This document is a presentation about using text analytics to accelerate insight generation. It discusses how text analytics can extract usable knowledge from unstructured text data to support decision making. Text analytics allows organizations to explore topics being discussed, sentiment, and provide a single voice of the customer experience. Two case studies are presented, one involving merging feedback from multiple sources for an airline, and another analyzing online discussions about diabetes to understand perceptions of a new drug class. Challenges of text analytics like multiple languages and garbage in/garbage out are also addressed.
2. 2Accelerating Insight Generation | November 2016 | Presentation |
In a nutshell…
Text analytics is a method for
extracting usable knowledge from
unstructured text data through
identification of core concepts,
sentiments, and trends, and then
using this knowledge to support
decision making
3. 3Accelerating Insight Generation | November 2016 | Presentation |
More and more unstructured data
Increasing number of channels
of interaction
Shorter MR questionnaires
Faster decision making process
Big data or
big headache?
The new normal
4. 4Accelerating Insight Generation | November 2016 | Presentation |
A well-balanced research ecosystem
that facilitates data integration
Analytical capabilities to make
sense of that data and turn it into
actionable insights in timely fashion.
Organisations need:
5. 5Accelerating Insight Generation | November 2016 | Presentation |
Accelerating the insight generation process across sources
Text analytics
Text
analytics
Complaints
Survey/
EFM verbs
Emails
Qualitative
data
Web data
Image/
Voice
6. 6Accelerating Insight Generation | November 2016 | Presentation |
So, what’s the big deal?
Cost
Better
insightsSpeed
Consistency
Scalability
Data
integration
Response
rates
Engagement
7. 7Accelerating Insight Generation | November 2016 | Presentation |
Detecting new issues
Concept and mind clouds
Robust quantification of content
into topics and sentiment
Text based driver analysis
Explore and quantify
Pattern detection Tracking and action planning1. 2.
9. Accelerating Insight Generation | November 2016 | Presentation | 9
Social listening
Customer comment
Staff comment
Three data
sources + One flexible text
analytics model = One holistic
feedback voice
Opportunity:
Global airline receives a lot of text feedback about the customer
experience from three different sources.
The background
Solution:
Using text analytics, data sources were merged to create a single
holistic feedback and provide compelling evidence for change.
Approach:
10. 10Accelerating Insight Generation | November 2016 | Presentation |
WHAT people are saying
We can identify 32%
26%
24%
18%
8%
5%
5%
5%
4%
2%
2%
1%
1%
Flight experience
Staff
Flight punctuality
Food and drink
Non-flight experience
Baggage
Check-in
Landing and arrivals
Communication
General satisfaction
Product offering
Boarding
Entertainment
• Customer comment: 500K verbatim
• Staff comment: 10K verbatim
• Social comment: 500K verbatim
% MENTIONS BY TOPIC
(Averaged by voice)
11. 11Accelerating Insight Generation | November 2016 | Presentation |
And who is talking most about what
Total Customer
Social
listening
Staff
Flight experience 32%
Staff 26%
Flight punctuality 24%
Food and drink 18%
Non-flight experience 8%
Check-in 5%
Baggage 5%
Landing and arrivals 5%
Communication 4%
General satisfaction 2%
Product offering 2%
Boarding 1%
Entertainment 1%
• Customer comment: 500K verbatim
• Staff comment: 10K verbatim
• Social comment: 500K verbatim
VOLUME OF COMMENT
Dark colours indicate a higher
volume of comment, lighter
colours a lower volume
12. 12Accelerating Insight Generation | November 2016 | Presentation |
We can set sentiment against volume
StaffFlight punctuality
Flight experience
Food/drink
Non-flight experience
Check-in
Boarding
Baggage
Entertainment
Landing/
arrivals
Product
offering
Communication
General satisfaction
Volume
Priority
improvement
Priority
maintenance
- Sentiment +
13. 13Accelerating Insight Generation | November 2016 | Presentation |
And split sentiment across the voices
Flight experience Staff Flight punctuality Baggage Food and drink
Customer Social listening Staff
PositiveNegative
Strength of sentiment is shown here not volume
The bigger the bar upwards, the more positive it is; the bigger the bar
downwards, the more negative it is (only top five categories shown)
-
+
14. Accelerating Insight Generation | November 2016 | Presentation | 14
WE FOUND AREAS OF
Ipsos MORI – Your WSBL
Only 4% of customers said the food was bad
Only 1% online said the food was bad
BUT
22% of staff think the food should be improved
Base: 29,770 comments; 27,510 customer comments, 1,072 web comments, 1,188 staff comments
DISSONANCE …
… AND AREAS OF
RESONANCE …
26% of customers mentioned delays
21% report delays online
AND
17% of staff want to better manage disruption
15. 15Accelerating Insight Generation | November 2016 | Presentation |
Text analytics allow organisations to explore:
Who is talking
What is being said
The sentiment
Action planning
Resonance identification
Understanding trend drivers
+
+
+
+
+
+
And provide one voice
of the customer experience
=
17. 17Accelerating Insight Generation | November 2016 | Presentation |
17
Understanding
what individuals
are saying about
diabetes online
Focus on ‘new kid on
the block’ – SGLT2
In-depth analysis from 390,000 posts on 19,000
topics, sourced from 3 diabetes patient forum
websites in the UK and US.
Goal:
to derive insight from comments in order to:
1. Build a picture around SGLT2
2. Understand the impact of the disease
3. Identify unmet needs and life-style
implications
18. 18Accelerating Insight Generation | November 2016 | Presentation |
Analytical steps
Starting with raw text divided into posts
TRADITIONAL
ANALYSIS
(frequency/
crosstab tables/
multivariate),
DATA
VISUALIZATION
Data
Cleaning
Grammatical
Annotation
Concept
Extraction
Topic
Categorization
& sentiment
19. 19Accelerating Insight Generation | November 2016 | Presentation |
About the SGLT2 Class?
What is Good & Bad
Positives
Convenience
Weight loss
Effectiveness - patients
share success stories
Low risk of
hypoglycaemia
Positive blood pressure
lowering effect
Negatives
Side effects such as UTIs
& increased urinary
frequency
Limited long term safety
data
Patients are aware that
FDA warnings exist
around ketoacidosis and
CV risk
FDA required additional
studies for liver toxicity
21. 21Accelerating Insight Generation | November 2016 | Presentation |
Patient Journey
Developing a better understanding of impact of disease on lifestyle
DIAGNOSIS LIFESTYLE CHANGES
FIRST
SYMPTOMS
Nausea
/dizziness
Anxiety
Patients recommended to:
►Monitor bg levels after meals each day
►Balance treatment with exercise and low
carbohydrate diet
►Reduce weight, if necessary
Potenital
weight loss
Social & Emotional
support
EMERGENCY ROOM VISIT
Blood Glucose Test
indicates possible
diabetes diagnosis
Shock & surprise
of BG results
HCP VISIT
BG monitoring at
home starts
Blood Glucose/A1c
test confirms
diganosis
Treatment
Initiation
Disappointment
Understanding and
acceptance
DIAGNOSIS
4x daily BG
monitoring around
meal times
Regular
Exercise
Adjust Diet Plan
DIETICIAN
HCP VISIT
Possible
Treatment
Adjustments
Social support
DIAGNOSTIC AND MONITORING
SYMPTOMS
DIABETES MANAGEMENT
STATE OF MIND
23. 23Accelerating Insight Generation | November 2016 | Presentation |
What‘s not the big deal?
Managing multiple languages
Systematic errors
Beware of the hype
Rubbish in / Rubbish out
24. 24Accelerating Insight Generation | November 2016 | Presentation |
The future is here!
Voice
Video
Image
Interactive survey
branching
Chat bots