Using Unstructured Text Data to Stay Ahead of Market Trends and Quantify Customer Perception

Using Unstructured Text
Data to Stay Ahead of
Market Trends and
Quantify Customer
Perception
Presented by:
Swaroop Johnson, Consultant (Analytics)
July 13, 2016

of data generated is
unstructured in nature,
and growing exponentially
of business executives
complain that they have
too much unstructured
text data and are unable
to interpret them
There are incremental
insights to be generated
from the text data given
willingly by your
customers
Our text analytics services
cover all phases of the
product life cycle, and the
customer journey
Imagine being able to
predict future trends
before it actually happens
And then design a
product based on
customer feedback,
and product reviews
And track each customer
touchpoint to identify
customer queries and
evaluate your customer
experience program
To evaluate net
sentiment of your
customer base with
deeper analysis and
insights than ever
before
Discover a new side of
story telling by
discovering hidden
insights from unused
sources of data
80%
40%
© 2016 Blueocean Market Intelligence

Enabling incremental insight generation through
comprehensive coverage of data sources
New Product Development
• Key opinion leader blogs
• Feedback/comments
• Product reviews
Customer Interaction Analysis
• Call centre data
• Feedback/comments
• Call logs
Digital Research
• Technology blogs and forums
• Technical papers
• Journals and magazines
Customer Experience Management
• Microblogging sites/social media data
• Feedback/user comments
• Product reviews
Brand Monitoring
• Customer feedback
• Survey Data
• Social media/Twitter data
Fraud Detection
• Emails
• Historical claims documents
• Financial statements

Some Practical Applications of Text Analytics
© 2016 Blueocean Market Intelligence 4
Aspect Extraction Text Classification Sentiment Analysis
Summarization Article Extraction Topic Modelling
Our Solutions
T
Customer Experience Management
Time and cost for identification of customer and employee issues will be reduced
Brand Monitoring
Help companies to keep a tab of the health of their companies brand image by analyzing trends over a period of
time
Digital Research
Reduces time related to topics and document searches by grouping documents

• Extract interesting and non-trivial patterns or knowledge from unstructured text documents
• Identify different aspects /components of a single problem
• Overall sentiments for a specific aspect under the business scenario can be extracted
• Determine the underlying conditions that give rise to the reasons for the problem/phenomenon
“the restaurant is situated at an excellent location and the food is very delicious. there are fresh barbeques served over the table
as starters including vegetables mutton chicken fish and prawns. all too good to enjoy. the main course and desserts are
available over the buffet table and the food variety is quite a lot to choose from both between veg and non-veg. well maintained
and good seating arrangement ideal for business parties or with friends. the only drawback was that the seats are limited and
during rush hours the guests need to wait until they get their turn. however it was a very good eating experience along with
work mates.”
Information extraction for deeper analysis of the business problem
Net Sentiment
0.489 (indicates positive
sentiment)
Different aspects of the user review
Location Excellent
Food Delicious
Reservation Bad
Experience Good
Service Good
Entities
Variety Excellent
Non veg Good
Main course Bad
Barbeque Fresh*
Desserts Good

Topic Modelling to Understand Customer Perception
Based on the article that we identified we were able to extract insights relevant to the
marketing, and competition for XXX’s new YYY range of processors
Topic modelling and article extraction solutions were deployed to create a story that divided
the entire 384 comments into 20 topics
These insights can be used to understand what customers are speaking and how they perceive
Intel’s new range of processors, as well as evaluate the marketing and branding strategy
Customers have expressed that YYY
processors enable power savings
Marketing Insights
Customers feel YYY X86 provides high
performance with low power consumption
End users feel YYY processors are worth upgrading to
as it a good improvement over previous generations
Oracle’s AAA solves the heartbleed issue
End users consider AAA to be a bit
expensive
Customers have rated XXX’s X86 processors as the
most efficient out there beating AAA
Competitive Insights

Breaking down consumer reviews to aspects so as to understand purchase
drivers, and drivers of positive, and negative sentiments (1 of 3)
Client: Technology
Approach
Industry:
Technology/Product/Smart Watch
Business Challenge
 The client, one of the leading chipset manufacturers wanted to understand the fall in
demand for its in-house smart watch product by studying consumer reviews from two
most popular online retail channels amazon.com, and bestbuy.com
 The partner wanted to identify top purchase drivers, and identify the reasons behind
negative sentiments, and the top activities the fitness tracker was used for
Aspect Extraction
• Identifying the key purchase
drivers
• Understand drivers of positive,
and negative sentiments
• Top use cases of the product
were identified
• Time wise trends with drill
down abilities to identify
reasons for declining trend
Business Impact:
Lexicon Based POS Tagging
Tools used:
UTAP, R, Python, KWIQVIS, SQL
Server 2014,
14© 2016 Blueocean Market Intelligence

Results
 Identifying the key purchase
drivers
 Understand drivers of positive,
and negative sentiments
 Top use cases of the product
were identified
Business Impact:
Wrist Band
 Painful, Wears, Complaint, Flawed, Issues,
Irritation, Return, Burn
Sync
 Crashed, Restart, Fiasco, iPhone,
Unstable, Sadly
REM
 Validity, Quality, Incorrect, Poor
Bluetooth
 App, Phone, Unpair, Mediocre, Worst
Fitbit
 Versus, Comparison, Awesome, Biking
Firmware
 iPhone, Confusion, Pathetic, Miserable,
Fix, Crappy, Unresponsive
Top Aspects across Star Ratings And Their Associations
1 and 2 star rating 3 and 4 star ratings 5 star rating
Battery Life
 Phenomenal, Fantastic, Charge, Reliable,
Activities, Great
Screen
 Impressive, Suit, Pretty, Customize,
Perfect
App
 Phone, iPhone, Mapmyrun, Pair, Update,
Comfortable, Great, Personalized
Heart rate
 Accuracy, Appreciate, Run, Reliable,
Fantastic, Customizable, Notifications,
Swimming
Waterproof
 Shower, Heart rate, Notification,
Appreciate, Traveling, Reliable, Swimming
Notification
 Text, Voicemail, Smart, Calendar,
Reminders, Customizable
App
 Phone, Update, Watch, Interface, Sync,
Good, Annoy, Reconnect
Heart Rate
 Sleep, Aerobic, Dancing, Positives,
Runners, Comfort, Stylish, Accurate
4% 8% 19% 4% 26% 32% 7%
Highly Negative Negative Slightly Negative Neutral Slightly Positive Positive Highly Positive
Aspect Extraction

Output
 An interactive dashboard to enable understanding of customer perceptions, enabling the partner to identify drivers of
positive, and negative sentiments, based on time wise trends.
 Consumer reviews are color coded with star ratings distributions, with drill downs available, for each individual sentiment,
and particular aspect
Aspect Extraction

• User friendly, flexible user interface: One touch data pre-
processing (removal of junk/stop words/URLs, lemmatization
etc...)
• Automated text classification: Query classification for deeper
and better understanding of customer queries
• Makes text classification 60% faster than traditional methods
• Preview screens for data extraction, and data cleansing
• Sliders and input boxes for easy definition of parameters and
data split
• Pre-defined classification and sampling performed with the click
of a button
• Get Accuracy Score, Confusion Matrix Score and Classification
Report to be downloaded in .pdf/.doc format
Unstructured Text Analytics Platform (UTAP)
WINNER
Innovative Technology of the Year 2016
Big Data and Analytics Awards

Low cost per comment, making it economic for
large data volumes
Economic
Ability to handles millions of rows of textual data
per day
Scalability
Each step of the unstructured text classification
process is pre-built
Automated
Comprehensive coverage of data sources
(structured and unstructured)
Comprehensive
Value add through innovative platforms

Summary
80% of data generated today is unstructured in nature – CANNOT be ignored any
more1
2
3
5
4
Call center data, social media comments/posts, open ended survey data etc.
are some of the sources of unstructured data
Incremental insights related to customer perception/trends/sentiments
can be extracted by mining unstructured data
Determine the underlying conditions that give rise to the reasons for the
sentiment/perception/customer issue
Unstructured Text Analytics Platform (UTAP) make the text classification
exercise easier, faster, and efficient
Digital Research help companies stay ahead in identifying upcoming trends6

Thank You
For more information:
Swaroop Johnson
Consultant (Analytics)
Blueocean Market Intelligence
Email: swaroop.j@blueoceanmi.com
Website: www.blueoceanmi.com
United States| United Kingdom | Dubai | India | Singapore

Using Unstructured Text Data to Stay Ahead of Market Trends and Quantify Customer Perception

More Related Content

What's hot (18)

Viewers also liked (20)

Similar to Using Unstructured Text Data to Stay Ahead of Market Trends and Quantify Customer Perception (20)

More from Course5i (17)

Recently uploaded (20)

Using Unstructured Text Data to Stay Ahead of Market Trends and Quantify Customer Perception