimation of price
(cross) elasticities, etc. have emerged and evolved with the increasing availability of various types of
data (Benoit, et al., 2020, p. 236). While deciding on the target customer profile in marketing, defining
information sources, and ensuring customer participation in product design, businesses benefit from big
data (Zhan, et al, 2018, p. 584). Big data analytics is a powerful analytical tool to understand market and
customer needs for products or services and enable them to make better decisions (Saggi and Jain, 2018,
p. 782) and serves as a tool to identify customer value creation strategies, define an overall marketing
strategy, develop product concept features, identify value creation opportunities, and develop customer
value metrics (Smith and Colgate, 2007, p. 15). The main research topic of marketing is to understand
the target audience correctly. Based on the results of field research; while 88 per cent of IT managers
believe their organization truly understands its customers, only 61 per cent of consumers believe that
companies understand their needs (Talend, 2021). Big data analysis offers important opportunities for
businesses to understand and respond to consumers’ needs and expectations in a timely and accurate
manner. (Buganza, Trabucchi and Pellizzoni, 2020, p. 58). Big data is the biggest opportunity that changes
the rules of the marketing and sales ecosystem and allows us to understand the target audience in more
detail since the internet became mainstream nearly 25 years ago (McKinsey, 2015, p. 4). As machine
learning evolves and analyzes different types of data faster; Big data is playing a more central role in
marketing than ever before, making the pace of development of marketing messages a more important

227
 
The Role of Big Data Analysis and Machine Learning in Marketing Communications

revenue item for the business (Luenendonk, 2019). Prior to the recent interest of businesses in analyzing
marketing data at a large scale, rapid advances in network and database technology have led to the devel-
opment of a variety of software applications. This technology not only allowed business processes to be
further integrated by interconnected computers but also stimulated interest in analyzing data where the
volume and diversity are large, such as the growth of the internet, interacting with potential and existing
customers, gathering information about competitors (Benoit, at al, 2020, p. 236). While marketers are
trying to use big data analysis commercially, it is seen that the interest of academicians on the subject
has increased in the literature. According to the results of Amado et al.’s (2018: 6) study of 1560 articles
published between 2010 and 2015 on the use of big data in marketing, big data in marketing has received
increasing attention over the years and the number of publications doubles every year compared to the
previous year. As a result of the use of big data in marketing activities by researchers and practitioners,
businesses have made significant gains. The results of the research conducted by Suoniemi et al. (2020)
on 301 marketing managers reveal that the use of big data improves firm performance by improving the
firm’s market-oriented capabilities.
The main way to make businesses competitive and differentiate them in an internationally competi-
tive environment is healthy communication with the target audience. Marketing communications can
provide a sustainable competitive advantage to any organization that taps into its potential (Mihaela,
2015, p. 1447). Marketing communication is the coordinated use of promotional messages and related
media to communicate with a market, where marketing communication messages are communicated
through one or more channels such as digital media, printed materials, radio, television, direct mail,
and personal selling (AMA, 2021). Integrated marketing communication, which expresses the harmony
of communication tools in the business, is an activity that allows creating profitable customer relations
and creates value for the product/brand or company. It aims to ensure the consistency of the message
conveyed through various communication tools such as advertising, sales promotion, direct marketing,
public relations, online communication (Mihaela, 2015, p. 1949).

Customer Oriented Marketing

In customer-oriented marketing, it is aimed to collect consumer information, to make effective inferences

as a result of detailed analysis of this information, to predict the future from the analysis of past customer
interactions, and finally to respond to customer needs in real-time (Gulati and Oldroyd, 2005, p. 97). For
this purpose, customer-oriented marketing and customer-oriented communication become more effec-
tive with big data analysis, which is one of the most effective data collection and analysis methods. The
increased focus on customer experience arises due to the complexity of the communication process as
customers interact with companies through multiple channels and numerous touchpoints in the media,
due to the redundancy of tools and messages in marketing communication. (Lemon and Verhoef, 2016,
p. 69). The integration of technology and big data into business strategy increases the complexity of
marketing communications and fosters the need for advanced marketing performance analytics (Buhalis
and Volchek, 2021, p. 1). Effective management of the complexity in the communication process is
possible with big data analysis tools. Big data enables strategic communicators to analyze the needs,
views, attitudes and behaviors of their stakeholders in greater detail. From planning to evaluation, big
data analytics make enterprise communication more analytical and potentially more strategic.

228

The Role of Big Data Analysis and Machine Learning in Marketing Communications

Communication professionals use advanced analytics tools to capture and process the needs, opinions,
attitudes and user behavior of target groups from all over the world. Marketers by connecting this data to
market data can monitor the changing perceptions of their companies, brands and products in different
countries and this in more or less real-time. Insights from big data analytics allow companies to conduct
more personalized communications in marketing, public relations and internal communications than in
the past (Wiencierz and Röttger, 2018, p. 43). A study was conducted by Wiencierz and Röttger (2018,
p, 42) with 35 big data and social media analytics experts on the use of big data in the field of commu-
nication, and the results of the research show that big data offers benefits to marketing communication
practitioners in four main areas. a) providing communicators with a wider knowledge base for situation
analysis b) formulating database-based strategies c) automated communication applications d) providing
evaluation based on larger data stocks. Marketers benefit from big data analytics in individual product
and brand communication, online touchpoint analytics, predictive analytics of product and label mes-
sage spread, topic or content marketing identification, real-time processing of customer data for online
marketing, market share determination through share analytics, and effective communication. In order for
big data analysis to be successful in marketing communication, it should be clear which of the goals of
the business serves the big data, it should be repeated several times until the results of the efforts such as
data generation, extraction and analysis are compatible with the determined goal, and interdepartmental
cooperation should be supported (Wiencierz and Röttger, 2018, p. 42).

Behavior-oriented Marketing

The analysis of consumer behavior is of great importance in the effectiveness of advertising decisions,
which is one of the leading marketing communication tools. While advertisements in traditional media
can present common content regardless of differences, advertisements can be personalized with big
data technology, allowing the delivery of advertisements suitable for the region, culture and person
(Yavuz and Erdoğan-Gençyürek, 2019, p. 214). Behavioral advertising is one way to better target the
most valuable consumers. For example, if past purchasing behavior in one product category is positively
correlated with product preference in another category, consumers who purchased the previous category
may be more heavily advertised (Shen and Villas-Boas, 2018). Netflix, one of the most valuable media
companies in the world, owes its success to its impressive customer retention rates, and the secret of
Netflix’s customer retention is that it is more successful in determining what they want by communicat-
ing effectively with its viewers compared to its competitors. Netflix collects data from its 151 million
subscribers and applies data analytics models to explore customer behavior and purchasing patterns,
then uses this information to recommend movies and TV shows based on their subscribers’ preferences
and more than 75% of audience activity is based on personalized recommendations with offers deliv-
ered through different marketing communication tools (Dixon, 2019). At the macro level, the two main
components of interactive integrated marketing communication are big data marketing information and
its technical system and interactive marketing communication plans (Tang, 2020). Social media is the
medium where big data is most active with marketing communication. After the analysis of the big
data obtained from social media, communication with the target audience is established with different
marketing communication tools in line with the strategies compatible with the business objectives and
this is a continuous cycle. In other words, while marketing communication tools are used to provide big
data, it is also a basic tool used for effective communication after the analysis of big data (Figure 1).

229
 
The Role of Big Data Analysis and Machine Learning in Marketing Communications

Figure 1. Big Data Usage Cycle in Marketing Communications. Source: (Authors, 2022)

It is seen that different people, institutions and organizations have made big data analysis in social
media the building block of their communication strategies and have achieved significant success.
Coca-Cola is able to respond to its customers in real-time by analyzing and evaluating the information
of Twitter users around the world, using services from Sysomos, a social media analytics business.
Former US President Obama predicted the intentions and tendencies of undecided voters by using the
Boatbuilder system, which can manage various databases in the 2008 elections, closely analyzed the
voter movement and responded to them with a direct communication strategy in real-time and won the
US presidential election (Bitnine, 2016). The critical point in the effective design and implementation
of these steps and the development of strategies is to collect and analyze as much qualified information
as possible from the target audience. With big data analysis, businesses can fully understand consum-
ers’ concerns, reach opinions about competitors and the industry, and find the opportunity to build their
marketing communications on these findings.

Active vs Passive Recipient’s Impact on Marketing

In the era of social commerce, the first part of the 21st century, the basic idea is that customers do not
become passive recipients of information from marketers, but interact with each other and with the
business to influence consumer purchasing and business decisions. These interactions can take the form
of product reviews, product usage descriptions, “homemade ads”, blogs, and other consumer-initiated
contributions (Fader and Winer, 2012, p. 369). Businesses need to develop timely and effective strategic
responses to contributions such as consumer review information, as silence to incoming messages is
often not a solution (Chen and Xie, 2008, p. 488). It now has access not only to purchasing and marketing
mix data but also to rich information about how consumers interact with each other and the products/
services they consume. This provides a much more complete picture of the effects on buyers and has

230

The Role of Big Data Analysis and Machine Learning in Marketing Communications

the potential to significantly improve modelling efforts and ultimately marketing decisions (Fader and
Winer, 2012, p. 369). Because analysis of social media data provides valuable information for market-
ing research purposes, businesses should use this data as a form of exploratory research (Malthouse,
et al, 2013, p. 270). By observing what consumers write about products in a category, firms can better
understand the online discussion and marketing opportunities, market structure, competitive landscape,
and characteristics of their own and competitors’ products (Netzer, et al, 2012, p. 521). Businesses soon
realized the potential of using internet-based social networks to influence customers and incorporated
social media marketing communications into their strategies to strengthen their businesses (Moro, Rita
and Vala, 2016, p. 3341). A system that can predict the impact of individual posts would be a valuable
advantage in promoting products and services (Chen and Xie, 2008, p. 477). 12 factors can be used to
measure the performance of social media posts (Table 1).

Table 1. Social Media Posts Performance Aspects. Source: (Moro et al. 2016, p. 3343)

Characteristic Indicator
Number of people who saw the post on the page
Full access to the post The number of times a post is viewed regardless of whether it is clicked or not. People can see more than
one.
They are the impressions of the same article. For example, someone might see a page update in their
Total views
news feed once, and then see it a second time if a friend shares it.
Users interacting The number of people (unique users) who clicked anywhere on a post.
Consumers The number of people who clicked anywhere on a post
Post consumption Number of clicks anywhere on a post
Impressions of people who
Total impressions from people who only liked a page
like a page
Reach of people who like
The number of people who saw a page post because they liked that page (unique users).
a page
People who like a page and
The number of people who liked a page and clicked anywhere on a post (Unique users).
are interested in a post
Comments Number of comments on the post
Likes Number of likes on the post
Shares Number of posts shared
Total interactions Total of likes, “comments and shares”

This dataset is also linked to all phases of branding (cognitive, affective and cognitive phases) and
shows that predictions about social media interaction can also be used to predict brand building. Managers
will be able to use this information to make informed decisions about the posts they publish, increasing
their impact and thus contributing to successful brand communication and brand building (Moro, et al.,
2016, p. 3350).

231
 
The Role of Big Data Analysis and Machine Learning in Marketing Communications

METHODOLOGY

Within the scope of this study, the technical dimensions of big data analysis and machine learning in
marketing communication, the problems experienced and the future perspective will be investigated. As
a research method, exploratory analysis will be conducted to obtain data by in-depth interview method
from experts who work in Amazon. Afterwards, as a case study, Amazon AP, which is the product
Amazon offers to other businesses, will be analyzed in accordance with the information provided by
the experts, the information obtained from the product and its web page. In face-to-face interviews, the
participants were asked to answer in-depth 15 questions spanning over different paradigms including
marketing problems and big data, ML solutions, services used and offered by Amazon itself, problems
they encounter and the future of machine learning and big data in marketing communication.

Technical Aspects Of Marketing Communication

With Bigdata And Machine Learning

While the rapid growth in user content on the Internet has a huge impact on marketing practice at the
same time, brand managers are no longer able to fully control the messages they use (Fader and Winer,
2012, p. 369; Malthouse, at al., 2013, p. 270). The huge amount of data makes it difficult to track and
measure information, and this rich but unstructured set of consumer data is qualitative in nature. A
combination of a text mining device and a network analysis framework can be used to overcome these
challenges (Netzer, et al, 2012, p. 521). Unstructured big data from online product reviews are generated
in real-time, easily accessible, and contain the messages consumers want managers to hear. Analyzing
such data has the potential to revolutionize market research and competitive analysis, but how can the
messages be deduced? How can large amounts of data be concentrated on insights to help drive business
strategy? How can large amounts of data be concentrated on insights to help drive business strategy?

Data Collection

Different types of data can shed light on different aspects of CX. Depending on the requirement, clients
can decide what type of data to be collected on their websites. As in this study, we are focussing on
marketing communication, customer interaction and behavioral data is of importance and thus we will
focus on strategies to collect that data. For big data applications, customer data is collected in three ways,
asking customers directly, by tracking data indirectly and by acquiring data from other companies. There
are tools which can be enabled in a website to track what item the customer is looking at, which buttons
and links he/she clicked and how much time the customer spent over a specific section or webpage. All
these data points when fed into a big data analysis tool or ML algorithm can generate insights that can
help in understanding and tracking user interests and behavior.
But there are some hurdles that companies face during the data collection phase. With the latest
development in privacy controls and new regulations being placed by governments, data collection is
becoming a more and more challenging task. In research, it has been observed that after Apple provided
an option to customers to block any type of data tracking, almost 25% of customers have opted out of it.
This is a huge impact as it means 25% less data for training models. Government regulations like CCPA
and GDPR give power to customers for the right to be forgotten. This brings an obligation on companies
to delete a customer’s data when opted. Due to these new controls and regulations, data collection needs

232

The Role of Big Data Analysis and Machine Learning in Marketing Communications

to be very transparent. Indirect tracking of data and selling of data is not encouraged to practice. True
that data collection is a matter of concern but also it has a big role to play in improving CX by helping
in identifying patterns in customer behavior and hence allowing companies to create targeted offers
and advertisements. Companies like Facebook, Snapchat and Twitter which earn most of their revenue
by advertisements are on the receiving end of the push towards privacy controls. They are working on
changing their underlying architecture to minimize the impact. In this direction data collection methods
like interviews, surveys, GPS based location tracking and data collection through IoT devices are being
used more and more.

Data Preprocessing

Most of the time raw data available is not suitable to be used in machine learning models as is. Data
preprocessing is a very essential part of the whole process. The quality of inferences directly depends on
the quality of the data. If the data is incomplete or incoherent, the results will be unreliable. To ensure
that proper inferences are derived, a lot of effort is dedicated to data preprocessing. There are certain
techniques like sampling and imputation to clean and prepare data before being used as input to ML
models. Once the data is validated to be in good form, the ML model is trained and tested using this
data for analysis.

Figure 2. Data Preprocessing. Source: (Authors, 2022)

For training machine learning models, the data is divided into train, test and validation data. There
are many ways to do this. Like data can be divided into 70:15:15 shares where 70% data will be used
for training, 15% data for testing and 15% for validation. In another approach, there can be overlapping
sets that will have some data in common. In some approaches the data set is divided into N number of
sets and training is performed using a certain number of sets each time.
There are three main challenges of any data pre-processing stage which can be defined by three Vs,
Volume, Variety and Velocity. As their literal meanings, volume means amount of data, variety means
different types of data and velocity signifies the speed with which data is processed. All three of these
directly impact the data preprocessing step and the quality of results hence obtained. In today’s world
where IoT and sensors are all around us, millions of data points are obtained every second. To save,

233
 
The Role of Big Data Analysis and Machine Learning in Marketing Communications

identify, filter and process these data points requires a lot of computing resources. Today hardware
designers are working on creating more and more efficient processors to speed up the calculations and
hence provide results as quickly as possible. With these advancements in computing power, processing
large amounts of data has become possible. But even then as the data increases with every day, new
approaches are being researched. Distributed processing is one such field where companies like Google
and Amazon have excelled a lot.

Figure 3. Training, Validation and Testing Dataset

Source: (Authors, 2022)

Distributed Data Processing

There are a lot of softwares for big data analysis. Spark is an open-source distributed computing framework
used to analyze big data quickly. Its ability to divide the task among workers to process chunks of data in
parallel allows it to analyze large amounts of data quickly, hence making it apt for big data applications.
For marketing, communication spark can be used to analyze the data of millions of customers in real-
time and thus provide inferences that can be integrated with marketing messages instantly. Engineers and
scientists can use Spark APIs to analyze user data from different sources to generate insights (Alzahrani,
2021). Hadoop is another distributed data processing framework that uses maps and reduces algorithms
to distribute tasks and combine results to get final inferences. The main difference between Hadoop
and Spark is the place where data is stored. Spark saves data in RAM while Hadoop uses a hard disk to
save data. This makes Spark faster than Hadoop. Thus Spark is good for applications that need real-time
inferences while Hadoop is used for applications that require batch processing. Amazon EMR or Elastic
Map Reduce service is another big data platform based on Hadoop that provides the capability to run
distributed data processing jobs using open source frameworks like Spark. Amazon EMR distributes
tasks among virtual servers Amazon EC2 instances and saves data on Amazon S3. Due to being on the
cloud, this service has the capability to scale up and down dynamically depending on the requirement.
Although these programs are not very expensive to acquire, professionals are needed to use them.

234

The Role of Big Data Analysis and Machine Learning in Marketing Communications

Data Analysis Models for Marketing Communication

Artificial intelligence tools driven by machine learning algorithms are rapidly transforming the business
world and this transformation is increasing the interest of researchers. Machine learning methods have
flexible model structures that can handle large-scale and unstructured data and provide strong predictive
performance (Ma and Sun, 2020, p. 481). Although it is argued that big data will remove the human fac-
tor (Akçan, 2016, p. 311) or not (Altunışık, 2015, p. 72), it is a fact that it will enrich the human factor
qualitatively. Marketing communication consists of three main types of content, text, images and videos.
There are different approaches to analyze each of these types effectively using AI. In this section, we
will discuss some of the models for each data type and how they can impact marketing communication.

Text Data

Text analysis of user-generated content provides access to information provided by active, content-
producing consumers, providing useful information from the social structure of a brand’s follower base,
regardless of whether it creates or consumes content. This is particularly useful for predicting perceptual
traits that consumers may be less likely to mention in brand conversations rather than key product features
(Culotta and Cutler, 2016, p. 347). In order to make these analyzes, Twitter is a very important tool in
big data analysis, as it is popular, relevant, social connections are public, and organized as it is organized
by users into topic-based lists (Culotta and Cutler, 2016, p. 348). Unlike traditional transaction data,
social media is a difficult medium to research due to data volume, diversity and speed (Arrigo, Liberati
and Mariani, 2021). Big data analysis and suggestions can be made by people as well as artificial intel-
ligence tools can be used. There are many AI approaches to analyze text data today. Topic-based analysis
is one such methodology where an AI model tries to find a topic for each section of the text and then
derives an overall topic. This is also known as topic modeling. Latent Dirichlet Allocation (LDA) is a
good example of this approach.
LDA is a statistical model that works on the problem of topic discovery. It works by changing each
word to its root lexical forms and then checking which topic they relate to. For example Fly, Flying,
Flies, Flew will all be converted to Fly and then the system can correlate that word to most probable
topics like a bird or plane. So whenever there is an ambiguous situation like this, LDA will try to predict
probabilities of all probable topics and present them to us. Plate annotations can be used to represent
such probabilistic graphical models (Blei et al. 2003).
In the representation of LDA, their is only one observable variable in the whole process, all others
are latent variables which are inferred. There are three levels in LDA, corpus, document (outer box) and
words(inner box). LDA can analyze large amounts of text and identify content by focusing on attributes
that do not appear with a certain weight. For example, in examining a graphic text, it can be analyzed by
focusing 70% on the story and 30% on the graphics. Collecting content from a large number of consumers
allows us to understand what is on the minds of consumers collectively, from which it can be revealed
what consumers care about and which features are viewed positively or negatively. Thus, for example,
if advertisements will be given in newspapers, healthy communication will be established by providing
the target audience to concentrate on the areas that the business wants to concentrate on while saving
on advertising expenses by including the features that the target audience sees positive. LDA is able to
map the relative strategic positions of competitors in the minds of consumers where they matter most
(Bendle and Wang, 2016, p. 115). Any method that tries to understand the message must extract the

235
 
The Role of Big Data Analysis and Machine Learning in Marketing Communications

common themes between the interpretations and thus bring together the unseen qualities expressed at
various levels of abstraction. As one of the effective methods of this grouping, LDA has similarities with
the principal component methods performed in customer surveys in terms of extracting common core
characteristics. The advantage of the method is that it works with a large amount of unstructured data
where not every customer review has most attributes. While LDA is somewhat technical, it is relatively
easy to implement for statistically trained analysts and the data required for analysis can be accessed from
social media. Without in-house expertise, data can be easily and relatively inexpensively purchased from
third parties. LDA reveals the hidden nature of the invisible attributes behind comments. This provides
more insight into what consumers think, telling management which areas of the customer experience
need attention (Bendle and Wang, 2016, p. 123). Thus, very critical information is obtained in terms of
strategic communication decisions.

Image Data

Advanced machine learning algorithms integrated with big data power recommendation systems on
e-commerce sites and content platforms like Amazon and Netflix. Learning engines analyze and tag
billions of images on social media sites like Facebook. Automatic bidding algorithms analyze a user’s
profile on a millisecond timescale to determine the most appropriate bid for ad serving; chatbots engage
in human-like conversations with customers to maintain relationships and loyalty (Ma and Sun, 2020,
p. 482). Despite significant advances and growing interest, the use of machine learning methods for big
data in marketing is still at an early stage and current work is somewhat scattered. For example, there
is no consensus on how machine learning methods should be incorporated into marketing research (Ma
and Sun, 2020, p. 482). One challenge with using big data in marketing communications is that while
automation can be more efficient for businesses, people still want to talk to people (Ariker and Manuel,
2015, p. 99).

Video Data

The ability of machine learning methods to handle unstructured big data is quite high and unstructured
media data (text, audio, video recordings), consumer monitoring data, network data, hybrid data created
by the combination of different data are transformed into unique information for marketing managers by
machine learning tools (Ma and Sun, 2020, p. 497). Video-based analysis derives real-time inferences by
analyzing the customer videos for behavior patterns that can then be materialized for enhancing marketing
communication. With the increase in computation power now it is possible to analyze video data with
more speed and in more volume. Thus it is apt time to use video-based analysis to derive useful insights.
One tool that can be used to benefit from video data in marketing communication is the Video-Based
Automatic Suggestion (VAR) System. The system uses a camera that captures the customer’s behavior
in front of the mirror to make inferences about their preferences based on facial expressions and clothing
viewed by consumers at points in time. Second, the system identifies customers with preferences similar
to the focus customer from a database of shoppers whose preferences, purchasing and/or evaluation deci-
sions are known. Finally, recommendations are made to the focal customer based on the preferences of
these like-minded customers and their purchasing and/or evaluation decisions. Each of the three steps
can be implemented in various variations, and a retail chain can choose the very specific configuration
that best serves its purpose (Lu, Xiao, and Ding, 2016, p. 484).

236

The Role of Big Data Analysis and Machine Learning in Marketing Communications

General design of the VAR system can be understood by example where a video feed from the chang-
ing room can analyze the behavior of the customers and their likings. Then the insights can be used to
recommend better choices in marketing communication. If the VAR system is to be implemented in
Walmart which has 10526 branches and 2.3 million employees in 24 countries (Walmart, 2021), the size
of the data and the scope of its analysis will be understood. Unknown to the consumer, a significant part
of the purchasing journey is guided by automated systems.
Search results are generated by an advanced Google ranking system, determined in part by advertisers’
bids automatically generated using bidding machines (Ma and Sun, 2020, p. 481). Content on websites
is customized according to user profile and questions are answered by chatbots. Reviews read by con-
sumers stand out because they are deemed useful by a rating algorithm, and ads they see repeatedly are
delivered through retargeting algorithms with real-time bidding. Coupons offering personalized prices
to a consumer are generated just in time by the pricing engine. Finally, their posts on social media are
collected by social listening engines and analyzed for emotion and feedback. These automated systems
that make instant and context-dependent decisions are artificial intelligence tools implemented using
state-of-the-art machine learning algorithms.

CASE STUDY: AMAZON AS A MARKETING

COMMUNICATION SYSTEM PROVIDER

Founded in 1994 as a bookselling site, Amazon sold books to people in the US and 45 different countries
in its first month (Hartmans, 2021). With 1.3 million employees today, Amazon generated $21.3 billion
in annual sales of more than $386 billion in 2020 (Fortune, 2021). Amazon ranks second in the Fortune
top 500 list with a market capitalization approaching $2 Trillion (Fortune, 2021). As a leader in customer
experience innovation and defining many aspects of customer experience, Amazon is reorganizing its
marketing activities around AI and machine learning efforts, not only taking things to the next level but
showing how other businesses can do the same. Amazon provides big data analysis services to other
large, medium and small businesses as well as serving the American Space and Aviation Administration
(NASA) with its solutions based on big data analysis (Morgan, 2018). One of the most effective uses of
big data and analytics, Amazon collects more than 2,000 historical and real-time data points on every
order, using machine learning algorithms to find transactions with high fraud probability and stopping
millions of dollars of fraudulent transactions every year (InsideBigData, 2019). Sharing this informa-
tion alone is part of Amazon’s effective marketing communication for the long-term relationship, not
the instantaneous one.

Amazon Personalize (AP) As A Marketing Communication Tool

Amazon provides infrastructure for the use and analysis of big data to other businesses through its AP
service. AP makes it easy for businesses to create applications that can offer a wide variety of person-
alization experiences, such as specific product recommendations, personalized product rankings, and
customized direct marketing. AP is a fully managed machine learning service by training, tuning and
deploying custom machine learning (ML) models beyond strict static rules-based recommendation
systems to provide customized recommendations to customers in industries such as retail, media, en-
tertainment (Amazon, 2021b).

237
 
The Role of Big Data Analysis and Machine Learning in Marketing Communications

Deep Dive Into AP’s Client Space

Amazon Personalize is a service that Amazon provides to its clients who want to obtain recommendations
based on their user data. Though AP is being used by companies in many fields most of the AP clients
come from two main industries, media entertainment and retail. Amazon pioneered personalization in
e-commerce more than 20 years ago so it is a natural fit, bringing the Amazon Personalized service to
retail clients, with all the historical learning and experience. Due to Amazon’s vast experience in retail,
AP has been developed in such a way that it resonates very well with the requirements and expectations
of its retail clients(Jory, 2022).
In media entertainment, the scope of AP’s use cases is very broad. AP’s clients can be media com-
panies that are streaming video on demand or streaming sports or streaming short videos. In addition
to these, there are use cases where clients are streaming music or textual data like news and articles.
An example of an AP client is Discovery, which is using AP to personalize all aspects of their user
experience including recommendations on the content rails. Down under, AP doesn’t really care if it’s
a product client is trying to sell, a video client is trying to stream or a song client is trying to get people
to listen to. It just looks at how users are interacting with those items within the client’s environment
including its application, emails, web app, mobile app and so on (Jory, 2022). The results obtained from
AP depend a lot on the quality of data that is fed into the models. The data is owned by the client. AP
provides algorithms in which clients can ingest their own data to obtain insights. AP is not like other
services which provide already pre-trained models in the form of APIs which clients can call. AP takes
the data from the client, uses the recipes to create a trained model. For this clients can save data in AWS
S3 buckets or other storage services. The information obtained from Amazon’s cloud storage service
and the Personalize interface is analyzed and presented to the company, with the models optimized with
the right parameters, and presented to the company simultaneously with the models. With AP service
clubbed with other AWS services, Amazon has provided an easy-to-use and highly scalable platform
that helps clients to generate recommendations for users. By providing the necessary infrastructure,
Amazon manages the entire pipeline, including processing data, recognizing features, using the best
algorithms, training, optimizing and hosting models (Amazon, 2021b). So in the end, every client has
their own dedicated trained model. The focus of AP is on providing the clients with a quality ML model
which can be used by them to generate useful insights(Expert, 2022).
The easy integration of the system with different marketing communication tools is a very important
feature and it ensures that the people who use the system are not directed to only one communication
tool. The most important question in mind is the confidentiality of information, and businesses may
rightly be concerned about the confidentiality of shared information. (Altunışık, 2015, p. 61). Ama-
zon’s offering for businesses states that the data will be privately and securely encrypted and will only
be used to generate recommendations to customers and will not be shared with customers or Amazon.
com. (Amazon, 2021b). The benefits of applying big data analytics for the retail industry are listed as
follows: Personalizing user home pages with product recommendations based on their shopping history,
suggesting similar items on product details pages to help users find what they are looking for easily,
helping users quickly find relevant new products, deals and promotions, improving marketing commu-
nication through personalization of push notifications and marketing emails with individualized product
recommendations, improving sales with relevant product rankings and high-quality up-sell/cross-sell
suggestions in the cart (Amazon, 2021b).

238

The Role of Big Data Analysis and Machine Learning in Marketing Communications

Data

One important aspect of the data is temporal details. AP can analyze the temporal details in real-time to
help the clients understand the evolving nature of customer interaction. This temporality aspect is really
powerful because it allows AP to understand how customer interests are changing over time in contrast
to the collaborative filtering approach that’s just using a matrix of items. “When did the customer inter-
act” becomes really important when the client is looking to capture “what the customer is interested in
now” as opposed to two weeks ago. Sophisticated combination of data and various ML models which
are called recipes are used by AP under the hood to provide clients with the best results. These recipes
are structured targeting different personalization use cases and hence provide a wide choice of selec-
tion to clients(Expert, 2022). Temporal aspect is a groundbreaking discovery as it allows the model to
understand the evolving intent of a customer and can predict what he may buy next or alongwith an
item already in his cart. This knowledge helps AP models to provide solid recommendations to clients.
There’s another aspect to the data that clients can provide to AP and it’s very powerful. It is what
is called contextual data. This data is very specific to an interaction. Usually clients need to provide a
user ID and item ID and a timestamp when a user interacts with an item at a particular time. But the
client can also provide additional data about that interaction like when did the interaction occur? Like
seasonality, is it winter, fall or spring. So if a client is selling clothing, user interest varies based on the
weather conditions for that season. Another example is day of the week, where consumers consume
video content differently on weekends. Maybe they watch long movies with their family. But on week-
day mornings, maybe they’re looking for short 30-minute news clips or sitcoms, whether commuting
to work. This type of data. Device type is another really common contextual data. So are customers
consuming content on their mobile phones, or on their desktops, or maybe on their set-top boxes. All
of this contextual information can help AP learn the nuances of how the same user interacts differently
based on different contexts. This allows AP to make even more relevant recommendations based on
the current context when you’re providing those recommendations. The dynamic filters included in the
system create better user experiences by allowing clients to tailor the business rules for each user when
recommendations are created, providing the opportunity for clients to control the recommendations in
real-time while responding to the individual needs and preferences of the users (Chwastek, Pooniwala
and Ashman, 2020).
Examining the customers, it is seen that AP users are international companies (Subway, Yamaha,
Dominos etc.). However, the presence of promotions, free apps, training, free trials, minimum fees and
no commitments shows that the AP target audience includes other businesses. The potential for SMEs
to access these services is very exciting. These developments show that easy and relatively inexpensive
applications started to exist shortly after Altunışık (2015) found that “it is extremely difficult to obtain
and benefiting from its benefits is extremely difficult, but also very troublesome”. AP offers personal-
ized impressions, similar products, personalized redirection, personalized promotions/notifications for
businesses using its own infrastructure (Amazon, 2021b).
LotteMart, one of the users of the AP system, is a supermarket chain in South Korea with 189 stores
and more than 600,000 users, including an online store (Lottemart, 2021). AP enabled LotteMart to
improve the customer experience with tailored recommendations. By focusing only on datasets for
predefined interactions, users and items helped Lottemart in time savings tune to almost 50% of what
they used to spend earlier in getting insights from the data. The use of coupons, a tool of marketing
communication, has more than doubled since the introduction of personalized recommendations. In

239
 
The Role of Big Data Analysis and Machine Learning in Marketing Communications

addition, there was a 1.7-fold increase in the frequency of new product sales. (Park and Hwang, 2020).
By using the AP system at Subway, the fast-food restaurant with branches in 100 countries, can offer
personalized recommendations to its customers without the need for machine learning expertise of
company personnel(Amazon, 2021b). It is a fact that subjects such as big data, artificial intelligence,
data mining, and machine learning are complex and require expertise. It is seen that there are simple
package customization services that include all these concepts in the AP system and will be sufficient
for the use of the enterprise in practice. AP is now being used in a variety of fields spanning blogs in
which the users share their information, training videos, visuals, frequently asked questions and answers
sections, certified training programs, customer service support consisting of professional experts, and
many others(Amazon, 2021b).
With the effective use of AP, it is seen that the communication with the target audience of the enter-
prise can be improved in a more effective, fast and healthy way, which can be relatively cost-effective.
With the AP product in the literature, as a result of big data analysis, the subject of determining market-
ing activities: limited access to big data, trust problem, costs, complexity, lack of necessary skills (Boyd
and Crawford, 2012, p. 662; Fosso et al. 2015, p. 244; Johnson, 2012; Altunışık, 2015, p. 61; Ariker,
Breuer, and McGuire, 2015, p. 91) are partly eliminated.
In the growing space of media and retail which sees most of the AP clients, competition is very fierce.
With time it is getting extremely easy for customers to switch from one provider to another. Customer
experience was never as important as it is in today’s market. It is very essential for providers that they
give the customer the best quality and experience. For this using APs recommendations, clients can
segment customers that have an affinity for the promoted items. This gives a much better chance of
engagement of customers, increasing the probability of getting an open click and then conversion to an
order. It’s like every customer have their own dedicated, trained model to them (Expert 2, 2022). This
is also what differentiates AP from other competitors who provide recommendations as an out of the
box SAAS which can be used by calling an API. Since these AP alternates have to focus on end to end
products, they somewhere lose the focus on quality of big data analysis and that is where AP excels.
These services will provide clients with already built-in libraries that clients have no control to modify
as per their use case. So having this one model fits all use-case approach is not correct as every cli-
ent might have different demographics that will be impacting the sales. AP’s approach to personalize
the model to individual client and not group them as part of segmented solution brings out the quality
recommendations(Expert 1: 2022).

Amazon Personalize (AP) Technical Operating Process

Process Start with first conversation. Every customer that is planning to use Amazon personalized ser-
vice, Amazon does a touch point meeting to set and decide the things. Amazon starts engaging with a
client that’s actually looking to use personalized advertising to improve their marketing. So using that
user segmentation capability of Amazon mentioned earlier AP enables clients to more intelligently target
users that would have an affinity for products that they are trying to promote.
The goal of different AP teams as a whole, including operations, developers, and solution architects
is to make the service so easy that clients don’t need them at all. Because of this tenet AP have a lot of
customers that self adopt the service which means they organically read about the service usings docu-
mentation, blog posts and sample codes. AP is so simple to use that most of the time clients never have
to contact AP teams at all. Being simple is APs core mission.

240

The Role of Big Data Analysis and Machine Learning in Marketing Communications

Moving to the data being consumed in AP services, clients bring their own data. AP is an industry
agnostic tool. Any client can bring their own data and start playing with the recipes provided by AP. AP
requires interaction and behavioral data to generate quality recommendations. These data points can be
as simple as time a user stayed on an item page. AP takes these datasets and trains the model behind the
scene. Different use cases that can benefit from this design are item recommendations based on similar-
ity, popularity and other parameters. Clients have the option to either use already available recipes for
some of these common use cases or create their own. AP provides the ability to clients to create multiple
versions for testing and then productionalize the best model by using it as an API endpoint from their
website to get the inferences(Expert 2, 2022).
The challenges that clients face while working with AP are similar to what they will face with any
ML service. The quality of ML service is usually measured by the quality of its recommendations.
Quality of recommendations depends a lot on the quality of data used to train and test the models. are
usually related to integration issues on either client side or Amazon side. To achieve data quality, cli-
ents usually require to do some feature engineering, data cleaning and other data preparation activities.
AP is very smart and takes care of feature engineering for clients. Clients can bring their raw data and
AP will feature engineer the parameters, perform data splitting for training and evaluation providing
performance metrics. Clients however need to develop capability on their side to cache the results, to
handle the AP recommendation flow, scaling of application, integrating the application with AP API,
and how to productionalize the whole flow. All these tasks need to be taken care of by the client itself.

Selection of Metadata

Selecting the right meta data for the use-case while using AP service can be a game changer in targeted
marketing communication. This allows AP clients to optimize results and generate more customer rel-
evant recommendations. This section discusses how to select the right meta data depending on different
aspects of a use-case. Right metadata plays a crucial role in training and testing of models. High quality
of data ensures high quality results and optimized models. Right metadata allows models to identify
patterns of user behavior and preferences. Due to availability of the same data from multiple sources,
one needs to compare and determine which data set to use and which to ignore. When a customer visits
a website, different aspects of his behavior are captured. This includes data points like how long he was
on a web page, how long he hovered over a certain image, which items he added to his wish list etc. All
these data points are collected but not all will be useful for every use-case. A proper selection can help
in identifying the customer preferences and hence will help in personalized recommendations.
AP has different combinations of models and datasets which are together called recipes. Different
recipes lead to different recommendations. Irrespective of the recipe an AP client chooses, there are three
main type of data sets that AP deals with. These are, user data, item data and user-item interactions. User
data can have metadata like, age, gender, location, ethnicity etc. Similarly, items will have metadata like,
size, color, originated from, price etc. The third and most revealing type of dataset is user-item interac-
tion. This dataset will have metadata like, time spent on the item page, item added to wish list etc. The
minimum number of interactions Amazon Personalize expects in order to start making recommendations
is 1,000 interactions from a minimum of 25 users(Hood and Cleopas, 2020).

241
 
The Role of Big Data Analysis and Machine Learning in Marketing Communications

Table 2. Different AP Data Types. Source: (Hood & Cleopas, 2020)

Dataset Type Required Fileds Reserved Keywords

User USER_ID (string)
Item ITEM_ID (string) CREATION_TIMESTAMP (long)
USER_ID (string) EVENT_TYPE (string)
Interactions ITEM_ID (string) IMPRESSION (string)
TIMESTAMP (long) EVENT_VALUE (float, null)

Usually there are two types of users that visit a website, guest users and member users. It is always
more optimal to use data from members as they are authenticated as well as easy to track then guest or
anonymous user. In AP, each data set has set specific requirements and a schema needs to be defined.
This ensures a proper expected structure of data and parameters which AP can then use.
One of the latest feature that has been launched is user segmentation (AWS re:Invent, 2021) . Clients
can train the models to segment users liking a particular item. So rather than asking what a particular
customer will buy, the problem statement becomes, who are all the prospective customers of a specific
item. This changes the whole game because now rather than sending email of the prospective items to
the customer, clients can send out marketing emails to all the prospective clients of an item. So from a
client’s perspective these will look like generalized communication to a user segment, but from a cus-
tomer’s perspective this will look like a personalized email of the items he/she likes.

Recommendation Models

In the previous section we discuss how careful and well researched data selection can lead to quality
recommendations and optimized models. There are two recommendation models that are used in AP,
collaborative filtering and content-based filtering. Collaborative filtering focuses on the user similari-
ties. If two users show similar patterns and interests, then the recommendations to one user depends on
the products that other similar users have bought. On the other hand, in content based filtering, focus
is one finding similarities between items. The recommendations are the items having similar metadata
as the one that the customer likes. Due to the intrinsic differences in the approach, both these models
used different data. Collaborative filtering uses user data while content based filtering uses item data.
A hybrid model will use both of these data sets.
While a quality data set results in optimized model, quantity of data also makes a big difference too.
Too much data will result in creating noise in model while too less data will create an underperform-
ing model. If meta data that has no relation to user is included in training data set, it makes it harder
for the learning algorithm to find a relevant pattern in the data. So this displays how a wrong approach
in metadata selection can make it hard for algorithms to find patterns and learn anything from data.
Recommendations generated by such a model which is developed using skewed, non-relevant data will
not be as personalized as expected.

Findings and Suggestions

Marketing communication plays a very critical part in sales of any business. Sales in the soul of any
company. A good marketing campaign can make or break the yearly sales revenue. Thus it is of utmost

242

The Role of Big Data Analysis and Machine Learning in Marketing Communications

importance that any insight that can be used to make marketing messages as effective as possible must
be utilized. Big data analysis holds this capability to provide useful insights into customer behavior and
thus helping create personalized marketing communication. From our analysis of different aspects of
using big data analysis to drive marketing campaigns we present some important findings in this section.
One of the most important findings is the impact of metadata on the quality of insights. Specific use
cases require specific types of data and thus it is very important to carefully choose the metadata. This
is challenging as it is not always easy to find the required parameters. In such cases work arounds like
interpolation, mutation and other such techniques can be used to get inferential data which is close to the
actual data. Another important finding is the selection of models to process the data and get insights. A
single model cannot cater to all use cases. So deciding which will be the best model is a challenge. With
this the challenge to train the model in such a way that we can minimize any noise comes into picture.
This is closely related with the quantity and quality of data. To decide how much data will be sufficient
to properly train a model needs a lot of iterations and validations. Last but not the least the requirement
of high speed processing of big data to get quick real-time results is another frontier on which a lot of
advancement is required. All these problems bring up the question of cost-benefit analysis of using big
data insights in making targeted marketing campaigns more efficient and personalized. No doubt it will
bring a lot of value in terms of increased sales and high rate of customer acquisition for the companies.

CONCLUSION

Big data analysis is a very important tool for marketing communication at the point of timely and accurate
understanding and the response of consumers whose needs and expectations are constantly increasing.
With the development of communication technologies and increasing competition, the messages sent to
the consumers and the feedback of the consumers are getting more complex day by day and businesses
have to give more importance to healthy communication. Online commerce is increasing very rapidly,
and many aspects of communication and product recommendations in online stores are becoming more
important. The comments made by the users are read by millions of people within a few minutes, and
millions of replies are received to these comments, and if there is no strategic and instant intervention,
processes leading to the bankruptcy of the enterprises can be experienced. Big data has a very critical
function for effective marketing communication in this environment. Identifying consumer requests,
needs, observing comments, complaints and satisfaction, and as a result, determining communication
strategies will increase the effectiveness of marketing communication for the business while reducing
costs on the other hand. The business will be able to make effective interventions for the right target
audience with the right tools.
In addition to this, attention should be paid to the fact that people prefer to talk to people, especially
when making decisions about digital communication strategies after big data analysis. Until recently,
serious IT personnel, technical equipment, additional personnel, etc. were required by businesses to en-
able them to collect, analyze and interpret big data in millions of channels within their own structure.
Today, outsourced services make big data analytics services relatively accessible to businesses. In the
AP example, free usage rights, discounts, training, protocols to ensure security/privacy are indicators
that infrastructure services will be offered to a much wider audience at affordable costs.
In this study, big data analysis and machine learning in marketing communication are discussed with
different dimensions. In future studies, it is evaluated that big data analysis, which will be carried out

243
 
The Role of Big Data Analysis and Machine Learning in Marketing Communications

by using primary data, will be beneficial to the literature and practitioners in research on the use cases
in the marketing communications of the enterprises, the obstacles encountered while using it, and the
solution proposals.

REFERENCES

Akçan, B. (2016). Big Data: Producers and Consumers. New Media Studies, (2).
Altunışık, R. (2015). Big Data: My Source of Opportunities or New Bundle of Problems? Yildiz Social
Science Review, 1(1), 45–76.
Alzahrani, S. M. (2021). Big Data Analytics Tools: Twitter API and Spark. 2021 International Conference
of Women in Data Science at Taif University (WiDSTaif), 1-6. 10.1109/WiDSTaif52235.2021.9430205
AMA. (2021). American Marketing Association. https://www.ama.org/topics/marcom/
Amado, A., Cortez, P., Rita, P., & Moro, S. (2018). Research Trends on Big Data in Marketing: A Text
Mining and Topic Modeling Based Literature Analysis. European Research on Management and Busi-
ness Economics, 24(1), 1–7. doi:10.1016/j.iedeen.2017.06.002
Amazon. (2021a). What is Big Data. https://aws.amazon.com/tr/big-data/what-is-big-data/
Amazon. (2021b). Personalize. https://aws.amazon.com/tr/personalize/
Ariker, M., Breuer, P., & McGuire, T. (2015). How to Get the Most from Big Data? McKinsey. In Mar-
keting & Sales Big Data, Analytics, and the Future of Marketing & Sales. McKinsey.
Ariker, M., & Manuel, N. (2015). Want Big Data Sales Programs to Work? Get emotional. In Marketing
& Sales Big Data, Analytics, and the Future of Marketing & Sales. McKinsey.
Arrigo, E., Liberati, C., & Mariani, P. (2021). Social Media Data and Users’ Preferences: A Statistical
Analysis to Support Marketing Communication. Big Data Research, 24.
Bendle, N., & Wang, X. (2016). Uncovering the Message from the Mess of Big Data. Business Horizons,
59(1), 115–124. doi:10.1016/j.bushor.2015.10.001
Benoit, D. F., Lessmann, S., & Verbeke, W. (2020). On Realizing the Utopian Potential of Big Data
Analytics for Maximizing Return On Marketing İnvestments. Journal of Marketing Management, 36(3-
4), 233–247. doi:10.1080/0267257X.2020.1739446
Bitnine. (2016). Introduction About Big Data Marketing. https://bitnine.net/blog-useful-information/
introduction-about-big-data-marketing
Blei, D. M., & Ng, Y. A., & Jordan M. (2003). Latent Dirichlet Allocation. Journal of Machine Learning
Research, 3(4–5), 993–1022. doi:10.1162/jmlr.2003.3.4-5.993
Boyd, D., & Crawford, K. (2012). Critical Questıons for Big Data. Information Communication and
Society, 15(5), 662–679. doi:10.1080/1369118X.2012.678878

244

The Role of Big Data Analysis and Machine Learning in Marketing Communications

Buganza, T., Trabucchi, D., & Pellizzoni, E. (2020). Limitless Personalization: The Role of Big Data
İn Unveiling Service Opportunities. Technology Analysis and Strategic Management, 32(1), 58–70. do
i:10.1080/09537325.2019.1634252
Buhalis, D., & Volchek, K. (2021). Bridging Marketing Theory and Big Data Analytics: The Taxonomy
of Marketing Attribution. International Journal of Information Management, 56, 1–14. doi:10.1016/j.
ijinfomgt.2020.102253
Chen, Y., & Xie, J. (2008). Online Consumer Review: Word-of-Mouth as a New Element of Marketing
Communication Mix. Management Science, 54(3), 477–491. doi:10.1287/mnsc.1070.0810
Chwastek, M., Pooniwala, P., & Ashman, S. (2020). Amazon Personalize Now Supports Dynamic Filters
for Applying Business Rules to Your Recommendations on the Fly. https://aws.amazon.com/tr/blogs/
machine-learning/amazon-personalize-now-supports-dynamic-filters-for-applying-business-rules-to-
your-recommendations-on-the-fly/
Culotta, A., & Cutler, J. (2016). Mining Brand Perceptions from Twitter. Marketing Science, 35(3),
343–362. doi:10.1287/mksc.2015.0968
De Mauro, A., Greco, M., & Grimaldi, M. (2016). A Formal Definition of Big Data Based on İts Es-
sential Features. Library Review, 65(3), 122–135. doi:10.1108/LR-06-2015-0061
Dixon, M. (2019). How Netflix Used Big Data and Analytics to Generate Billions. https://seleritysas.
com/blog/2019/04/05/how-netflix-used-big-data-and-analytics-to-generate-billions/
Erevelles, S., Fukawa, N., & Swayne, L. (2016). Big Data Consumer Analytics and The Transformation
of Marketing. Journal of Business Research, 69(2), 897–904. doi:10.1016/j.jbusres.2015.07.001
EU. (2021). Big Data: Definition, Benefits, Challenges (infographics). https://www.europarl.europa.eu/
news/en/headlines/society/20210211STO97614/big-data-definition-benefits-challenges-infographics
Fader, P., & Winer, R. (2012). Introduction to The Special Issue on The Emergence and Impact of User-
Generated Content. Marketing Science, 31(3), 369–371. doi:10.1287/mksc.1120.0715
Fortune. (2021). Fortune 500. https://fortune.com/company/amazon-com/fortune500/
Fosso, W. S., Akter, S., Edwards, A., Chopin, G., & Gnanzou, D. (2015). How ‘Big Data’ Can Make
Big Impact: Findings from A Systematic Review and A Longitudinal Case Study. International Journal
of Production Economics, 165, 234–246. doi:10.1016/j.ijpe.2014.12.031
Gulati, R., & Oldroyd, J. B. (2005). The Quest for Customer Focus. Harvard Business Review, 92–101.
PMID:15807042
Hartmans, A. (2021). Jeff Bezos Originally Wanted To Name Amazon ‘Cadabra,’ And 14 Other Little-
Known Facts About The Early Days Of The E-Commerce Giant. Businessinsider: https://www.busines-
sinsider.com/jeff-bezos-amazon-history-facts-2017-4
Hood, A., & Kleopas, I. (2020). Selecting the right metadata to build high-performing recommendation
models with Amazon Personalize. https://aws.amazon.com/blogs/machine-learning/selecting-the-right-
metadata-to-build-high-performing-recommendation-models-with-amazon-personalize/

245
 
The Role of Big Data Analysis and Machine Learning in Marketing Communications

InsideBigData. (2019). How Amazon Used Big Data to Rule E-Commerce. https://insidebigdata.
com/2019/11/30/how-amazon-used-big-data-to-rule-e-commerce/
Johnson, E. J. (2012). Big Data + Big Analytics = Big Opportunity: Big Data is Dominating the Strategy
Discussion for Many Financial Executives. As These Market Dynamics Continue to Evolve, Expectations
Will Continue to Shift About What Should Be Disclosed, When and to Whom. Financial Executive, 28(6).
Lemon, N., & Verhoef, C. (2016). Understanding Customer Experience Throughout the Customer Jour-
ney. Journal of Marketing, (80), 69-96. doi:10.1509/jm.15.0420
Lottemart. (2021). https://www.lotteon.com/p/display/main/lottemart?mall_no=4&chno=100195&ch_
dtl_no=1000617
Lu, S., Xiao, L., & Ding, M. (2016). A Video-Based Automated Recommender (VAR) System for Gar-
ments. Marketing Science, 35(3), 484–510. doi:10.1287/mksc.2016.0984
Luenendonk, M. (2019). Uses of Big Data in Marketing. https://www.cleverism.com/best-uses-big-
data-marketing/
Ma, L., & Sun, B. (2020). Machine Learning and AI in Marketing – Connecting Computing Power to
Human Insights. International Journal of Research in Marketing, 37(3), 481–504. doi:10.1016/j.ijres-
mar.2020.04.005
Malthouse, E. C., Haenlein, M., Skiera, B., & Zhang, E. W. (2013). Managing Customer Relationships
in The Social Media Era: Introducing The Social CRM house. Journal of Interactive Marketing, 27(4),
270–280. doi:10.1016/j.intmar.2013.09.008
McKinsey. (2015). Marketing & Sales Big Data, Analytics and the Future of Marketing & Sales. https://
www.mckinsey.com
Mihaela, E. O. (2015). The Influence of the Integrated Marketing Communication on The Consumer
Buying Behavior. Procedia Economics and Finance.
Morgan, B. (2018). How Amazon Has Reorganized Around Artificial Intelligence And Machine Learning.
https://www.forbes.com/sites/blakemorgan/2018/07/16/how-amazon-has-re-organized-around-artificial-
intelligence-and-machine-learning/?sh=4f75f1573618
Moro, S., Rita, P., & Vala, P. (2016). Predicting Social Media Performance Metrics and Evaluation of The
İmpact On Brand Building: A Data Mining Approach. Journal of Business Research, 69(9), 3341–3351.
doi:10.1016/j.jbusres.2016.02.010
Netzer, O., Feldman, R., Goldenberg, J., & Fresko, M. (2012). Mine Your Own Business: Market structure
Surveillance Through Text Mining. Marketing Science, 31(3), 521–543. doi:10.1287/mksc.1120.0713
Park, S., & Hwang, K. (2020). Increasing Customer Engagement and Loyalty With Personalized Cou-
pon Recommendations Using Amazon Personalize. https://aws.amazon.com/tr/blogs/machine-learning/
increasing-customer-engagement-and-loyalty-with-personalized-coupon-recommendat ions-using-
amazon-personalize/

246
 
The Role of Big Data Analysis and Machine Learning in Marketing Communications

Saggi, G. M., & Jain, S. (2018). A Survey Towards An Integration of Big Data Analytics to Big Insights.
Information Processing & Management, 54(5), 758–790. doi:10.1016/j.ipm.2018.01.010
Shen, Q., & Villas-Boas, J. (2018). Behavior-based Advertising. Management Science, 64(5), 2047–2064.
doi:10.1287/mnsc.2016.2719
Simpson, J. (2017). Finding Brand Success In The Digital World. https://www.forbes.com/sites/forbesa-
gencycouncil/2017/08/25/finding-brand-success-in-the-digital-world/? sh=774a012626e2
Smith, J. B., & Colgate, M. (2007). Customer Value Creation: A Practical Framework. Journal of Mar-
keting Theory and Practice, 15(1), 7–23. doi:10.2753/MTP1069-6679150101
Suoniemi, S., Meyer-Waarden, L., Munzel, A., & Straub, A. R. (2020). Big Data and Firm Performance:
The Roles of Market-directed Capabilities and Business Strategy. Information & Management, 57(7),
103365. doi:10.1016/j.im.2020.103365
Talend. (2021). Big Data in Marketing 101: Why it’s Important, Where it’s Going, and How to Get
Started. https://www.talend.com/resources/big-data-marketing/
Tang, Q. (2020). Construction of Interactive Integrated Communication Marketing System in Big Data
Era. Journal of Physics: Conference Series, 1550(3), 1550. doi:10.1088/1742-6596/1550/3/032142
Walmart. (2021). https://corporate.walmart.com/our-story/our-locations
Wiencierz, C., & Röttger, U. (2018). Big Data-Based Strategic Communication. https://www.akademis-
chegesellschaft.com/fileadmin/webcontent/Presse/1803_Comm_Director_CW_UR_2018_BD_based_
Strat_Comm.pdf
WorldEconomicForum. (2012). Big Data, Big Impact: New Possibilities for International Development.
http://www3.weforum.org/docs/WEF_TC_MFS_BigDataBigImpact_ Briefing_ 2012.pdf
Yavuz, V., & Erdoğan-Gençyürek, M. (2019). Advertising Applications on Internet News Sites. Current
Marketing Studies. in Image Publishing House.
Zhan, Y., Tan, K. H., Ji, G., & Tseng, L. C. (2018). A Big Data Framework for Facilitating Product In-
novation Processes. Business Process Management Journal, (270), 577–595.

247

View publication stats