The Changing Digital Dynamics of Multichannel Marketing: The Feasibility
of the Weblog - Text Mining Approach
for Fast Fashion Trending
Tracy Anna Rickman,
Robert M. Cosenza,
Extended Abstract
Copyright 2006-2008 by the authors and the DMA. This article cannot be reproduced
or disseminated in any form without the express permission of the authors.
The authors provide an argument for the use of weblog textmining as a tool for multichannel decision making in the fashion industry. Since getting the goods to the market segments to defend and/or increase market share requires quick understanding and response within and between all constituents in the fashion supply chain, the authors focus on forecasting. They
argue that the fashion/style company that can tap the continual flow of information in the present, contrast it with a stored set of information from the past, and adjust based on repeated cycles, will have the best insight into the lingering trend, changing trend, or dynamic trend. Thus they will make the best competitive decisions. This is apparent both for manufacturers and later for resellers in areas like online catalogs, direct mail, etc.
Solomon (2006) defines fashion as the process of social diffusion by which a new style is interpreted as a context dependent code, and then adopted by a group of consumers. Fashion contains all types of cultural phenomena including adornment, clothing, art, music, and architecture. He further defines “a” fashion as synonymous with style. As such, a fashion refers to a particular combination of attributes. To be “in style” means that some reference group positively evaluates the combination. Style is then likened to a bond, a communicative code that represents the DNA of the genre. This communication/connectiveness forms the basis of most post-modern style groups. The discovery and understanding of the changing dynamics of these groups form the basis of the multichannel decisions that have to be made in this style industry.
Although style has
morphological variability, it can be consistently defined as a manner of expression
of a particular individual or group. The
definition also has subsumed elements or patterns unique to the person that
individuate him or her. To know a style
group, one must follow them into a common space – to find their commonalities.
Commonalities become the expressive elements
of the segment. Following these elements
and observing the style can give a good indication of what style is or is
not. The quickness of discovering “in style” elements
of a segment can give multichannel fashion distribution systems incredible
competitive advantage – from handshakes to channels to customers in just a
few days or weeks. In a world of possibility,
laggards end up losers. Shortening
forecasting cycles, spotting shifts in demand, and fine tuning your company
to deliver to the market in weeks, not months is the key survival in the dynamic
multichannel world.
Trend monitoring
from the street usually begins with but is not limited to (other methods have
been used, focus groups, etc.) the collection of a time oriented set of visuals
from various global markets around the world.
Minimally, data like this, to be critical to the forecaster, must have
commonality, similarity, difference, and clusters.
Here lies the science and art of forecasting – the model that incorporates
the data in a reality of the future, and the decision maker that sets the
rules for discovery – developing the way the model works (finding commonality,
similarity, difference, clusters) and how the predictive outcome is used. Thus, the current state of the street forecast
will benefit those in the industry with evolved supply/value chains who can
apply fast fashion techniques, managing their growth trends. But, it will do nothing to find a decay in a
trend or the next real pull trend from the customer. This must occur in a longitudinal fashion and
incorporate street data into large databases and database management. Due to the proprietary nature of the data in
the industry, this is unlikely to happen.
Because consumers under the age of 25 have spending power and a greater interest in fashion, they are influential in setting trends (Brannon, 2005). Knowing what these consumers want could be the basis of a new street trend forecasting revolution. Feitelberg (2001) has described the worth of individual data in all forms, especially fashion by eloquently stating “When you collect (data) from the same individuals every month, you get a good idea about what’s in their closets.” Unique personalization is used in the context of chronicling, storing and/or receiving information that is transformed by an individual into a personalized form and delivered to a larger body of the population that has an interest in the information, person, or the group. It forms the basis of the development of very unique and personal information portals, databases. This is called the popularized Weblog or simply the blog (Baker and Green, 2005). Thus, the fashion blog would contain thoughts, opinions, experiences, and in most instances visual content. A consequence of the pervasive use of computers is that most weblogs originate in digital form. Although not all of them are fashion blogs, Technorati, a blog search engine, is now tracking 19.6 Million weblogs, and the total number of weblogs tracked continues to double about every 5 months. What this means to the world of fashion forecasting is a repository of untapped fashion data that is individual, to what they are and what they want to be and what they don’t want to be, and to make it even more exciting, personalized for the most part with cultural, demographic, shopping data and more – all just waiting to be extracted to find the next real “trend”. Is this the new street, the digital repository of the internet? Only if we can ferret out the fashion information from the textual and visual personalized fashion data. Although data mining has been used for years for prediction, new algorithms, rules, models, and methods have recently been developed and tweaked to make sense out of the text (pictures are another revolution called object mining) using semantic rules for creating similarity and commonality, clusters, and differences in clusters. In other words, tuning in to the next real trend or tipping point.
Text mining aims
to automatically determine various attributes of a free-form text (e.g., a
news article, office memo, technical report, etc.) including key features,
frequently occurring words, summary, category, etc (Weiss, et al, 2004; Sullivan,
2000). Unstructured information makes up about 80-90%
of all information available for decision making. Although numbers are preferred by data analysts,
context-sensitive text mining is a great tool for structuring this content
for decision making. Therefore, when
structured, this text information becomes an asset for any corporate decision
making, value delivery, and especially forecasting (Anthes, 2004; Fickenscher,
2005). Although text mining has been
around for many years, it is not a panacea for finding worth in text data.
However, the fashion industry seems to be in a quagmire when it comes
to forecasting. As with all industry, they are searching out
the latest and greatest technology tools that can give them the boost to determine
dominant fads and real trends in a more reliable and accurate manner.
The text mining technology, coupled with the historical and current
repository of unstructured information on the Internet (Weblogs) seem to promise
at the least an alternative and/or concomitant technology to deal with shorter
lifecycles, competitive selling environments, and more comprehensive style
and color offerings (Baker, 2004).
Alexander Halavais, a researcher at SUNY Buffalo, believes that analyzing weblogs could provide a vivid picture of the future social landscape (Futurist, 2005). Future convergence, using the cell-phone, will also impact the diffusion/proliferation of weblog information, especially in the less than 25 age group (New Media Age, 2005). Striving to keep an eye on the changing dynamics of their consumers and to operate in an efficient supply chain/value chain, furniture retailers are using the Weblog-Text-Trending. What exactly is the buzz?
Weblog mining is a special case of data mining (Mena, 1998). The objective is to determine a variety of structural patterns to text data contained within Weblogs. The models marry text mining algorithmic attributes and operations with unstructured content from the Internet Blogspehere to determine present and future patterns. Okumura (2005) presents a weblog mining methodology that captures trends on Japanese blogs. Okumura’s weblog mining system automatically extracts and mines “burstiness” (trend, frequency, time-span), “hot words”, and favorable and unfavorable opinions toward objects (i.e. fashion objects) from a collection of specified weblog pages. Although it is in rough form, Fukuhara (2005) developed algorithms and methods that evaluate Chinese Weblogs and real time social occurrences to find matches, lag, and leading indicators of social concerns. Building on the work of Kumar et al (2003) and Gruhl et al (2004), Nakajima (2005) proposed and tested a methodology to search weblogs to find important bloggers. They found two groups of important bloggers, the agitators, and the summarizers. The agitators are able to generate the buzz – analogous to the trend setter. Although in its infancy and seminal in nature, there is some evidence to indicate that textmining Weblogs in general and fashion weblogs in particular could generate musing that could identify the next real fashion innovation/trend.
Obviously, the fashion
world is driven by its own fads. Cool
forecasting, a fad of the last few years, has given way to Trend forecasting.
Gina Piccalo (2005) a writer for the LA Times sums it all up beautifully. “Trends are hot – cool isn’t. As culture morphs at internet speed, forecasters
fight to stay ahead of it all”. And
the traditional trend watchers will not provide the needed speed to predict
short run color, fabric needs, and hot fashion and definitely not real long
run innovation/trend.
Arguably, the less than 25 year old group, especially the male component,
currently rules the web. And if technology
morphs like it has since the first IBM computer accessed the Internet, this
less than 25 year old cohort will continue to rule the digital roost.
They will be the group that will provide all of the data necessary
to create a new forecasting dynamic from the marriage of the weblog and textmining.
The weblog buzz is alive and growing.
The tippers are out in blogspace. They
need to be discovered in the digital world. Then as the weblog diffuses to the later adopters,
which for all practical purposes has started (look at the buzz created on
the web to sort out the new Medicare pharmaceutical policies), forecasters
will be able to take the technology to different fashion cohorts.
Future fact or fiction, textmining the weblog is not going away – especially
in a short cycle, hard to forecast, dynamic business/fashion global environment. Whether we like it or not, the traditional methods
have to morph. Qualitative research
will need a whole new generation of human muses that go beyond the traditional
methods and sift through web based digital data/information to find trend.
There are simple tools available to
get a sense of weblog trend textmining. One such tool is available from Blogpulse trend
tools. It is possible to compare search
terms/links in isolation, or use all three fields to compare search terms/links
against others. As fashion forecasters
get more sophisticated with the models and procedures, forecasts will get
very interesting.
Most in the field
of direct marketing would have to agree that the field has one commonality
that drives its existence. Direct marketing
is all about capturing the attention of a consumer long enough to make an
impression and a sale. It is about
connecting and communicating benefits. (Cymfony Blog, 2006)
In this scenario, blogs are the storehouse of the content and data
necessary for future connection and communicating.
Semantic and image mining of the web will be the next frontiers of data mining
and trend spotting. Initially from
the street level, the methods will morph to other cohorts as they are verified.
These emerging technologies will enable
researchers to find semantic meaning hidden in data and documents, share and
integrate information with supply chain/value chain members, and find more
valuable insights. Some future thoughts: