2.1 Introduction To Literature Review
2.1 Introduction To Literature Review
LITERATURE REVIEW
15
16 AI in Fashion Industry
synthesis of the collected literature in these two broader fields has been done to
frame theoretical constructs. The frameworks thus derived are discussed in
Section 3.2.
Fashion reflects the sentiments of various people who perceive its significance
in their lives in many different ways. To some, it may mean what models wear
on the ramp; to others, it may mean what people wear daily. In this section, we
have attempted to cover a holistic definition of fashion from different authors
and present 360 degrees view of fashion.
Stone (2001), Tortora and Eubank (2005), and Easey (2009) advocate
fashion to be a phenomenon that is admired and followed by the masses
within a specific locational boundary at any given point of time in an era.
Hence, it can be anything ranging from a popular dance, the words we speak,
a haircut or popular music. For example, if one member of a specific group
decides to pierce an eyebrow, it is not a fashion; however, if many group
members decided to do the same, then eyebrow piercing can be termed as a
fashion within that particular group. Fashion is the behaviour of many
members of a reference group in a similar direction (Yurchisin & Jhonson,
2010). However, another essential characteristic of fashion is rejection.
Fashion demands change. The very survival of the fashion industry depends
upon change. Many consumers must be persuaded or freely choose to replace
their products; otherwise, not functionally obsolete (Sproles, 1981). If the
group members wear a particular clothing product for several years and do not
change it, they may be unfashionable. If a person does not change his
appearance or looks for a more extended period, he is unfashionable at an
individual level. If he changes it rapidly and drastically but not in consonance
with his peer group, he may also be termed unfashionable. In order to be
fashionable, one has to follow his peer group. To summarize, we may say that
the term fashion can be characterized by two words – change and mass
acceptance.
According to Easey (2009), fashion has reached beyond the clothing and
lifestyle industry. Fashion applies to an idea, practice or object (Rogers, 1983).
Fashion can be found in arts, science, food, architecture, cars, movies, medical
treatments, television programmes, books, political thoughts, communication
styles and so forth in the world around us. However, for our research, we have
specifically addressed fashion in the clothing and apparel industry in this
thesis.
Literature Review 17
It is not a surprise that there is a connection between many life areas (e.g.
apparel, cars, food). For example, consumers wearing a particular brand of
clothing may also have a liking and preference for a few varieties of cars, and
there may be a relationship. Hence, we may say that fashion represents the
Zeitgeist of time and place (Zeitgeist is a German word, which means ‘Spirit of
the times’). Historically, obtaining clothing has been difficult than obtaining
food (Kidwell & Christman, 1974). Crane (2000) proposes that consumer
choices of clothing are subjected to people’s interpretations of a specific form
of culture for their purpose that includes healthy social norms about appro-
priate appearance at a particular point of time.
One good thing about Fashion is that she never breaks with her old friends.
She is faithful and reunites with them at a later date. Fashion is cyclic, where
almost everything that has emerged as a new fashion has a predecessor (Terms,
2015). Fashion is evolutionary and has been in existence for ages. It evolves or
changes with time and may take a few years or decades or even a century to
repeat itself. For example, the cycle of changing dress lengths of women from
short to long and again short (Blumer, 1969; Sapir, 1931).
demand and price might be directly proportional to each other for certain
goods (Veblen goods), and a price increase may trigger more purchases
(phenomena of luxury goods, wines, watches, etc.) Later, in 1950, economist
Professor Harvey Leibenstein (1950) opined that such drastic changes or rise
in the consumption pattern might be defined as a ‘non-functional demand’. His
pioneering work comprises two of the most fantastic theories, which explains
fashion behaviour. The ‘bandwagon effect’, deriving from the psychological
need to become a part of the crowd, the ‘snob effect’, arises from the need for
differentiation. The same phenomena have also been highlighted by socialists
like Merton (1949) and Barber and Lobel (1952). In a unique work by Zech
(1975), the bandwagon effect has even been associated with voting in elec-
tions. The author concluded that every election would enjoy being with the
winning side by observing the fellow voters. Since the bandwagon effect
spreads more with publicity and advertisements, ‘greater is the publicity
awarded, greater the chances of bandwagon effect and hence winning’. This
work is of particular importance to us because social media is one of the
newest means of promotion and self-expression, and hence it may expedite the
flow of information through the bandwagon effect.
A psychology scholar, Scitovsky (1976), opined that opposite human
requirements drive demand for fashion. First is the need for newness and
differentiation, and second is the need for social conformance. Stigler and
Becker (1977) throw a different light on the above phenomena and state that
consumer tastes do not vary over time and may not be included in the ‘theory
of demand’.
King and Ring (1980) have tried to draw fashion components on a
2-dimensional plane where one axis represents the fashion object, and the
other highlights the process. Their work defines various parameters such as
dressing styles, architectural designs, popular music, etc., as a part of fashion
objects where the fashion process has been discussed on its movement
parameters from the introduction stage to the stage of popularity. The stages
take the shape of a curve that indicates its introduction through innovation,
early adoption, mass dissemination and eventual declination over a particular
period. This has been reflected in the diagrammatical representation of
Fashion Cycle given in the figure below:
Wasson (1968) explained the dynamics of a fashion product life cycle
(Fig. 1). A fashion product slowly gains acceptance among its followers,
reaches its peak and then sales slowly decline before it goes out of the market.
It originates from a few designer labels (haute couture or high-end fashion)
then goes to the masses as ready-to-wear before it reaches discounted stores.
The fashion product price sharply changes from very high to low so as the
Literature Review 19
There are few studies found in the area of gender-related effects on fashion
behaviour as well. Tigert, King, and Ring (1980), based on their different
cultural studies, stated that the new women’s fashions are monitored by the
women consumers who regularly keep an eye on the changing fashion sce-
nario. They concluded that women have a greater degree of association with
fashion. McCracken and Roth (1989) worked on interpreting apparel codes
and found out that women recognized and were more sensitive to fashion
cues than men (Auty & Elliott, 1998). Few research articles also prove that
men and women grasp advertising cues in different ways (Meyers-Levy &
Sternthal, 1991), in that women are far more aggressive in paying attention
to advertising. Elliott (1994) stated that women follow fashion advertise-
ments more than men. Stith and Goldsmith (1989) also opine the same and
state that women have a greater inclination towards fashion apparel inno-
vations; opinion leaders do not mind spending more than men. There exists a
relationship between dressing and one’s expression of Fashion. That may be
why people prefer the same brands, colours and fashion tendencies on an
everyday basis.
with seduction. Wilson (2007) states that beauty and sex appeal go hand in
hand. However, the impact of beauty on sex appeal varies from person to
person. Flugel (1930) states that as clothes help to conceal and reveal the body,
clothes may trigger the sexual imagination of the opposite gender leading to
strong sexual desires.
Fashion has also been subjected to strict control under political environment of
a country or region. There are several measures by which ruling government/
monarch exercises this control. Dress code, heavy taxation on imports, gov-
ernment policies forbidding wearing of some dresses, consumer laws, social
laws, gender-based discrimination, etc. are some of them (Jhonson, 2002).
Sometimes fashion has also been used as an instrument of protest, for example,
when King of Swaziland made it compulsory for women to wear tassels with a
belt assuming it may preserve their virginity and prevent spread of HIV–AIDS,
Swazi women protested by dropping tassels in front of the royal palace (Russel &
Armstrong, 1985). For ages, political environment constrained and imposed
restrictions on social identities. Joseph and Houston (1986) identify few
important phenomena such as uniforms (for military, police, religious, etc.) and
certain styles of clothing such as straight jackets and corsets for women during
Victorian ages, which irrespective of restricting and constraining individual’s
movements and mannerism, were accepted as mass fashion because of the theory
discussed above.
22 AI in Fashion Industry
The term identity was popularized following Erikson’s (1946) paper and
sparked interest in social scientists in interpreting human behaviour. Stone
(1962) expanded that communication via appearance is more important than
communication via discourse and highlighted that dressing has a priority in
social encounters because it can be seen before communication can be initiated.
Wiegert et al. (1986) also offered an extensive literature review on the rela-
tionship between social identity and oneself. A reference group through dress
communicates identities as it announces the wearers’ social position and the
observers’ during a particular interaction situation. There are a few identities
that are assigned at birth. Antecedents include gender, race, ethnic group, body
variations (e.g. handicap) or kinship group. Through time, an individual
acquires many other identities. Notably, no two people’s identity is the same as
no two individuals acquire the same environmental circumstances. An individ-
ual fulfils a social role, which can be communicated by his taste of consumption
and dressing in particular (Higgins, Eicher, & Johnson, 1995). Stone (1962)
explained how dress contributes to the acquisition of identities. According to
him, people acquire identities by participation or memberships in social rela-
tions. He is keen on how his peers view him and how others will react to his
dress; Stone called this review from a social reference group. For gaining
acceptance, an individual must be able to predict others’ behaviour towards his/
her ways of dressing up that reflects the mood of his/her reference or the peer
group. Many other factors that influence fashion may include family (a family
injects socio-cultural values), occupation, economy, politics, religion, member-
ships and associations, leisure and sports activities, health, hygiene and illness
(Higgins et al., 1995). According to Tarafder (2015), self-esteem, self-image,
gender, age and occupation affect consumers’ choice of fashion. Diana Crane
(2012), in her book ‘Fashion and Its Social Agendas: Class, Gender and Identity
in Fashion’ elaborates that fashion identity for men and women varies with
gender, where men seek masculinity, desire to express their class, lifestyle and
popular culture which they follow, women seek sophistication, want to reveal
sexuality and beauty, however, constrained and bounded by strong social
norms. In various places like the United States and Europe, various aspects of
identity were highlighted by clothing choice. Such identities are influenced by
occupation, regional identity, religion and social class. Crane (2012) also pro-
poses that fashion identity may be further factorized into individual self-identity
and social identity at any given point in time (p. 3). Self-image, a prominent
factor behind self-identity, comprises lifestyle, age, gender, sexual orientation
and ethnicity. Gidden’s (1991) documentation of self-identity with time and
Literature Review 23
These discussions are essential for our present work; in Chapter 3.2, we
have used the above discussions to deduce the conceptual framework of
fashion identity.
Not everything that is on the runway becomes a Fashion. Stone (2008) did an
interesting piece of research to study the process of acceptance of fashion. In a
runway show of fashion brand Prada in 1996, cropped pants were showcased.
Few women started wearing cropped pants in 1997. By 2007, every woman in the
United States has either owned or tried a cropped trousers pair. Hence, he
concluded that the introduction and imitation are essential components of con-
sumers’ psyche in fashion. Fashion is a business of social copying (Stone, 2008).
When a new style is introduced, it either becomes a fad (a small scale and
limited-time fashion) or flop (a style which never gains mass acceptance), a
fashion, or sometimes a classic (something which becomes timeless, e.g. white
shirt or blue denim). A style can never become a fashion until and unless it
gains mass acceptance (Nystrom, 1928). Celebrity or fashion icons have a
significant impact on the fashion business. For example, actress Merlene
Dietrich and Katherine Hepburn introduced pants in social events in the
1960s, when most women in the United States wore either skirts or dresses in
social events (Buckland, 2005).
Fashion directly relates to consumer purchasing power; wherever there is
purchasing power, there is interest in fashion. Fashion determines both the
character and direction of consumption. Factors such as price and the con-
venience of use, durability and craftsmanship also play an essential role in
consumer purchase decisions and its fashion quotient, which should be
according to the prevailing trend that attracts the consumers most (Hsaio &
Dickerson, 1995).
Hayes and Jones (2002) suggest that women’s increased income levels have
significantly contributed to the rise of fashion economic activities through
increased purchase power. They have empirically proved that wants rather
than needs drive consumers to spend more on fashion apparel and clothing.
Hence in an economic depression, fashion is one of the first industries to be
affected. Kilduff (2005) says that the fashion industry has become turbulent
and vibrant with the onset of many brands and rapidly changing styles. While
there are a series of challenges in the fashion industry, any fashion retailer’s
biggest challenge is unsold inventory selling at discounts. Money is lost in
between what does not sell and what sells (Subramanian, 2014). A significant
reason for this remains the wrong forecast. Fashion is a hostile business for
many; especially in the absence of modern forecasting apparatus.
In the business of fashion, faux pas can be costly. To get confidence over their
forecasts, almost every fortune 500 fashion retailer depends on professional
trends forecasting companies such as WGSN, Style sight and Promostyl, to
name a few but the most popular ones. With an average annual fee varying
between $7,000 and $15,000, they offer ideas on colour, fabric and cuts. The
traditional gathering methods include trawling art exhibitions, restaurants,
concerts, events and even scientific journals. Fashion forecasters often
photograph trendy people, travel to fashion cities, follow catwalk shows and
spot vintage as well as upcoming trends. The Internet’s data are proving a
boon to them in the form of cost savings (Gordon, 2013). Another traditional
way of forecasting includes analyzing consumer publications such as Glamour,
Vogue, InStyle, Harper’s Bazaar and GQ (Dickerson, 2007). Companies such
as ITC in India use big data to analyze past designs to bring out the aspects
that have been popular and accepted by the masses and feed this information
back into the design or new product development process, thus strengthening
the designer’s intuition and creativity (Balaji, Rajnish, & Sharma, 2013).
Gardener (1995) compares fashion forecasting with ‘chasing the future with
a butterfly net’. The people who forecast picks up upcoming and prevailing
trends from available resources. (Popcorn, 1991). Product developing com-
panies and forecasting agencies help short-term predictions for colourways,
fabrics, silhouettes and styling directions. Fashion forecasts help merchan-
disers to anticipate change and predict future deliverables (Levenbach &
Cleary, 1981). A trend forecast may identify the source, the reasons, under-
lying pattern, direction and the tempo of the trend. Forecasting comprises
28 AI in Fashion Industry
tools and techniques that can be applied systematically. The six steps of
developing forecasts have been enumerated in the following paragraph.
(1) Analyze past trends and forecasts (2) Analyze the reasons for change as
well as its effects (3) Comparing the past forecast and the actual sales (4)
Identify the cues that will affect the future trends (5) Use forecasting tools and
techniques in order to get accurate and reliable information (6) Continuous
follow-up to diagnosticate any deviation (7) Regular review. Season after
season, product developers, merchandisers and product managers look for
short-term trend forecasts of colours, textiles, silhouettes, and overall style
directions. Forecasting comprises tools and techniques that can be applied
systematically, but among all ‘Scanning’ plays a significant role. Forecasters
look at historical sales data, communication with supply chain partners,
competitive scan, environmental scan, consumer scan and fashion scan, which
eventually translates into colour, fabric, style, silhouette and overall look for
the upcoming season (Keiser & Garner, 2008). The traditional gathering
forecast data included trawling art exhibitions, restaurants, concerts, events
and even scientific journals. Fashion forecasters often photograph trendy
people and celebrities, travel to fashion cities, follow catwalk shows and spot
vintage as well as upcoming trends. Fashion auxiliary services offered by the
media like fashion magazines, general consumer publications, information
resources like fashion consultants and information services are essential
ingredients for new product development and long-term forecasting (Stone,
2001). These works reveal that fashion forecasting’s traditional nature was
based on observation, scanning of the environment, expert judgements and
human intuitions.
In brief, we may tell that forecasting helps convert available information to
form a strong basis of decision-making and minimize risks (Brannon, 2010).
Fashion forecasting has shaped like a big industry with some major players
like WGSN, Promostyl and Style sight in line with mass production and
development of organized fashion retail. A shift has been witnessed in fashion
forecasting from the 1960s. Fashion forecasting is not stating a single domi-
nant fashion trend. Instead, it has emerged as a pluralist approach affected by
the expansion of mass media and increasing sophistication of the consumers.
The period post World War II saw a slower-moving fashion market. During
this time, forecasting was easy. The forecasters used to collect stories and
themes and compiled them into books/magazines that could deliver visual and
tactile information (Mckelvey & Munslow, 2008).
Scanning the outside environment is an essential ingredient of the fashion
ecosystem. It helps to identify the various external factors like economic,
political, social, sustainable, technological and cultural determinants that may
Literature Review 29
affect fashion. Current events, the arts, sports, science and technology affect
short-term forecasting.
Past sales data, supplier’s communication, competition mapping, environ-
ment scanning, consumer scanning and fashion predictive scanning translates
into colours, fabrics, silhouettes and other production elements. This forecasting
may be adjusted according to the target market (Keiser & Garner, 2008).
Fashion auxiliary services offered by the media like fashion magazines, general
consumer publications, information resources like fashion consultants and
information services are essential for new product development and forecasting.
For over 25 years, Promostyl, headquartered in Paris, has predicted fashion
change and helped its clients profit from its advice on international fashion and
design consulting bureau. Linking fashion with social trends is the key to Pro-
mostyl’s methods (Stone, 2001). Fashion scan includes steps such as travelling to
fashion cities (New York, London, Paris, Milan and Tokyo), scanning the
streets, scanning printed and broadcasted material as well as web sources,
meeting fashion leaders and experts from the field of creativity such as arts,
architecture, design, cosmetics and entertainment to identify fashion cues.
Consumer scan is done by first sub-dividing the target consumer into
smaller segments, e.g. working women can be divided into married, single,
number of children, ethnicity or other characteristics. Various data gathering
methods include focus groups, surveys and observations. Point of Sale (POS)
data are used for identifying various price points and quantities. The industry
uses consumer scan for short-term forecasts. Social media can play a more
significant role here as consumers of almost all fashion brands are present on
social media and actively creating and uploading content about their tastes
and preference.
It is imperative to mention the work of Nystrom (1928). Nystrom
endeavoured to list the factors that guide and influence the character and the
dimension of fashion. His framework is represented below:
Nystrom’s framework is used for macro-level forecasting, also called long-
term forecasting.
Many forecasting and market research companies are using the Internet as
a tool for trend and style research. People are looking at designers and retailers
alike websites that can be found easily on the Internet (Next, Paul Smith, La
Redoute, etc.). This emerging area of ‘e-research’ has benefits like speed and
cost savings; however, problems inaccuracy as the Internet population is not
representing a brand’s target population (Easey, 2009).
Other fashion forecasting methods include collecting market information
from fashion editors and buyers, style testing (a consumer involvement tool),
test marketing of new styles, colours and silhouettes.
30 AI in Fashion Industry
Over the last many years, the fashion industry has always been engaged in
different marketing strategies, with traditional cataloguing being made refer-
ence to on proxy without adequate segmental provisioning. This is based on
designers’ intuition with some forecasting tools and instruments (Yang &
Huang, 2011). Thus, the fashion industry relies on previous data, mental
intuition and imaginations to predict and meet the customers’ demands.
However, with the recent fast-dynamic fashion trends, the design market has
become highly competitive and renders the present mechanism obsolete and
irrelevant (Guercini & Runfola, 2016; Jacobs, Cambré, Huysentruyt, &
Schramme, 2016; Roubelat, Brassett, Mcallum, Hoffmann, & Kera, 2015).
The traditional method of brainstorming on designs is time-wasting and
reduces productivity. Hence, it is difficult for fashion designers to place their
designs on mannequins due to outdated styles. There is a high probability that
what they are making is not the trend (Jacobs et al., 2016). Some recent works
have tracked customer demand using different methods through comments,
shares and re-tweets from the social media website Twitter.
The advent of the World Wide Web and social media revolutionized the
fashion forecasting industry. However, the tactile nature of traditional fore-
casting is still predominant in the industry. In the fashion industry, the
dominant trait of non-conformance of behaviours according to academic
standards have given rise to decision-making and analysis as an art; however,
‘a sizeable array of decision aids’ for a smooth decision-making may get
‘capitalized on a big data analytic approach’ (Beach & Lipshitz, 2017; Zhao,
Liu, Zhang, & Huang, 2017).
A few works in existing literature also reveal that researchers have tried to
address this field with some analytical models. Garro (2011) addressed sales
forecasting with fast-fashion retailer Zara by developing regression models
based on past sales data and unsold inventory. Ferreira et al. (2016) used a
similar model but drew regression trees. Huang and Liu (2017) used adaptive
neuro-fuzzy inference system (ANFIS), which outperformed the traditional
artificial neural networks in forecasting demand for fashion products. Some
more studies report the usage of past sales data for forecasting the demand for
fashion, Tehrani and Ahrens (2016) using classification algorithm, Choi et al.
(2014) extreme learning machine and grey model, Ren et al. (2015) Panel data
model and a variety of other predictive models as well as deep learning (DL)
approaches such as Decision trees, Random forest, Support Vector Regression,
Literature Review 31
(Wong, 2012; Zhou et al., 2018). E-commerce and social media are the main
leverages of the fashion industry in recent times. Social networking sites such
as Facebook and Instagram generate tons of data, both structured and
unstructured, that may be analyzed using AI-based algorithms for better
operational decisions such as designing or merchandise planning of certain
products before reaching the peak of their popularity (Franco, 2018).
According to Scoltz (2017), these data have characteristics like volume, vari-
ety, velocity, veracity, variability, complexity and viability, making it humanly
impossible to process and make decisions without a machine interface.
Machine vision first became popular in the area of radiology with its appli-
cations in medical diagnosis. With advances, it was introduced to the fashion
industry with image processing features and its ability to classify images based
on colours, textures, shapes and sizes (King & Lau, 1996). Processing images
to get business insights has become popular in the present era of AI-driven
technologies. Many industries are processing visual images to create tasks that
aid their current practices (Mohapatra, 2019). Many big corporations like
Google, Microsoft and IBM provide image processing-based solutions to the
fashion industry. In April 2018, software giant Microsoft published an article
on their developer blog webpage, where they introduced two of their recent
applications developed jointly with an international fashion retailer. The first
application, named ‘Grabcut’, enables users to extract photographs of fashion
products from typical day-to-day photographs of users. It uses a Gaussian
mixture model with five components and uses an iterative optimization
algorithm that removes the background unwanted noise from the pictures. The
second application, ‘Tiramisu’, is an image segmentation architecture that uses
a deep-learning-based algorithm and binary coding of foreground and back-
ground pixels for classification, factorization and segmentation. Microsoft
claims that this model achieved 94% accuracy during training mode, making it
fit for further use (Microsoft Developer Blog, 2018).
Computer-enabled image processing, by and large, has become a hot topic
in both practitioners as well as academic literature; however, a more system-
atic as well as holistic investigation of the entire fashion system may lead to a
conceptual model, which we have addressed in the analysis and results section.
Design thinking is an approach that uses the designer’s sensibility and methods
for problem-solving to meet people’s needs in a technologically feasible and
commercially viable way. In other words, design thinking is human-centered
Literature Review 33
innovation. People’s need and demand for better performance with less effort
steers the design thinking process that can be best addressed through successful
application of technology in the designing process, often around materials
choices: balancing performance with aesthetics for the ideal product experi-
ence. It requires integration of quantitative technical perspectives with
descriptions of aesthetics that are often variable and culturally dependent. The
concept of aesthetics is associated with philosophical reflections of once
experience of any product, environment or intangible emotional expression of
human beings.
An aesthetic experience is a part of product experience that includes the
entire set of experiences such as emotional experience and meaningful expe-
rience. The experience deals with the degree to which all our senses are
gratified to senses is known as aesthetic experience. The rest of the experience
deals with the faculties of human minds, i.e. cognition and emotions. They
should be conceptually separated either while designing for user experience or
while making an aesthetic design decision (Bairisal & Kumar, 2019). Aesthetic
experience seems to be a soft and subjective attribute to measure, but studies
have shown that its indicators can be computed. The sense of the soft or hard,
cool or warm, dull or bright, matt or shiny can be quantified and linked to
material properties (Brownell, 2014). The physical, chemical and thermal
properties determine the senso-aesthetic attributes of materials (Laughlin &
Howes, 2014).
The value of a product is a measure of the degree to which it meets or
exceeds the expectation of the user in functionality and satisfaction. Simplicity
and restraint can be more appealing than drama and display. There is inter-
action with materials through products, which in the present study is the
fashion product or garment. Aesthetic experience stimulates the sensory
attributes of material. Aesthetic ability arouses interest and kindles the five
senses, particularly the sense of vision. It is through the senses that we expe-
rience materials, their texture, colour and patterns on the fabric. Designers
manipulate these senses and the reactions to each sense to create a product’s
identity.
To measure aesthetic sensitivity, Bairisal and Kumar (2019) have consid-
ered the stimulus to contain only aesthetic component and not take into
account emotional and semantic components such as colours, personal desires,
brand names, brand identity, social, cultural-moral values and symbols. In
other words, aesthetic design stimuli should contain only aesthetic components
such as aesthetic design principle and free from emotional components,
semantic interpretation and symbolic association. Whereas, according to
Desmet et al. (2007), product experience is summation of aesthetic experience,
34 AI in Fashion Industry
Social media and e-commerce in recent years have created a gold mine of data
for researchers and practitioners. Pictures, texts, hashtags, likes, shares and
comments carry pieces of information on users’ behaviours at any given time.
Many fashion brands analyze their consumers’ conversations on Web 2.0
Literature Review 35
where users were also empowered to create content and also share them. This
new age of the web is web 2.0.
Social media and Web 2.0 are different. Web 2.0 is associated with online
applications; social media, on the other hand, are those Web 2.0 applications,
which have a social aspect. For example, Google Maps is a Web 2.0 appli-
cation as we can create content (for example, we can tag our house); however,
it does not have a social side. Hence, it is not a social media website; the same
goes with Wikipedia. Social media has become a phenomenon today, associ-
ated with 80% of global consumer spending way back in 2010 (Nuttney,
2010). One fundamental property of social media is massive content creation
by its users. Facebook users are topping all social media sites in content cre-
ation and spreading (Manovich, 2009). For example, some 350 million pho-
tographs are uploaded on Facebook every day, followed by 100 million hours
of video views (Source: Facebook Newsroom, September 2016).
Some 83% of fortune 500 companies are using Web 2.0 based services to
connect, hear and exchange information on the web with their consumers and
prospective consumers (Naylor, Lamberton, & West, 2012). Hollebeek (2011)
proposes that people want to reveal themselves, their identities and personal
interests in social media. These are then received by other users in this network
who also start behaving the same. Fashion is all about self-expression, and
social media gives immense self-expressions opportunities (Kozinets, 2006).
Some companies like wooplr.com have taken a step ahead and created a
fashion discovery service out of social media. They call it fashion social
commerce. Their fashion curators keep a close eye on trends through the social
media pages of thousands of college-goers, designers and fashion leaders.
Consumers are promoted to post pictures of their outfits, which are then
surfaced and showcased to other consumers and the information of where they
can buy that. Social media is also a tool for the co-creation of value with
customers (Clark & Melancon, 2013). Since value creation is one of mar-
keting’s objectives (Kotler et al., 2008), social media is an essential component
of an organization’s marketing ecosystem. Companies want to have an in-
depth understanding of their customers, facilitated by social media through a
digital medium in the form of digital interactions. Organizations can have a
closer view of their vital segments’ needs, wants and desires, enabling them to
produce better business results in terms of profits, sales and brand image
(Mohammadian & Mohammadreza, 2012). Kaplan and Haenlein (2010), also
Weinberg and Pehlivan (2011) and Bernoff and Li (2008) are repeatedly
emphasizing the usage of social media by companies for enhanced business
results. Marion, Barczak, and Hultink (2014) have suggested that companies
can also utilize social media to design and manage new products. This has
38 AI in Fashion Industry
Social
interac2on
Surveillance/
knowledge Informa2on
seeking
about others
Informa2on
sharing Pass 2me
Why
Social
Media?
Expression of
opinion Entertainment
Convenience
u2lity Relaxa2on
Communicatory
u2lity
forward, one can use Kietzmann’s seven building block to arrive at a specific
business decision by tracking certain social media activities.
Liang (2016) has mentioned four different analytical tools to analyze social
media-based information; however, all of these are related to texts (majorly).
Nowadays, even images and videos are being analyzed to gain business
insights.
In the framework mentioned earlier, the Honeycomb model acts as a prism,
and the outcome of this first stage is then subjected to analysis as a filter with
the subsequent refraction providing the results in the form of interpretation.
The model has a fit with most of the social networking sites.
In another framework by Stieglitz, Dang-Xuan, Bruns, and Neuberger
(2014) (Annexure 3) data pertaining to structural attributes, opinion,
sentiment-related and topic/trend-related information has been advised to
40 AI in Fashion Industry
PRESENCE
The extent to
which users
know if others
SHARING are available
RELATIONSHIPS
The extent to
which users The extent to
exchange, which users
distribute and relate to each
receive content other
IDENTITY
The extent to
which users
reveal REPUTATION
themselves
CONVERSATIONS
The extent to
The extent to which users
which users know the social
communicate standing of
with each other others and
GROUPS content
The extent to
which users are
ordered or form
communities
analyze using tools such as content analysis, sentiment analysis, social network
analysis, statistical analysis and trend analysis. He named this framework as
social media analytics framework. However, the researchers did not mention
any tools or methodologies related to images or videos. Such a framework is
useful in analyzing information retrieved from Twitter or blogs, mostly in
texts; however, social networking sites such as Facebook and Instagram pro-
duce high volumes of unstructured data, such as images and videos. A separate
framework may be needed for such a purpose.
Behesti et al. (2018) proposed and validated a framework named ‘social
media analytics framework’ for trend detection in fashion buying process in
which they used a snowball sample of 74 German-written fashion blogs using
the KNIME analytics, platform text analytics. They successfully identified the
most trending words: BOHO, Fringe, Metallic, 60s, XXL sleeves, etc. Boho
Literature Review 41
Many exciting works in the area of fashion and social media has been done in
recent years. Touchette, Schanski, and Lee (2015) performed a content anal-
ysis of 1,443 Facebook posts over 50 apparel brands. They concluded that
most dominant brands use photos and advertisements in entertainment using
play theme frivolity, which uses fun and short acts through which users are
engaged in particular products, thereby enhancing sales. Shafiq and Imran
(2014) discovered that fashion consciousness and social media are associated,
but they did not specify which social media websites have the highest associ-
ations with the other variables. Ahmad and Ashiq (2015) studied the rela-
tionship between social media and fashion with four variables and tested their
model using regression. Their research proved that there is a scope of pre-
dicting the fashion industry using social media.
Technology has boosted consumer interaction with brands. This has
resulted in high awareness levels and recalls of products and brands engaged in
social media activities. The majority of fashion brands connect to their con-
sumers through Facebook, Twitter, Instagram, Fashion blogs, Pins and You-
Tube videos. In the beginning, luxury brands refrained from using social
media, perceiving that their relationship with the consumers may get diluted;
later on, Gucci launching its line of sunglasses through its social site boosted
the confidence levels of other brands (Kim & Ko, 2012). Marketing through
social networks spreads very vigorously and is often termed viral marketing.
This spread as word of mouth (WOM) from consumer to consumer (Alreck &
Settle, 1995; Arndt, 1967). Many brands are using social media as a tool for
viral marketing. Ahn, Hwang, and Jung (2010) tell that since information
spreads and diffuses faster through social media, some companies are using it
as a promotional channel.
42 AI in Fashion Industry
from his college dorm room to enable his fellow students to communicate
better. The initial name was Facemash, which later changed to ‘the facebook’
and finally Facebook. Facebook had around 1.79 Billion Monthly Active Users
(MAU) and 1.18 Billion Daily Active Users (DAU) worldwide as of September
2016 (Source: Facebook Company Info, September 2016). Today, it has 2.45
Billion MAUs. Surprisingly, more than 90 % of its users are on mobile. Today,
Facebook has become a worldwide phenomenon, with 85% of its users
outside the United States and Canada. Facebook has an interactive user
interface that enables users to engage in discussions, share pictures, send
messages, express likes and dislikes, and hear and watch audio and video. The
United States is Facebook’s No. 1 market with comprehensive coverage. Many
brands are using Facebook for extracting business benefits in the United States
(Arnold, 2006).
According to Clark and Melancon (2013) and Saravanakumar and
Sugantha-Lakshmi (2012), 92% of social media users are on Facebook, fol-
lowed by Twitter 84% and LinkedIn 71%. Also, 71% of online American
adults are active Facebook users (Duggan & Smith, 2014).
From the industry point of view, Facebook may yield better business results
than other social sites. Some 20 million people interact with nearly 1.6 million
brand pages daily (Jeanjean, 2012). Coca-Cola has more than 40 million fans
on its Facebook page (Araujo & Neijens, 2012). Costumers engage with
Facebook brand pages through likes, shares or comments. Customers also
produce and share contents and derivatives of the contents through peer-to-
peer engagement (Bonsón & Ratkai, 2013; Bortree & Seltzer, 2009).
This makes Facebook a crucial area of academic and industrial attention.
India, a demographically young country, has a majority of its working
population under 35. India has the second-highest number of Facebook users,
i.e., 7% of Facebook’s global user base is from India (Deloitte, 2014). India
has got more than 120 million Facebook users. With growing mobile con-
nectivity, it is growing faster. People open Facebook more often than any other
mobile application (Sandberg, 2014).
2010; Thackeray, Neiger, & Keller, 2012). Malhotra (2013) proposed that
shares were more important because they signified that a customer was willing
to publicly align with an organization. The model of Facebook engagement as
suggested by Achen (2017) is given below (Fig. 4):
The above mentioned model was designed for a study of Facebook
engagement in sport, hence may be used as a reference. A similar conceptual
model may be necessary for studying fashion.
Babac (2011) suggested an updated framework of engagement based on
Kietzmann’s (2011) model of seven building blocks. He suggested the Face-
book fit of the existing model and named the Facebook activities, which are
associated with each of the seven building blocks. His model is given forward
(Fig. 5):
Babac’s model proves that Facebook variables such as like, share, post,
comment, status update, profile update, video, link, message, chat and groups
memberships are measurable. Each of these variables may be directed to
measure Facebook engagement of users in the light of Kietzmann’s seven
building framework.
Many of the Facebook activities suggested in his framework may be used as
an independent construct and can be measured with specific items with a
research instrument’s help. This may exactly predict the amount of involve-
ment a specific consumer has with Facebook also customer segments can be
formed.
Like Share
Visit Comment
Facebook
Read Message
Engagement
Presence
facebook
Profile
Status
Sharing Chat
Checkin Relationships
Photos
facebook
Videos Friends
Links Groups
Notes Identity Personal Groups
Status Profile
facebook
Likes Profile
Shared Content
Status
Conversation Likes Reputation
facebook
Status Friends
Shared Content Shared Content
Chat Status
Groups
Messaging Checkin
Likes
facebook
Groups
Personal Groups
Pages (Company,
Community,Fan)
and many more in almost every field of business. A few standard features of
these new-age companies are their heavy dependence on technology, analytics,
automation, AI, social media and a host of other similar features. These
companies are not only limited to forecasting, creating, promoting and
distributing the right products at the right time, in the right way, at the right
price, to the right people but also creating a superior post-purchase experience
to the consumers through 24*7*365 engagement, anytime-anywhere reach,
microscopic level of customizations and business solutions as well as contin-
uous and path-breaking innovations. Professor Kotler and his co-researchers
H. Kartajaya and I, Setiawan (2016), coined the term Marketing 4.0 in his
book (Marketing 4.0: Moving from traditional to digital), explained this
digital shift in companies throughout the world. Marketing 4.0 relies on
communication 4.0, based on transcending the existing communication
framework from awareness to action from awareness to advocacy. Unlike
traditional marketing, where the role of communication in marketing was first
introducing a product (awareness), creating a liking and interest (appeal),
giving consumers clues to buy (ask), making the buying process comfortable
(act), marketers in communication 4.0 increasingly believe in creating brand
advocates hence the existing framework has been extended till advocacy.
These new-age consumers become brand advocates, and they become the voice
of marketers in marketing 4.0. Hence in marketing 4.0, consumers act as the
new sellers or marketers.
Kotler et al. (2016) further explain that whereas marketing 1.0 was
product-centric, 2.0 was consumer-centric, 3.0 was value-driven and mar-
keting 4.0 is digital-centric. Marketing 4.0 is equipped with cognitive tech-
nologies where analytics (Descriptive, predictive, prescriptive and visual) is
embedded in the product itself, enabling the consumer to consume the product
and the related technology. For example, when a consumer tries to book a car
through Uber, Uber’s AI system looks at millions of data points to give the
consumer the best rate, service, route options and availability under time,
distance and cost constraints. Another example can be the listing of products
on an e-commerce platform such as Amazon. It is highly customized, user-
oriented, uses cognitive level intelligence to predict what other products the
consumer may buy, based on detailed analysis of his/her browsing history,
cookies on his browser, last purchase history and so many other factors. The
result is that the consumer can view only those products towards which he
may have the maximum possibility of liking. There is credible evidence that
users in marketing 4.0 associate strongly with brands in co-creation, sharing
and feedback (Świeczak, 2017). Marketers 4.0 use social media-based listening
(brand pages, Twitter, etc.) and social media-based intelligence (what users are
Literature Review 47
talking about them). They also believe in spreading e-WOM and engage
consumers in conversations in their social communities. Other salient social
media features of marketing 4.0 include feedbacks, ratings, reviews, likes,
shares and community-based discussions. Kotler et al. (2016) even said that
social media communities are the new target markets. Zarco et al. (2019)
argue that the rise of big data technologies in recent years has mostly
contributed to the transformation of marketing from 3.0 to 4.0 and
strengthening consumer-brand relationships.
There is a rise of marketing 4.0 based companies in the fashion business as
well. Some of these examples we already discussed in Sections 2.14 and 2.15.
Easey (2009) highlights that product is the central element of a fashion mar-
keting mix. Without having the right or consumer-loveable products, the
conversation may not begin at all. In Section 2.12, we have given a few
examples of fashion enterprises that are rapidly switching from traditional to
digital by using modern day forecasting tools. Thus, the present work may
document fashion companies’ effort in this transition from traditional to the
digital age of marketing 4.0. Using social media data for forecasting, which
has been well-documented in the present work through frameworks and a case
study in the upcoming sections, companies are co-creating fashion products
and consumers. Through social media sites, users upload their photographs
and videos, which contain information on what they wear. Practitioners may
use our framework to co-create products that may be closer to consumer tastes
and preferences, thereby minimizing the risk of failures. Hence, this work may
be considered as an addition to the marketing 4.0 literature for fashion
companies in the area of product 4.0.
(1930), Lavers (1969), Tigert, King, and Ring (1980), McCracken and Roth
(1989), and Meyers and Sternthal (1991) and many more were explored.
These works helped us to understand the mutual connection of different
constructs and their effect on fashion. We developed a deep understanding of
fashion not just from product perspective, rather psychological behaviour of
individuals.
We found that the major challenge of fashion business is wrong forecast
leading to loss of sales, deep markdowns, lost profits, reduced customer
satisfaction and more unsold inventory building. We further explored the
literature on fashion forecasting and explored many works in the recent era
where we understood that the traditional form of forecasting which is based
on human judgement is changing rapidly. Literature in this section mainly
covers some recent development in digital tools and rise of events such as
Human–robot Interface, AI, ML, Natural Language Processing, Artificial
Neural Networks, social media and e-commerce.
From here, we identified few clear research gaps which aided us to decide
our research objectives. We also developed a few frameworks which were
discussed in Section 3.2.1 to 3.2.4.
(1) There was a dearth of literature observed in the area of fashion identity.
The term itself appears to be an under-researched term in the research
literature, even though Erikson’s (1946), Weigert et al. (1986), Higgins
et al. (1995), and Crane (2012) emphasized its importance, more studies
would have certainly contributed in enlarging its scope and would have
helped in defining the background of fashion behaviour.
(2) Concepts of self-identity and social identity and their relationship with
fashion identity through a conceptual design were absent in the previous
literature.
e.g. Gidden (1991), Crane (2012), and Leung et al. (2015). However, a
conceptual and holistic definition of self-identity was yet to be discovered
for establishing its relationship with fashion identity.
(5) Many developments were noticed in fashion forecasting, with the earliest
works as old as Nystorm (1928), significant documentation was found in
different decades with its changing nature. A plethora of works was
noticed in the last 20 years, out of which the last 10 years have been
phenomenal for this industry after the onset of digital tools in fashion
forecasting. However, systematic documentation of these events was
missing, especially in the academic literature. We cited many works in this
dissertation: online articles, podcasts, YouTube videos, journals or books,
clearly proving our argument.
(8) Some of the emerging tools like AI, ML, DL, human–robot interface, etc.,
were noticed to impact the area of fashion forecasting profoundly; how-
ever, a framework and systematic documentation was found to be missing
in the pertinent literature.
(9) Although e-commerce, social media, and Web 2.0 possess to be a ware-
house of fashion information today, a systematic review, documentation
and framework development for promoting future research in this area as
well as taking full advantage of this field for fashion product forecasting is
an area where existing literature seems to be blank.