help 5 pages
pfa
5 pages excluding references(references from academic journal)
Find a quantitative, a qualitative and a mixed-method thesis that may be closely related to the selected topic journal on bigdata. Describe the method used, the sample, the population chosen, was there a survey involved, or a set of questions asked as in a qualitative study. Finally, can you identify the problem the thesis tried to examine.
TRADITIONAL
VS. BIG-DATA FASHION TREND FORECASTING:
AN EXAMINATION USING WGSN AND EDITED
by
Mikayla DuBreuil
A thesis submitted to the Faculty of the University of Delaware in partial
fulfillment of the requirements for the degree of Master of Science in Fashion and
Apparel Studies
Spring
2020
© 2020 Mikayla DuBreuil
All Rights Reserved
ProQuest Number:
All rights reserved
INFORMATION TO ALL USERS
The quality of this reproduction is dependent on the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted.
Also, if material had to be removed,
a note will indicate the deletion.
Published by ProQuest LLC (
ProQuest
). Copyright of the Dissertation is held by the Author.
All Rights Reserved.
This work is protected against unauthorized copying under Title 17, United States Code
Microform Edition © ProQuest LLC.
ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 – 1346
27832243
27832243
2020
TRADITIONAL VS. BIG-DATA FASHION TREND FORECASTING:
AN EXAMINATION USING WGSN AND EDITED
by
Mikayla DuBreuil
Approved: __________________________________________________________
Sheng Lu, Ph.D.
Professor in charge of thesis on behalf of the Advisory Committee
Approved: ___________________________________________________________
Hye-Shin Kim, Ph.D.
Chair of the Department of Fashion and Apparel Studies
Approved: __________________________________________________________
John Pelesko, Ph.D.
Dean of the College of Arts and Sciences
Approved: __________________________________________________________
Douglas J. Doren, Ph.D.
Interim Vice Provost for Graduate and Professional Education and Dean
of the Graduate College
iii
ACKNOWLEDGEMENTS
First, I want to thank Dr. Sheng Lu, my advisor, for providing me the
opportunity to research a topic that is changing the fashion industry and has helped me
grow as a student. I am so grateful for your unparalleled mentorship, constant support,
and critical insight that made this such an exciting, thought-provoking journey. Any
student that has the opportunity to work with Dr. Lu is lucky—he is simply one of the
best! I cannot exaggerate how happy I feel to have been your advisee for the past two
years.
To my committee, Professors Brenda Shaffer and Katya Roelse—thank you for
sharing your insight and helping to make my research a success. Your expertise has
broadened the scope of my study and made it more industry-relevant from both
business and design perspectives.
Additionally, I want to thank Dr. Cao, Director of Graduate Studies, for
encouraging me to pursue the graduate program and making it a reality. I am grateful
for the time he has allowed me to grow these past two years in the University of
Delaware Department of Fashion and Apparel Studies.
Finally, to my friends and family, thank you for your support, smiles, and
laughs along the way. Kendall and Cheyenne—I couldn’t have done it without you
i
v
and I’m so glad that graduate school brought us closer together. Mom and Dad, thank
you for your encouragement and for advocating for my success!
v
TABLE OF CONTENTS
LIST OF TABLES…………………………………………………………………….….
vii
LIST OF FIGURES.…………………………………………………………………….viii
ABSTRACT………………………………………………………………………………ix
Chapter
1
…………………………………………………………………………………………..1
1.1 Introduction ………………………………………………………………………………………………….1
1.2 Research Question ………………………………………………………………………………………..3
1.3 Key Definitions …………………………………………………………………………………………….4
2
…………………………………………………………………………………5
2.1 Fashion Trend Forecasting and Related Theories. ……………………………………………..5
2.2 Using Big Data in the Fashion Industry ……………………………………………………………7
2.3 Debate on the Application of Big Data for Fashion Trend Forecasting ………………..8
2.4 Summary ……………………………………………………………………………………………………10
3
…………………………………………………………………………………12
3.1 Data collection ……………………………………………………………………………………………12
3.2 Data analysis ………………………………………………………………………………………………15
4 RESULTS AND DISCUSSION……………………………………………………..23
4.1 Descriptive Analysis ……………………………………………………………………………………23
4.2 Statistical Analysis ………………………………………………………………………………………31
vi
5 IMPLICATIONS AND FUTURE RESEARCH AGENDAS ………………………………..
35
5.1 Findings……………………………………………………………………………………………………..35
5.2 Implications………………………………………………………………………………………………..35
5.3 Future Research Agendas …………………………………………………………………………….
37
REFERENCES………………………………………………………………………………………………………….40
vii
LIST OF TABLES
Table 1.1 Fashion Trend Forecasting Definitions……………………………………………………….4
Table 3.1 Fashion Trend Forecasts Generated by WGSN ………………………………………….13
Table 3.2 Fashion Trend Forecast by EDITED: Key Factors Considered ……………………15
Table 3.3 Fashion Trend Forecast Coding Scheme …………………………………………………..16
Table 3.4 Pattern definitions ………………………………………………………………………………….17
Table 3.5 Design details ………………………………………………………………………………………..19
Table 4.1 WGSN Fashion Trend Forecasts for Womenswear in the U.S. Market in the
Spring/Summer 2018 Season ………………………………………………………………….23
Table 4.2 EDITED Fashion Trend Forecasts for Womenswear in the U.S. Market in the
Spring/Summer 2018 Season ………………………………………………………………….28
Table 4.3 WGSN and EDITED Trend Forecasts: Results Comparison ……………………….31
Table 4.4 Results of Independent Sample T-test ………………………………………………………32
viii
LIST OF FIGURES
Figure 4.1 Example of WGSN Forecast Coding Illustration (for case #3) ……………………28
ix
ABSTRACT
Traditionally, fashion trend forecasting is conducted through a human-based process
that relies heavily on designers’ artistic viewpoints. However, with the emergence of data
science and the increasing availability of data inputs from consumers, the possibility of
using big data tools to forecast fashion trends is attracting growing interest among the
academia and practitioners in the fashion
industry.
This study empirically evaluated the similarities and differences of the results of
traditional human-based fashion trend forecasts with the ones generated by a big data
tool. Based on the comparison of 20 paired fashion trend forecasts for womenswear in the
U.S. retail market during the 2018 Spring/Summer Season (S/S 2018) generated by
WGSN (i.e., tradtional human-based approach for trend forecasting) and EDITED (i.e., a
fashion big data tool) and by using the independent sample t-test, the study finds that:
First, WGSN and EDITED were able to generate very similar trend forecasts for the
pattern. Second, WGSN and EDITED were able to generate overall similar trend
forecasts for the color. Third, the forecast results by WGSN and EDITED for the design
details were the least similar statistically.
The findings of this study fulfill a critical research gap regarding the feasibility of
using big data for fashion companies’ creative activity. Particularly, the findings suggest
the great potential of using big data tools to aid fashion companies’ forecasts and the
x
creation of new products. Additionally, the results of the study significantly increase our
knowledge of the benefits and limitations of using big data in fashion forecasting.
1
INTRODUCTION
1.1 Introduction
Fashion trend forecasting, which refers to the profession of envisioning future trends
in style and foreseeing consumers’ desires, is crucial to fashion companies’ business
success (Rousso, 2012; Furukawa, Miura, Mori, Uchida, & Hasegawa,
2019).
Fashion
companies use inputs from the trend forecasts to create products that are appealing to
consumers, drive retail sales, and reduce excess inventory (Jackson, 2007).
Traditionally, fashion trend forecasting is done through a human-based process of
examining artistic viewpoints, culture, societal attitudes, and current events to predict
future trends (Grammenos, 2015). In this process, fashion forecasters collect and analyze
inspiration to translate it into themes, design details, colors, and patterns that will be
popular in the next season (Rousso, 2012). Creativity is considered central to this process
as consumers desire originality and uniqueness (Ming Law, Zhang, & Leung,
2004).
However, traditional fashion trend forecasts created by designers also face
limitations and critics. For example, some criticize traditional fashion trend forecasting as
‘opinionated guesswork’ due to designers’ tendencies to rely on their ‘gut feel’ to predict
trends (Tehrani & Ahrens, 2016; Fox, Graul, & Peng, 2018). Particularly, it is of concern
that the traditional designer-based fashion trend forecast involves high business risk
2
because of designers’ over-reliance on their artistic vision and too little inputs from the
commercial point-of-view (Visual Next, 2018; Israeli & Avery, 2018).
The shortcomings of the traditional approach of forecasting fashion trends along
with the emergence of data science have inspired fashion companies to explore new ways
of trend forecasting (Israeli & Avery, 2018; Chaudhuri, 2018). Big data typically refers to
large data sets that require advanced computational tools to analyze (Merriam-Webster,
2019). Fashion retailers have been actively using big data to support their business
operations (Fox et al., 2018). Big data tools, in particular, have demonstrated
effectiveness in helping fashion companies improve product assortment for targeted
markets, more accurately predict future sales, and optimize inventories across all selling
channels (Fox et al., 2018). In comparison, the use of big data to support fashion
companies’ creative activities, such as trend forecasting, remains nascent, yet promising
(Choi & Hui, 2011; Ren, Chan, & Ram, 2017). Potentially, big data tools could create
fashion forecasts and reveal patterns, trends, and predictions in consumer preferences by
leveraging the breadth and large quantities of data and advanced analysis (Joshi, 2018,
para. 2; Ren et al., 2017). These insights may allow fashion companies to more accurately
forecast what particular apparel patterns or colors will be market-popular, as well as the
duration of these trends (Joshi, 2018). Nevertheless, the extent to which big data tools can
be used to forecast fashion trends effectively remains largely unknown (Gaimster, 2012;
Sun & Zhao, 2018).
3
1.2 Research Question
Given the growing interest in exploring the use of big data for creative activities
in the fashion industry and the lack of existing research on the topic, this empirical
study intends to compare the results of traditional designer-based fashion trend
forecasts with the trend forecasts generated by big-data, focusing on evaluating
their similarities and differences. The study is important because:
First, current literature on the use of big data tools in fashion companies focuses
on merchandising strategies such as improving product assortments, markdown
optimization, and producing sales forecasts (Joshi, 2018; Visual Next, 2018). Whereas,
the results of this study will significantly increase our knowledge of the benefits and
limitations of big data as a creative tool to fashion forecast. Second, both fashion
companies and educational institutions can benefit from this study by gaining a greater
understanding of how technology and data are changing the fashion industry, as well as
the role of creativity and traditional methods in fashion forecasting. Additionally, the
findings of the study will produce valuable empirical results that can inform how
business-minded individuals, creatives, and educators alike can maximize the usage of
big-data based fashion trend forecasting as well as traditional methods.
4
1.3 Key Definitions
For the purpose of this research, we define the following terms, as shown in Table
1.1:
Table 1.1 Fashion Trend Forecasting Definitions
Terms Definition
Fashion Trend
Forecasting
“the practice of predicting upcoming trends based on past and
present style-related information, the interpretation and analysis
of the motivation behind a trend, and an explanation of why the
prediction is likely to occur” (Rousso, 2012, p. 296)
Traditional Fashion
Trend Forecasting
Fashion trend forecasting, which is produced by humans through
creative methods, where artistic viewpoints, culture, societal
attitudes, and current events serve as inspiration (Grammenos,
2015).
Big-data based
Fashion Trend
Forecasting
Fashion trend forecasting, which is generated by big data tools
through artificial intelligence and quantitative analysis
(EDITED, 2019).
5
LITERATURE REVIEW
This section will go over the pertinent fashion theories, the current application of
big data in the fashion industry, and critically evaluate the possibility of using big data as
an alternative to the traditional designer-based approach to forecast fashion trends from a
theoretical perspective.
2.1 Fashion Trend Forecasting and Related Theories.
Conventional fashion forecasting is based on human-centric methods, where
forecasters examine the world around them—from culture, business, and arts to science
and technology (Gaimster, 2012). Typically, the forecasters would gather inspirations
from multiple design disciplines to spur creative thinking and predict consumers’ desires.
Forecasters are required to gain deep cultural insights into upcoming trends and to predict
trends through curtailing gathered information on existing fashions (Gaimster, 2012).
Ultimately, fashion forecasters’ role is to recognize upcoming trends, produce trend
reports, and provide methods of implementation to improve business’ product lines and
sales (Kim, Fiore & Kim, 2013).
6
Fashion theories suggest that three aspects, namely color, pattern, and design
details, are the most critical components for trend forecasting (Blaszcyk & Wubs, 2018;
Jackson, 2007). First of all, color (such as blue, yellow and red) is essential to fashion
trend forecasting since it is a building block for creating trend-right, and top-selling
designs (Blaszcyk & Wubs, 2018). Studies find that when shopping for clothing, both
online and offline, consumers are often initially attracted to a garment by its color (Park,
Kim, Funches, & Foxx, 2012). Through differentiating the fashion colors that will be
popular either long-term throughout the whole season or short-term that rapidly rise,
peak, and fizzle out, companies will be able to place the right product design at the right
time in the market (King, 2012, p. 540). Empirical studies also suggest that the accuracy
of color forecasting can be critical to a garment’s ultimate success in retail sales (Choi,
Hui, Ng, & Yu, 2012).
The second aspect is the pattern (such as stripes, spots, and checks), which refers
to ‘a repeated decorative design that can be printed, stitched or woven into a fabric’
(Ambrose & Harris, 2007, p.182). Like the color, the pattern also plays a vital role in
fashion trend forecasting since it is another primary clothing attribute that influences a
consumer’s purchasing decision (Rousso, 2012). For example, in 2018, garments and
footwear that used leopard print proliferated among young consumers, helping fashion
brands and retailers that carried such products achieve substantial business success (Yau,
2018). As another evidence of the importance of trend forecasting for pattern, data
indicates that apparel featuring floral and lace patterns overall achieved a higher
7
inventory turnover and greater profit margins than those with spots and stripes patterns in
the U.S. retail market between 2017 and 2018 (EDITED,
2019).
The third aspect is the design details, which refer to the fabric type and shape of
the clothing, such as denim for fabric and a round neck for shape (Jackson, 2007). The
design details affect nearly every product offering, from jackets to dresses, making it a
crucial component of fashion trend forecasting (Resnick & Montania, 2003). Successful
forecasting for design details further allows retailers to maximize profits by
implementing them in multiple styles for a period that may even extend beyond the
current season (Jackson, 2007).
2.2 Using Big Data in the Fashion Industry
With the increasing availability of data inputs and the advancement of related
analysis tools, fashion companies have begun to take advantage of the threshold of
opportunities provided by big data to improve their business operations (Thomassey &
Zeng, 2018). Studies show that the current application of big data by fashion companies
is particularly popular in the business aspects, such as demand forecasting, pricing
optimization, supply chain management, and consumer behavior analysis (Silva, Hassani,
& Madsen, 2019). Scholars from multiple academic disciplines also have been
developing mathematical models and algorithms to explore new ways of using big data to
solve specific fashion business problems, such as improving speed to market and
controlling the inventory level (Choi & Hui, 2011; Ren, Chan, & Ram, 2017; Boone,
Ganeshan, Jain, & Sanders, 2019). For example, Choi & Hui (2011) develops the
8
Extreme Learning Machine Fast Forecasting model to yield more accurate sales demand
forecasts for fashion companies based on inputs from a wide selection of sales data at the
stock keeping unit (SKU) level. Similarly, Pantano, Giglio, & Dennis (2019) use machine
learning to analyze social inputs from consumers’ tweets rather than sales data to
understand the popularity of different fashion brands. However, the research methods
applied and the source of data used by the existing studies varied significantly, making it
challenging to compare the results and evaluate their application values to the fashion
industry.
On the other hand, a growing number of big data analysis tools dedicated to the
fashion apparel sector, such as EDITED and Trendalytics, are launched to the market.
These tools, in general, allow fashion companies to more effectively leverage data
collected from multiple sources, such as social media and e-commerce websites, to
improve their decision makings in product assortment and pricing (EDITED, 2019;
Trendalytics, 2019). Despite these big data tools’ popularity among industry users,
however, their usage by academic studies remains rare.
2.3 Debate on the Application of Big Data for Fashion Trend Forecasting
Compared with its much broader application in the business aspects, the use of
big data for fashion trend forecasting as a creative activity is still at its nascent stage, yet
remains promising (Choi et al., 2014; Israeli & Avery, 2018; & Ren et al., 2017). On the
one hand, some studies suggest that because consumers’ fashion taste stays relatively
stable over time, it is feasible to use historical data such as purchasing history to predict
9
what fashion patterns, colors or styles consumers may like in the future (Gaimster, 2012;
Israeli & Avery, 2018). The studies advocating the potential of big data-based fashion
trend forecasting also emphasize the ability of data science to target upcoming trends and
allow companies to more quickly create popular and best-selling items based on concrete
numbers (Joshi, 2018). Particularly, based on leveraging intelligence contained in
historical records, the big data-driven trend forecasting provides companies with
opportunities to gain more insights into the market dynamics, and better understand
consumers’ tastes, wants and lifestyles, helping to improve the accuracy of trend
forecasting than otherwise (Brannon, 2010; Chaudhuri, 2018). Further, some pioneering
studies have proposed theoretical procedures and methods to use big data tools for trend
forecasting in specific areas, such as color, although few empirical research using real
market data has been conducted so far (Gu & Liu, 2010; Thomassey & Zeng, 2018).
However, the big data-driven trend forecasting does not come without concerns
and challenges. First, despite its powerful advanced machine learning techniques, big
data has demonstrated difficulty in understanding and processing critical cultural factors
that influence fashion trends, such as societal attitudes, movements in politics, ethics,
emotion, and aesthetics (McDowell, 2019). Second, since big data does not have original
creative thinking skills or an aesthetic perspective, it is of concern that big-data based
trend forecasts may not be able to satisfy consumers’ desires for fresh, out-of-the-box
designs (Ming Law et al., 2004). Additionally, even a small and medium-sized retailer
today could carry hundreds of thousands of clothing in different categories, styles,
patterns and colors (EDITED, 2019). Such a large number of apparel items newly
10
released or sold every season makes it highly challenging for big data to pinpoint and
narrow down what design details will be the most popular, given the similar data inputs
used to generate the forecasts (Arte, 2017).
2.4 Summary
In summary, currently, there is no consensus regarding the possibility and
reliability of using big-data tool generated fashion trend forecasting over the traditional
fashion designer-based methods. The studies advocating the potential of big data
frequently emphasize the ability of data science that can quickly target upcoming trends
and allow companies to create popular, best-selling items based on concrete numbers
(Joshi, 2018; Israeli & Avery, 2018). However, other research contends that big-data
based fashion trend forecasts can be unreliable due to its limitations to assess cultural
data and satisfy consumers’ desire for the original product (Ming Law et al., 2004).
Studies that support traditional trend forecasting methods further stress the importance of
forward-thinking designers due to the unpredictable nature of fashion (Ming Law et al.,
2004).
Additionally, despite the heated debate, there is a lack of empirical studies that
evaluate the effectiveness of big-data based fashion trend forecasting compared with
traditional fashion trend forecasting methods.
This study will fulfill these critical research gaps, and provide fashion retailers
and academic institutions a greater understanding of the possibility, reliability, benefits,
and limitations of using big data tools to forecast fashion trends. Overall, the empirical
11
results of this study will enhance fashion retailers’ fashion trend forecasting decisions as
well as academic institutions’ curriculum on big data and fashion trend forecasting.
12
METHODS AND DATA
Given the competing theoretical views regarding the feasibility and effectiveness
of using big-data tools for fashion trend forecasting and the lack of empirical studies, this
study intends to fulfill this research gap. Specifically, the study will focus on evaluating
the similarities and differences of the results of fashion trend forecasts generated by the
traditional approach versus those based on using the big-data tool.
3.1 Data collection
For fashion trend forecasts generated in a traditional approach, this study
collected data from WGSN, a world-renowned service provider for fashion trend
forecasting (WGSN, 2019). WGSN’s trend analysis is consistently cited throughout
academic literature as one of the most trusted sources for traditional fashion forecasts
(Jackson, 2007; Rousso, 2012).
Specifically, based on data availability, this study used all the 20 fashion trend
forecasts created by WGSN for womenswear in the Spring/Summer 2018 (S/S 18) season
targeting the U.S. retail market. Notably, the United States is one of the largest apparel
consumption markets in the world, and womenswear typically accounts for over 60
percent of apparel products available in the market (EDITED, 2019).
13
As shown in Table 3.1, each of these 20 forecasts focuses on a particular product
category (such as ‘Dresses & Skirts’) during a specific time-segment of the S/S 18 season
(such as ‘Spring transitional’).
Table 3.1 Fashion Trend Forecasts Generated by WGSN
Case Market Season Product category
1 Womenswear Spring Transitional Trousers & Shorts
2 Womenswear Spring Transitional Jackets & Outerwear
3 Womenswear Spring Transitional Swimwear
4 Womenswear Spring Transitional Dresses & Skirts
5 Womenswear Spring Transitional Knitwear
6 Womenswear Spring, Mid-Spring & Festival Trousers & Shorts
7 Womenswear Spring, Mid-Spring & Festival Jackets & Outerwear
8 Womenswear Spring, Mid-Spring & Festival Swimwear
9 Womenswear Spring, Mid-Spring & Festival Dresses & Skirts
10 Womenswear Spring, Mid-Spring & Festival Knitwear
11 Womenswear Summer & High Summer Woven Tops
12 Womenswear Summer & High Summer Trousers & Shorts
13 Womenswear Summer & High Summer Jackets & Outerwear
14 Womenswear Summer & High Summer Dresses & Skirts
15 Womenswear Summer & High Summer Knitwear
16 Womenswear Summer Transitional Swimwear
14
17 Womenswear Summer Transitional Trousers & Shorts
18 Womenswear Summer Transitional Jackets & Outerwear
19 Womenswear Summer Transitional Dresses & Skirts
20 Womenswear Summer Transitional Knitwear
Data source: WGSN (2019)
In correspondence with these 20 traditional fashion trend forecasts, we generated
another 20 comparative trend forecasts by using EDITED, a big data tool, which covers
the real-time pricing, inventory and assortment information of over 30 million apparel
products at the SKU level sold by more than 90,000 brands and retailers in the U.S.
market since 2016 (EDITED, 2019). Each trend forecast generated by EDITED focused
on the exact same product category of womenswear targeting the U.S. market during the
exact same time segment of the S/S 18 season as their paired WGSN forecast. As
explained in Table 3.2, when using these millions of data points to generate trend
forecasts, we asked EDITED to consider three factors deemed as the most critical for
market-popular fashion items, including inventory level, retail price and markdown, and
frequency of replenishment (Sterlacci & Arbuckle, 2009; Heching, Gallego, & van
Ryzin, 2002; Burns, Mullet
& Bryant, 2016).
15
Table 3.2 Fashion Trend Forecast by EDITED: Key Factors Considered
Factors Justification
Inventory level The apparel items need to maintain a high inventory level and
always be in-stock during the examined period. The high stock level
implies that retailers were willing to carry these items because of
their popularity among consumers (Sterlacci & Arbuckle, 2009;
Balar, Malviya, Prasad, & Gangurde, 2013).
Pricing and
discount
The apparel items were always sold at the full price during the
examined period, which implies that retailers did not need to use
markdown to drive more sales because of the genuine popularity of
the items (Heching, Gallego, & van Ryzin, 2002).
Replenishment The apparel items need to be replenished at least once during the
examined period. Having replenishment suggests that retailers
intended to carry sufficient inventories and avoid stockout for these
items because of their popularity among consumers (Burns, Mullet,
& Bryant, 2016).
Note: For the purpose of the study, in addition to the factors stated above, the apparel
items shall also meet the following criteria: 1) womenswear; 2) sold in the U.S. retail
market.
3.2 Data analysis
Following Jackson (2007)’s principles of fashion forecasting, this study first
conducted a content analysis of each of these 40 paired trend forecasts (i.e., 20 generated
by WGSN and 20 generated by EDITED) and coded their respective predictions for the
16
color (such as ‘red’ and ‘green’), patterns (such as ‘Long Sleeve’ and ‘V-neck’) and
design details (such as ‘Stripes’). To generate objective and unbiased results, we
developed a coding scheme based on EDITED’s handbook for key terms and the industry
common practices as detailed in Table 3.3 (Saldana, 2015). Tables 3.4 and 3.5 specify
definitions for pattern and design details.
Table 3.3 Fashion Trend Forecast Coding Scheme
Design details Pattern Color
Long Sleeve Plain Black
Peter Pan Collar Patterns Grey
Round Neck Stripes Maroon
V Neck Spots Red
Flat Checks Pink
Heeled Floral Fuchsia
3/4 Length Sleeve Lace Purple
Beading Animal Blue
Denim Camo Navy
Fringing Conversational Teal
Jewels/Gems Aztec Aqua
Leather (incl faux leather) Geometric Green
Longline Graphic Lime
Maxi/Long Length Tile Yellow
17
Metallic Paisley Orange
Metallic Abstract Copper
Mini Length/Size Brown
Peplum Gold
Sequins Neutral
Sleeveless Silver
Tropical Pattern White
Velvet
Table 3.4 Pattern definitions
Patterns Definition
Plain
Refers to garments that contain no pattern (i.e. solid in color with no
recurring
motif)
(EDITED, 2019).
Patterns
Refers to garments that contain a pattern (i.e. displays a recurring
motif) (EDITED, 2019).
Stripes “Parallel bands of color” (EDITED, 2019).
Spots “Uniform circles that repeat themselves” (EDITED,2019).
Checks
“Gridded prints such as houndstooth, tartan or gingham” (EDITED,
2019).
Floral
“Prints that have flowers, including tropical but excluding lace”
(EDITED, 2019).
Lace “The fabric of lace or the print of lace (EDITED, 2019).
Animal
“Garment is made to resemble the pattern of the skin and fur of
an animal such as a leopard, cheetah or zebra” (EDITED, 2019).
18
Camo
“Design of irregular patches of dull colours (such as browns and
greens), as used in military camouflage” (EDITED, 2019).
Conversational
“Any print with a recognizable picture in it, such as cats, birds or
stars” (EDITED, 2019).
Aztec “Tribal inspired patterns” (EDITED, 2019).
Geometric
“Geometry is the use of straight lines and shapes to create a pattern.
For example, a garment printed with rectangles, hexagons or
squares” (EDITED, 2019).
Graphic
“Logos, images or text normally placement printed on the front centre
of the garment” (EDITED, 2019).
Tile
“Normally flowery design within a grid repeated over and over”
(EDITED, 2019).
Paisley
“A droplet-shaped motif of Persian origin” (Ambrose & Harris, 2007,
p.182).
Abstract
“The opposite of geometric. Painterly, abstract patterns” (EDITED,
2019).
Note: Pattern refers to “a repeated decorative design that can be printed, stitched or
woven into a fabric” (Ambrose & Harris, 2007, p.182). For pattern identification,
EDITED uses powerful image recognition software to analyze the product’s image and
notes the listed name of the pattern on the retailer’s website (EDITED, 2019).
19
Table 3.5 Design details
Design details Definition
Long Sleeve “a part of a garment covering an arm,” from shoulder to wrist
(Merriam-Webster, 2019).
Peter Pan
Collar
“a usually small flat close-fitting collar with rounded ends
that meet at the top in front” (Merriam-Webster, 2019).
Round Neck a circlular-shaped neck of a garment (Merriam-Webster,
2019).
V Neck “a V-shaped neck of a garment” (Merriam-Webster, 2019).
Flat “of a shoe heel : very low and broad” (Merriam-Webster,
2019).
Heeled “an element called a top piece that is added to the rear end of
the sole of a shoe, lifting the back of the shoe away from the
ground” (Ambrose & Harris, 2007, p.130).
3/4
Length
Sleeve
“a part of a garment covering an arm,” from shoulder to mid-
forearm (Merriam-Webster, 2019).
Beading “Beads that are attached to a fabric for decorative or other
purposes” (Ambrose & Harris, 2007, p. 34).
Denim “Durable, cotton fabric, primarily used to make jeans” (Ambrose
& Harris, 2007, p. 90).
20
Fringing “an ornamental border consisting of short straight or twisted
threads or strips hanging from cut or raveled edges or from a
separate band” (Merriam-Webster, 2019).
Jewels/Gems “Decorative objects worn on the person or clothes, often made
with precious metals such as gold, silver and platinum and
gemstones “ (Ambrose & Harris, 2007, p. 142).
Leather (incl
faux leather)
A material produced from the tanned hides and skins of many
different animals, but usually cattle, sheep, pig and goat
(Ambrose & Harris, 2007, p. 151).
Longline Hems that are longer than the normal length for a particular
garment (Rayment, 2015).
Maxi/Long
Length
“an ankle-length skirt” (Ambrose & Harris, 2007, p. 169).
Metallic “resembling metal: such as having iridescent and reflective
properties” (Merriam-Webster, 2019).
Mini
Length/Size
“A short skirt with a hemline that is typically at least 20cm or
eight inches above the knee” (Ambrose & Harris, 2007, p. 169).
Peplum “a short section attached to the waistline of a blouse, jacket, or
dress” (Merriam-Webster, 2019).
Sequins “a small plate of shining metal or plastic used for
ornamentation especially on clothing” (Merriam-Webster,
2019).
21
Sleeveless Apparel without “a part of a garment covering an arm”
(Merriam-Webster, 2019).
Tropical
Pattern
Prints that resemble tropical environments, such as “plants,
fruits, and animals” (The Fashion Folks, 2016).
Velvet “A tufted fabric made from silk, cotton or synthetic fibres with
evenly distributed cut threads to give a short, dense pile“
(Ambrose & Harris, 2007, p.107)
Note: “Design details” refers to aspects of a garment’s shape and fabric type (Jackson,
2007). Examples of garment shape include necklines, hem lengths, and collars (Rousso,
2012).
To locate products with specific design details, EDITED searches for products that
include the term in its description (EDITED, 2019).
Then, we rated the degree of similarities of the results of the paired trend forecasts
generated by WGSN and EDITED according to the following coding scheme:
• Value=2 (Very similar): EDITED’s forecast matched more than 50% of WGSN’s
forecast
• Value =1(Similar): EDITED’s forecast matched some but less than 50% of WGSN’s
forecast
• Value =0 (Different): EDITED’s forecast matched none of WGSN’s forecast
Next, we used the independent samples t-test for the rating scores in the previous
step to evaluate how statistically significant is the similarity of WGSN and EDITED’s
fashion trend forecasts (Ott & Longnecker, 2015). The independent samples t-test is
22
widely used by researchers to determine whether the means of variables between two
different groups are significantly different (such as North, de Vos, & Kotze, 2003 and
Jung & Shen, 2011). For this study, the WGSN forecasts and EDITED forecasts were the
two independent groups while the rating scores for the color, pattern, and design details
served as the variables for mean comparisons.
23
Chapter 4
RESULTS AND DISCUSSION
4.1 Descriptive Analysis
Table 4.1WGSN Fashion Trend Forecasts for Womenswear in the U.S. Market in the
Spring/Summer 2018 Season
Case Design Details Patterns Colors
1 Denim Plain, Stripes White, Neutral, Brown,
Grey, Black
2 long sleeve, VNeck,
Sleeveless, Leather,
Maxi/long
length
Plain Black, Grey, Neutral,
Brown, White
3 metallic, round neck,
sleeveless
plain, abstract,
geometric
red, copper, neutral,
brown, white, green,
black, silver
4 long sleeve, maxi/long
length, mini length/size
plain, checks,
stripes
blue, brown, white,
neutral, black
5 long sleeve, round neck
stripes, plain,
geometric
brown, white, red,
neutrals, blue
6 Denim plain black, white, brown,
neutrals
7 long sleeve, denim lace red, pink, orange, blue,
navy, white, brown
24
8 sleeveless, fringing, v
neck, round
neck
geometric, stripes,
lace, floral, tile
black, white, red, pink,
orange, blue, navy,
copper
9 round neck, maxi/long
length, long sleeve,
sleeveless
plain, stripes, lace red, pink, orange, blue,
navy, white, copper,
black, brown
10 long sleeve, fringing stripes, plain blue, white, red, pink,
orange, navy, brown
11 Longline, tropical pattern,
sleeveless
Stripes, plain, floral White, black, pink,
orange,
green
12 Tropical pattern Floral, plain Black, pink, green,
blue,
neutral
13 Maxi/long length, tropical
pattern
Floral, plain Teal, green, neutral,
pink, blue
14 Maxi/long length, long
sleeve, ¾ sleeve, v neck
plain White, blue, neutral,
green,
pink
15 Long Sleeve, VNeck,
Maxi/long length, round
neck, sleeveless
Plain, stripe Pink, blue, navy,
grey
16 Metallic plain Black, white, neutral,
maroon, copper
17 Others* plain Purple, gray, black,
white
18 Denim, longline, long
sleeve
plain Blue, navy, black,
green
19 Maxi/long length plain Black, blue, grey,
maroon, neutral
25
20 Longline, VNeck, Long
Sleeve
Plain, stripes Blue, neutral, white,
pink
Note *: No design details match options listed in Table 2.1; analyzed based on trend
forecasts from WGSN (2019)
The coded WGSN spring 2018 forecasts are summarized in Table 4.1. A detailed
coding example is illustrated in Figure 4.1.
According to the results, first, overall, most fashion trends seem short-lived and
unstable, with few spanning the entirety of the S/S 18 season. While a few key trends in
design details, pattern, and color were suggested to stay throughout spring, many other
trends were concentrated into shorter, distinct time periods, such as ‘Spring Transitional’
and ‘Summer & High Summer.’ This result is far from surprising. As suggested by
previous studies, specific design details, color, and pattern are often isolated to certain
time periods of the season due to weather or social events. Jackson (2007) describes these
time periods as ‘user occasions’ or when shoppers seek an item due to changing lifestyle
activities, weather, or special events. For example, retailers consistently offer a surplus of
festive party dresses in December to accommodate for the holidays (Blaszcyk & Wubs,
2018). Likewise, WGSN forecasted that ‘tropical pattern design details, floral patterns,
and pink, green, and blue colors’ were going to be popular for women’s trousers & shorts
during ‘Summer & High Summer’ in S/S 18—a period when many consumers go on
tropical vacations or to the beach. Therefore, to satisfy the ‘user occasion’ of going on
tropical vacations or to the beach, these trend predictions were isolated to ‘Summer &
High Summer,’ rather than the entirety of S/S 18, consistent with Jackson’s (2007) trend
theory. Understandably, it is not rare to see fashion retailers constantly add new products
26
during different periods in a season to accommodate consumers’ desire for freshness and
novelty (Ming Law et al., 2004).
Second, among the three dimensions of fashion trend forecasts examined, the
pattern had the highest trend stability, whereas the design details seem to be most
unpredictable. For instance, as shown in Table 3, the pattern ‘plain’ is present in almost
every forecast, and the pattern ‘stripes’ is present in nearly half of the forecasts,
evidencing their relative stability throughout the S/S 18 season. By contrast, most
individual design details are present in less than 20 percent of the total forecasts for the
S/S 18 season, suggesting the propensity of design details to be much less predictable.
Additionally, it is interesting to see color forecasts demonstrate the greatest diverse
results with a greater number of predictions than either design details or patterns.
Furthermore, the results of trend forecasts seem to vary substantially among
product categories. For example, of all product categories examined, knitwear appeared
to have the most stable trends—with ‘stripes, and plain patterns, long sleeve design
details, and blue color’ predicted to be popular throughout the S/S 18 season. Since
knitwear is often made up of classics or staple garments that serve to be worn year after
year (Brannon, 2005, p. 62), trends in the design details, pattern, and color for knitwear
understandably were forecasted to be more consistent. By contrast, swimwear, along with
jackets and outerwear, had the most variability in trend, with almost none predicting
trends that lasted all season. Since swimwear is worn less frequently and is often bought
for short trips or vacations, its designs are typically unique with a relatively higher trend
variability across different periods (Brannon, 2005, p. 61).
27
28
Figure 4.1 Example of WGSN Forecast Coding Illustration (for case #3)
Table 4.2 EDITED Fashion Trend Forecasts for Womenswear in the U.S. Market in the
Spring/Summer 2018 Season
Case Design Details Patterns Colors
1 Denim plain, stripes, floral,
checks
Black, white, grey, navy,
blue, brown
2 Leather, long sleeve, denim
Plain, stripes, floral,
graphic
Black, white, blue, navy,
grey
3 No result* plain, stripes, Aztec,
floral
Black, white, blue,
brown, pink, grey, red
4 Sleeveless, mini length/size,
V Neck
plain, lace, floral,
stripes
Blacks, white, pink,
grey, red, blue, navy
5 long sleeve, round neck, V
neck
Plain, graphic,
stripes, aztec
Black, white, grey, blue,
pink
6 Denim Stripes, floral, checks Black, white, grey, blue
29
7 Long sleeve Plain, checks, floral,
graphic, stripes,
lace
Black, white, navy, blue,
grey
8 No result*
Plain, floral, stripes,
abstract
Black, white, blue, pink,
brown, green
9 V neck, sleeveless, mini
length/size, maxi/long
length
Plain, floral, lace,
stripes
Black, white, pink, blue,
red, grey
10 Long sleeve, round neck, V
neck, sleeveless, metallic
Plain, graphic, stripes Black, white, grey
11 V neck, long sleeve,
sleeveless
Plain, floral, stripes,
checks, graphic
Black, white, navy, blue,
neutral, yellow, green
12 No result*
Plain, stripes, checks,
floral, spots
Black, blue, navy, white,
neutral, green
13 Long sleeve, leather Plain, checks, stripes,
animal
Black, blue, grey, white,
green, navy, pink,
neutral
14 Sleeveless, V neck, mini
length/size, maxi/long
length
Plain, lace, floral,
stripes
Black, navy, white, blue,
pink, red, neutral
15 Sleeveless, long sleeve,
round neck, V neck
Plain, stripes, floral,
lace
Black, neutral, white,
pink, navy, grey, blue
16 No result* Plain, floral, stripes Black, white, orange,
blue, navy, pink
17 No result* Plain, floral, stripes Black, navy, grey, blue,
brown, green, white
18 Long sleeve Plain, checks, stripes,
animal
Black, blue, green,
brown, navy, grey
19 V neck, mini length/size,
long sleeve
Plain, lace, floral,
stripes
Black, navy, red, blue,
maroon, white
30
20 Long sleeve, round neck, V
neck
Plain, stripes, checks,
floral, lace
Black, neutral, grey,
brown, pink, white,
green
Data source: Coded based on trend forecasts by EDITED (2019). The coding scheme is
detailed in Table 3.3
Note*: EDITED was unable to pinpoint a clear design detail to be popular in the market.
Table 4.2 summarizes the coded EDITED forecasts for womenswear in the U.S.
market during the S/S 18 season. Compared with the forecast results produced by
WGSN, trend forecasts generated by EDITED, as a big data tool, appeared to be much
more stable and coherent, with most trends spanning the entirety of the S/S 18 season.
This result, however, is far from surprising. Notably, EDITED-based trend forecasts used
popular selling items in the market as inputs, which include not only trendy apparel but
also basic fashion (EDITED, 2019). For instance, basic wardrobe essentials, such as a
pair of work slacks or a black dress, often generate stable sales revenue for retailers,
whereas few retailers carry trendy items only to avoid high business risks (Israeli &
Avery, 2018). Additionally, the results of the EDITED forecasts are based on items that
had been replenished in the stock frequently enough, evidence that the item has been
consistently popular in the market and thus is more likely to look ‘similar’ in fashion
trends throughout the season (Burns, Mullet, & Bryant, 2016).
Specifically, for the S/S 18 season womenswear fashion trends in the U.S. retail
market generated by EDITED:
First, the forecast result for the color is the most stable among all the three
dimensions examined. It is interesting to note that almost all forecasts, regardless of the
31
clothing type and sub-season, contain basic colors—‘black’ and ‘white’ in their trend
predictions (Min, 2015). Beyond basic colors, EDITED also identifies ‘blue’ and ‘pink’
to be popular throughout the S/S 18 season. The lack of color diversity in EDITED’s
forecast results echoes the concerns that generating inspirational, creative color palettes
could be a weakness of big data-based fashion forecasts (Tehrani & Ahrens, 2016).
Second, the pattern forecasts by EDITED overall displayed more diversity than
the results by WGSN. For example, Table 4.1 and Table 4.2 show that for a particular
category of womenswear in the U.S. market during a particular sub-season in S/S 18,
EDITED typically forecasts 3-4 patterns to be popular compared with only 2-3 generated
by WGSN. A possible explanation for the greater pattern diversity of the EDITED
forecasts is that big data tools can more easily monitor the popularity of all patterns
available in the market simultaneously, whereas humans can only validate pattern trend
predictions in a much narrower scope once at a time (Rousso, 2012).
Additionally, the trend forecasts by EDITED have fewer variations among
product categories than the results of WGSN. As shown in Table 4.2, except for the fad-
oriented swimwear, the EDITED-generated trend forecasts for other product categories
were fairly consistent and interconnected across different sub-seasons in S/S 18.
4.2 Statistical Analysis
Table 4.3 WGSN and EDITED Trend Forecasts: Results Comparison
Similarity/Content Design Details Patterns Colors
Very Similar 30% (N=6) 85% (N=17) 55% (N=11)
Similar 40% (N=8) 15% (N=3) 45% (N=9)
32
Different 30% (N=6) 0% (N=0) 0% (N=0)
As summarized in Table 4.3, the similarity in the results of trend forecasts
generated by WGSN and EDITED varied across the three dimensions examined.
Specifically, the forecasts for the design details by the two approaches turned out to be
the least similar with 30% (N=6) cases being totally ‘different’ (i.e., EDITED’s forecast
matched none of WGSN’s forecast). In comparison, the forecasts for the color and pattern
shared more common outcomes, with no cases being totally ‘different.’ Meanwhile,
WGSN and EDITED’s predictions for the pattern had the highest degree of similarity
overall, with 85% (N=17) cases being classified as ‘very similar’ (i.e., EDITED’s
forecast matched more than 50% of WGSN’s forecast). Likewise, the WGSN and
EDITED forecasts for the color matched well too, with over half of the cases (55%,
N=11) being classified as ‘very similar.’
Table 4.4 Results of Independent Samples T-test
Variables/Hypothesis Mean=0 Mean=1 Mean=1.5 Mean=2
Design Details 5.627**
(0.00)
0.000
(1.00)
-2.814**
(0.01)
-5.627**
(0.00)
Pattern 22.584**
(0.00)
10.376**
(0.00)
4.273**
(0.00)
-1.831
(0.83)
Color 13.581** 4.819** 0.4381 -3.943**
33
(0.00) (0.00) (0.666) (0.01)
**: p<.01 at the 99% confidence level; *: p<.05 at the 95% confidence level
To further statistically evaluate the degree of similarity between the forecast
results generated by WGSN and EDITED, we conducted the independent sample T-test
on the 20 paired trend forecasts (i.e., a total of 40 trend forecasts listed in Table 4.1 and
Table 4.2). The result of the test is shown in Table 4.4. Specifically:
First, according to the results, we rejected the null hypothesis that the mean value
for “design details” can be 0 (p=0.00), 1.5 (p=0.01) and 2 (p=0.00) at the 99% confidence
level. However, we could not reject the null hypothesis that the mean value for design
details is 1 (p=1.00>0.05) at the 99% confidence level, meaning WGSN and EDITED’s
forecasts for design details are statistically ‘similar’. Second, we rejected the null
hypothesis that the mean value for pattern can be 0 (p=0.00), 1 (p=0.00), and 1.5 (p=0.00)
at the 99% confidence level. However, we could not reject the null hypothesis that the
mean value for the pattern is 2 (p=0.83>0.01), at the 99% confidence level; thus, WGSN
and EDITED’s forecasts for the pattern statistically are ‘very similar’ as defined by the
study. Third, while we rejected the null hypothesis that the mean value for the color can
be 0 (p=0.00), 1 (p=0.00), or 2 (p=0.01) at the 99% confidence level, we could not reject
the null hypothesis that the mean value for the color is 1.5 (p=0.666>.05), at the 99%
confidence level, indicating that WGSN and EDITED’s color forecasts are between
‘similar’ and ‘very similar.’
Overall, the statistical analysis confirms the promise and potential of using big
data to forecast fashion trends since all forecasts were at least ‘similar’ (i.e., EDITED’s
34
forecast matched some but less than 50% of WGSN’s forecast). The results also suggest
that big data seems to be more effective in generating forecasts for the pattern and color,
whereas its capability of predicting the design details raises more questions. This result is
not entirely surprising. As suggested by previous studies, the pattern and color are more
straightforward and predictable, whereas the multiplicity of the design details makes it
more difficult for big data to forecast (Jackson, 2007).
35
IMPLICATIONS AND FUTURE RESEARCH AGENDAS
5.1 Findings
This study empirically evaluated the similarities and differences of the results of
traditional human-based fashion trend forecasts (WGSN) with the ones generated by big
data tool (EDITED). Based on the comparison of 20 paired fashion trend forecasts for
S/S 2018 womenswear in the U.S. retail market generated by WGSN and EDITED and
by using the independent sample t-test, the study finds that:
First, at the 99% confidence level, WGSN and EDITED’s forecasts statistically are
suggested as ‘similar’ (i.e., EDITED’s forecast matched some but less than 50% of
WGSN’s forecast) for the design details, such as fabric and shape of the clothing.
Second, at the 99% confidence level, WGSN and EDITED’s forecasts statistically
are suggested as ‘very similar’ (i.e., EDITED’s forecast matched more than 50% of
WGSN’s forecast for the pattern of the clothing.
Third, at the 99% confidence level, WGSN and EDITED’s forecasts for the color
statistically are suggested between ‘similar’ and ‘very similar.’
5.2 Implications
The findings of the study fulfill a critical research gap regarding the feasibility of using
big data for fashion companies’ creative activity and significantly enhance our
36
understanding of both the advantages and limitations of using big data for fashion trend
forecasting (Choi & Hui, 2011; Ren et al., 2017). The findings of the study also have
several important implications:
First, the findings of this study suggest the great potential of using big data tools
to aid fashion companies’ forecasts and the creation of new products. Notably, the results
of the study reveal that many of the paired WGSN and EDITED forecasts showed a high
degree of similarity, particularly regarding patterns and colors. As these two aspects are
critical components of fashion trend forecasts, fashion companies could consider
adopting big data tools to help improve the accuracy of pattern and color forecasts.
Several fashion companies, such as Gap, Madewell, and Puma, have already started to
incorporate big data tools to improve their fashion trend forecasting methods (Israeli &
Avery, 2018).
There is also a growing number of big data tools newly launched to the market,
such as Trendalytics, aiming to help fashion brands and retailers improve their business
operations through big data analysis (Trendalytics, 2019). The findings of this study
suggest that adding and strengthening the functions of fashion trend forecasts could be a
promising area for these big data analytics tools to expand their service to fashion brands
and retailers.
Further, it is interesting to note that even traditional players in the fashion trend
forecast business, such as WGSN, have begun to explore the role of big data in trend
forecasting. For example, WGSN recently launched several tools, including a consumer
37
insight survey and WGSN Instock, to help fashion companies identify emerging
consumer preferences for fashion products supported by data analysis (WGSN, 2019).
Second, the findings of the study also illustrate the limits of using big data tools in
fashion trend forecasts. Notably, WGSN and EDITED produced different predictions for
design details. This result, however, echoes the findings of previous studies, which shows
that consumers’ desire for novel, innovative, and exciting design details are shaped by
many unpredictable and complex factors, making it difficult for big data to predict based
on the inputs from historical data alone (Barnes & Lea-Greenwood, 2010; Ming Law et
al., 2004). a
It shall also be noted that some design details suggested by WGSN, such as ‘patch
pocket’, ‘wide leg’, or ‘dropped waist’, were beyond the coverage by EDITED (EDITED,
2019; WGSN, 2019). This result suggests that the feature and capacity of the big data
tool is another factor that may limit the flexibility and outcome of the forecasts.
5.3 Future Research Agendas
Despite the interesting results, this study also has several limitations that future
research might overcome.
First, while this study looked at womenswear only, future studies may yield trend
results by looking at other product categories, such as menswear and children’s wear. It
will be interesting to compare the capability of big data tools in forecasting the fashion
trends for different product categories. The results will help us understand the strengths
38
and weaknesses of using big data tools for creative activities in the fashion industry
further.
Second, the method of this study also can be applied to investigate fashion trends
in other seasons, such as Fall/Winter and Resort. Specifically, forecast results may
change due to seasonal variations in product categories associated with special events and
weather. For instance, most Fall/Winter fashion lines often include a ‘holiday’ collection,
featuring party dresses and eveningwear, to accommodate the frequent celebrations that
occur in November and December. Holiday collections may be easier for big data to
forecast since specific colors, such as red and green, as well as the design details, like
sequins, are consistently popular year after year—enabling big data tools to forecast
based on historical data easily. However, Resort wear collections may present a great
challenge for big data tools. These collections are commonly released in January—a
cold-weather month in the Northern Hemisphere—but provide apparel, such as
swimwear, for vacationing in warm areas.
Moreover, future research may explore alternative big data methods to fashion
forecast, such as considering inputs from social media or predictive analytics. Research
that integrates traditional and big data-based fashion trend forecasting techniques can
reveal valuable insight on how to optimize the advantages and disadvantages of both
methods. For example, traditional fashion trend forecasting can provide inspiration that
satisfies consumers’ desire for originality while big data-based methods reduce business
risk by providing commercial inputs (Israeli & Avery, 2018; Ming Law et al., 2003).
Finally, as the fashion industry becomes increasingly data-driven, understanding how to
39
incorporate data analytics into fashion education will be essential to promote the
preparedness and success of fashion students. Fashion educators are already beginning to
incorporate data analytics into fashion trend forecasting and merchandising courses by
teaching students how to use EDITED. Undoubtedly, big data’s creative applications in
fashion trend forecasting hold immense potential to foster fashion companies’ success
and largely benefit the fashion industry.
40
REFERENCES
Ambrose, G., & Harris, P. (2007). The visual dictionary of fashion design. Ava
Publishing.
Atre, S. (2017). Big data analytics: Theory & practice. Retrieved
from
http://mis.ucd.ie/events/big-data-analytics-theory-practice
Blaszczyk, R. L., & Wubs, B. (Eds.). (2018). The Fashion Forecasters: A Hidden History
of Color and Trend Prediction. Bloomsbury Publishing.
Ballmer, A., & Tobias, J. (2017). Trend forecasting: Collecting the history of the
future. Art Libraries Journal, 42(1), 19-25.
Barnes, L., & Lea-Greenwood, G. (2010). Fast fashion in the retail store
environment. International Journal of Retail & Distribution Management, 38(10),
760-772.
Beheshti-Kashi, S., Karimi, H. R., Thoben, K. D., Lütjen, M., & Teucke, M. (2015). A
survey on retail sales forecasting and prediction in fashion markets. Systems
Science & Control Engineering, 3(1), 154-161.
Boone, T., Ganeshan, R., Jain, A., & Sanders, N. R. (2019). Forecasting sales in the
supply chain: Consumer analytics in the big data era. International Journal of
Forecasting, 35(1), 170-180.
Brannon, E. L. (2010). Fashion forecasting. New York, NY: Fairchild Books.
http://mis.ucd.ie/events/big-data-analytics-theory-practice
41
Burns, L. D., Mullet, K. K., & Bryant, N. O. (2016). The business of fashion: Designing,
manufacturing, and marketing. Bloomsbury Publishing USA.
Chaudhuri, S. (2018). H&M pivots to big data to spot next big fast-fashion trends. Wall
Street Journal. Retrieved from https://www.wsj.com/articles/h-m-pivots-to-big-
data-to-spot-next-big-fast-fashion-trends-1525694400
Choi, T. & Hui, C-L. (2011). An intelligent fast sales forecasting model for fashion
products. Expert Systems with Applications, 38(6), 7373-7379.
Choi, T., Hui, C., Liu, N., Ng, S., & Yu, Y. (2014). Fast fashion sales forecasting with
limited data and time. Decision Support Systems, 59, 84. Retrieved from
https://search.proquest.com/docview/1507600748?accountid=10457
EDITED (2019). Market analytics. Retrieved from https://accounts.edited.com/
Fox, R., Grail, M., & Peng, A. (2018) Geek meets chic: Four actions to jump-start
advanced analytics in apparel. San Francisco: McKinsey & Company.
Furukawa, T., Miura, C., Mori, K., Uchida, S., & Hasegawa, M. (2019). Visualisation for
analysing evolutionary dynamics of fashion trends. International Journal of
Fashion Design, Technology and Education, 12(2), 247-259.
Gaimster, J. (2012). The changing landscape of fashion forecasting. International Journal
of Fashion Design, Technology, and Education, 5(3), 169-178.
Grammenos, C. (2015). The evolution of “big data” in fashion forecasting. Retrieved
from
https://www.hbs.edu/openforum/openforum.hbs.org/goto/challenge/understand-
https://www.wsj.com/articles/h-m-pivots-to-big-data-to-spot-next-big-fast-fashion-trends-1525694400
https://www.wsj.com/articles/h-m-pivots-to-big-data-to-spot-next-big-fast-fashion-trends-1525694400
https://search.proquest.com/docview/1507600748?accountid=10457
https://accounts.edited.com/
https://www.hbs.edu/openforum/openforum.hbs.org/goto/challenge/understand-digital-transformation-of-business/the-evolution-of-big-data-in-fashion-forecasting/comments-section.html
42
digital-transformation-of-business/the-evolution-of-big-data-in-fashion-
forecasting/comments-section.html
Gu, W., & Liu, X. (2010). Computer-assisted color database for trend forecasting.
In 2010 International Conference on Computational Intelligence and Software
Engineering (pp. 1-4). IEEE.
Heching, A., Gallego, G., & van Ryzin, G. (2002). Mark-down pricing: An empirical
analysis of policies and revenue potential at one apparel retailer. Journal of
revenue and pricing management, 1(2), 139-160.
Israeli, A. & Avery, J. (2018). Predicting consumer tastes with big data at Gap. Harvard
Business School (pp. 1-27). Retrieved from
https://files.transtutors.com/cdn/uploadassignments/2661391_2_gap
Jackson, T. (2007). The process of trend development leading to a fashion season. In
Fashion
Marketing (pp. 192-211). Routledge.
Joshi, N. (2018). Big data is stepping into the fashion world. Retrieved from
https://www.bbntimes.com/en/technology/big-data-is-stepping-into-the-fashion-
world
Jung, J., & Shen, D. (2011). Brand equity of luxury fashion brands among Chinese and
US young female consumers. Journal of East-West Business, 17(1), 48-69.
King, J. A. (2012). Colour trends and selection in fashion design. In Colour Design (pp.
535-550). Woodhead Publishing.
Kim, E., Fiore, A. M., & Kim, H. (2013). Fashion trends: analysis and forecasting. Berg.
https://www.hbs.edu/openforum/openforum.hbs.org/goto/challenge/understand-digital-transformation-of-business/the-evolution-of-big-data-in-fashion-forecasting/comments-section.html
https://www.hbs.edu/openforum/openforum.hbs.org/goto/challenge/understand-digital-transformation-of-business/the-evolution-of-big-data-in-fashion-forecasting/comments-section.html
https://files.transtutors.com/cdn/uploadassignments/2661391_2_gap
https://www.bbntimes.com/en/technology/big-data-is-stepping-into-the-fashion-world
https://www.bbntimes.com/en/technology/big-data-is-stepping-into-the-fashion-world
43
McDowell, M. (2019). Analytics are reshaping fashion’s old-school instincts. Retrieved
from https://www.voguebusiness.com/technology/data-trend-forecasting-google-
tracking-tools
Merriam-Webster. (2019). Dictionary. Retrieved from https://www.merriam-
webster.com/
Min, S. (2015). Korean fashion designers’ use of cultural expression and its influence on
their design. Fashion Practice, 7(2), 219-239.
Ming Law, K., Zhang, Z. M., & Leung, C. S. (2004). Fashion change and fashion
consumption: the chaotic perspective. Journal of Fashion Marketing and
Management: An International Journal, 8(4), 362-374.
North, E. J., De Vos, R. B., & Kotze, T. (2003). The importance of apparel product
attributes for female buyers. Journal of Consumer Sciences, 31(1), 41-51.
Pantano, E., Giglio, S., & Dennis, C. (2018). Making sense of consumers’ tweets:
Sentiment outcomes for fast fashion retailers through Big Data
analytics. International Journal of Retail & Distribution Management, 47 (9),
915-927.
Park, E. J., Kim, E. Y., Funches, V. M., & Foxx, W. (2012). Apparel product attributes,
web browsing, and e-impulse buying on shopping websites. Journal of Business
Research, 65(11), 1583-1589.
Rayment, S. (2015). How to wear: Longline Pieces. Retrieved from
https://www.voguebusiness.com/technology/data-trend-forecasting-google-tracking-tools
https://www.voguebusiness.com/technology/data-trend-forecasting-google-tracking-tools
https://www.merriam-webster.com/
https://www.merriam-webster.com/
44
Ren, S., Chan, H., & Ram, P. (2017). A comparative study on fashion demand
forecasting models with multiple sources of uncertainty. Annals of Operations
Research, 257(1-2), 335-355.
Resnick, M. L., & Montania, R. (2003). Perceptions of customer service, information
privacy, and product quality from semiotic design features in an online web
store. International journal of human-computer interaction, 16(2), 211-234.
Rousso, C. (2012). Fashion forward: A guide to fashion forecasting. New York: Fairchild
Books.
Salkind, N. J. (2017). Statistics for people who (think they) hate statistics. Sage
publications.
Silva, E. S., Hassani, H., & Madsen, D. Ø. (2019). Big Data in fashion: transforming the
retail sector. Journal of Business Strategy. Ahead of print.
https://doi.org/10.1108/JBS-04-2019-0062
Sterlacci, F., & Arbuckle, J. (2009). The A to Z of the Fashion Industry (No. 96).
Rowman & Littlefield.
Sun, L., & Zhao, L. (2018). Technology disruptions: exploring the changing roles of
designers, makers, and users in the fashion industry. International Journal of
Fashion Design, Technology and Education, 11(3), 362-374.
Tehrani, A. F., & Ahrens, D. (2016). Improved forecasting and purchasing of fashion
products based on the use of big data techniques. In Supply Management
Research (pp. 293-312). Springer Gabler, Wiesbaden.
https://doi.org/10.1108/JBS-04-2019-0062
45
Thomassey, S., & Zeng, X. (Eds.). (2018). Artificial Intelligence for Fashion Industry in
the Big Data Era. Springer.
The Fashion Folks. (2016). Tropical pattern. Retrieved from
http://www.thefashionfolks.com/blog/tropical-pattern/
Trendalytics (2019). Powering the future of commerce. Retrieved from
Visual Next. (2018). How big data is impacting the fashion industry. Retrieved from
https://www.visualnext.com/software/big-data-business-intelligence-fashion/
World Global Style Network (2019). Future trends—forecast. Retrieved from
https://www-wgsn-com.udel.idm.oclc.org/content/reports/#/Future+Trends/w
Wong, W. K., & Guo, Z. X. (2010). A hybrid intelligent model for medium-term sales
forecasting in fashion retail supply chains using extreme learning machine and
harmony search algorithm. International Journal of Production
Economics, 128(2), 614-624.
Yau, C. (2018). Your roundup of 2018 top trends. Retrieved from
https://edited.com/blog/2018/12/2018-top-trends/
http://www.thefashionfolks.com/blog/tropical-pattern/
https://www.visualnext.com/software/big-data-business-intelligence-fashion/
https://www-wgsn-com.udel.idm.oclc.org/content/reports/#/Future+Trends/w
- Chapter 1
INTRODUCTION
1.1 Introduction
1.2 Research Question
1.3 Key Definitions
Chapter 2
LITERATURE REVIEW
2.1 Fashion Trend Forecasting and Related Theories.
2.2 Using Big Data in the Fashion Industry
2.3 Debate on the Application of Big Data for Fashion Trend Forecasting
2.4 Summary
Chapter 3
METHODS AND DATA
3.1 Data collection
3.2 Data analysis
4.1 Descriptive Analysis
4.2 Statistical Analysis
Chapter 5
5.1 Findings
5.3 Future Research Agendas