Transcript Silicon India BI Summit Hyderabad – Aug-2011

12 Key Analytical Trends / Themes
Silicon India BI Summit
Hyderabad – Aug-2011
Derick Jose
Vice President-DAS ( MindTree )
Context
 12 Battle tested themes distilled from real life experiences
 Not based on Forrester or Gartner 
 Spans Emerging markets + Advanced markets
 Example + Specific call to action
Slide 2
Theme-1 : Information monetization
Slide 3 I
have terabytes of consumer data . How can I create a new revenue stream ?
Real life example : GPS info product catalogue
Travel behavior
Segmentation
Travel behavior
Media asset placement
and optimisation
Demographic data
GIS/Road data
Cut thru data
Village visitor data
Highway traveler
Travel
Behavior
Analysis
Congestion
Forecaster
Government road/town
planners
Misc data
Events data
Macro economic data
Understanding
drivers of traffic
Survey data
Census data
Retail / Mall planning
4 $ impacting actions from 3 info products
• Action-1 : Traffic Signal optimization
• Action-2 : Bill board setup
• Action-3 : Catchment interception and new outlet store creation
• Action-4 : Road creation/widening/ Policy changes to relieve congestion
buildup
SPECIFIC ACTION : Evolve SPECIALISTS to “curate” monetizable scenarios from data
Theme-2 : Analytics can solve problems customers never
thought existed
Are 6there patterns in
Slide
data which can signal an issue which are not apparent to my naked eyes ?
Real life example : Channel Cannibalization Detector
In emerging markets, did Modern trade cannibalize General trade
during summer season for beverage drinks
Analytical models can reveal if there is a systematic sales
cannibalization pattern
SPECIFIC ACTION : Don’t wait for customer problem statements. Prime latent demand
Theme-3 : Mining Digital Consumer Behavior
Which are the most engaged actions a Digital consumer takes on the platform ?
Slide 8
Real life example : Understanding WOM behavior
DAN
VINAY
FREE SAMPLE
LINK
NEW RAZOR LAUNCH
ARTICLE
MELANIE
SCOTT
HAIR CARE TIPS
ARTICLE
CONFIGURE UR OWN
PRODUCT LINK
LINK TO RESONATING
DISCUSSION
KRISHNAN
FREE PERFUME
SAMPLE LINK
BARGAIN
COUPONS
GOOD REVIEW
OF FACE WASH
NUGENT
SANJOY
RADHA
FREE SAMPLE
LINK
ACTION : Think Digital Platform as a Lab where understand drivers of consumer intent
Theme-4 : Unstructured Data Mining
Slide 10
What hidden patterns reside in unstructured data ?
Real life example : Sentiment analysis of opinions on www.yelp.com
Sentiment metric
Value
Posts by source
(For ABC Brand)
136
ABC Brand Sentiment index
2.70
ABC Brand Buzz velocity
(for 2008-2009)
8.03 posts/month
(+ve)
Competitive Buzz index
(with respect to Compet Brand)
0.56
More No. of posts entered (on yelp.com) for Compet Brand
than for ABC Brand
Competitive Sentiment index
(with respect to Compet Brand)
2.18
Ratio of ABC Brand sentiment index to McDonald sentiment
Index is greater than 1, this implies consumers have better
experience with ABC Brand as compared to Compet Brand
Competitive buzz velocity
(For Compet Brand, for 20082009)
12.64 posts/month
(+ve)
For this exercise MindTree sourced the postings from yelp.com
Eopinions and Twitter have more entries on ABC Brand sentiments and can be mined subsequently
No. of Positive sentiments is greater than No. of Negative
sentiments (Based on Yelp.com)
Indicates 8.03 posts are coming up per month at yelp.com.
Further slope of buzz curve is +ve.
Indicates 12.64 posts for Compet Brand are coming up per
month at yelp.com. Further slope of buzz curve is +ve.
ACTION
: Consider mining unstructured data for insights in addition to structured data
Slide 11
Theme-5 : Examining value at the intersect
What new questions can I ask at the intersection of online/offline behavior ?
Slide 12
Real Life Example : Search & Booking
SEARCH DATA - Logged & Flushed 
BOOKING DATA
ACTION : Dig deeper at the intersect !
Theme-6 : Ensemble models
How can we combine analytical models in innovative ways ?
Slide 14
Real life example : Segmentation + Text mining in Fleet ind
Av. Gallons
pcpm
Av. No of
Av. Ancillary
Transactions
Revenue
pcpm
Av. No. of
Retention
Calls
Av. Late Fees
Av. Activation
Rate
Av. MOM
Growth
Population
106.4
5.5
248.8
0.7
324.1
0.7
1.03
Stable Underdogs
62.5
3.6
83.4
0.3
87.7
0.5
1.05
Miniature Laggards
77.3
4.6
90.0
0.2
118.3
0.9
1.03
Cash Cows
179.9
8.0
1098.6
0.6
2965.3
0.7
0.96
Dark Horses
122.3
6.1
1196.2
0.4
534.6
0.7
0.93
Sulking Mediocres
101.5
5.2
63.2
4.6
279.2
0.6
1.07
Front-runners
276.4
12.4
163.8
0.7
371.0
0.7
1.04
Slide 15
Ensemble Model in Telecom Churn Prediction Models
Behavorial
Segmentation
CDR data
Billing data
Social
Network
analysis
Maven
list
Inbound calls
-Activations
-Inquiries
-Service calls
Outbound calls
-Collection calls
-Campaign calls
Cluster-2
High credit risk
Cluster-1
Low credit risk
Customer = “Scott Nugent”,
Customer = “V Komrala
Text
mining
keywords
Regression model
Customer = “Joe Henry”,
Customer = “A Sampath
Neural networks
Customer = “Scott +
Nugent”,
churn score+= Scoring
1
Customer+= Social
“Joe Henry”,
churn score
=0
ACTION : Combine Segmentation
text mining
Models
media
analysis
Customer = “V Komrala”, churn score =0
Customer = “A Sampath”, churn score =1
Theme-7 : Verticalized Data Model Frameworks
•
•
•
•
•
ARTS -Retail
ACCORD -Insurance
SCOR -Supply Chain
CDISC -Clinical Trials
CMAT – Customer Analytics
Which
Slide 17 prefabricated industry specific data models and analytical processes can I use to
jumpstart my solution ?
Real life example : ACCORD based Claims model for a
leading US based Insurance provider
ACTION : Which industries and sub processes do I need to build knowledge in Vertical specific
Slide 18
standards ?
Theme-8 : Analytics penetrating unconventional areas
Telemedicine + Student scoring models
Slide 19
Real life example : Location based intelligent alerts
Here when the district health care
officer clicks on location the most
statistically significant patterns
from t/chi / regression test will be
filtered and shown
ACTION : Which non traditional, data rich processes can we apply analytics in to unlock value?
Theme-9 : Need varied ways to triangulate predictors & hypothesis
What are the various means to unearth causal levers which influence a business outcome ?
Slide 21
4 important learning's in surfacing predictors and behavioral hypothesis
HUNCHES
+
STATISTICS
= FORCE MULTIPLIER
1. Informed Market “Intuition's + Analytics = Force multiplier
2. Intuitions are NOT bad ! Its not “either or” …”And” mindset
3. 6 ways to get “informed hunches” regarding what potentially caused a business
outcome
4. Triangulate !
Real Life Example : Triangulating Surfacing policy renewal predictors
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
Recency of a claims denial
Tenure of agent
Overall experience of agent ( total experience )
Automated deduction or cash/cheque based ( payment mode )
No of unanswered call center calls in last 8 weeks
Frequency of outbound triggers for renewal
Recency of phone bound renewal trigger
% Change in renewal commissions to the agent ( driven by policy )
3 month ratio of inbound calls to outbound calls
Range of channels for interaction – Agent / Internet / Mobile / call center
Outbound watch list : Frequency of occurrence of specific keywords in outbound call interaction
Inbound watch list : Frequency of occurrence of specific keywords in inbound call interaction
Recency of last payment
Policy attributes : Type of policyholder/location / type of coverage/Policy cost / Sum assured / Issue age / Policy tenure /
Range of products covered
ACTION : Consider spending a “immersive day in life of session” to surface real world predictors
Theme-10 : Increase in demand for “data driven story tellers”
•
•
•
•
•
Wired for narratives
Package data patterns in narrative format
Ability to connect the dots
From Presenting Insights to presenting ‘data dr
Right brained + left brained
How can I maximize the impact of data patterns to business users ?
Slide 24
The “T” shaped Analytics Individual
Breadth of Knowledge
Depth of
Expertise
Right brained and left brained analytics professional
Frame issues- creative
Blue print – left brained
Interpretation - creative
ACTION : Have the right cross functional team for an analytical project
Theme-11 : Reverse innovation in Analytics
2 real world examples of reverse innovation
Fraud
: Open source
Telemedicine
: Mobile analytics
ACTION : Identify opportunities at the intersection of open source and industry sub processes
Slide 28
Telemedicine – Patient Side VC
Telemedicine – Doctor Side Application
Tabs to view Case sheet,
initiate VC, etc
Patient Case
Sheet, Review
records
Option to view
ECG –
Offline/Live
Patient Vital
Statistics
Telemedicine – Doctor Side
Live ECG and VC.
ECG viewer
Integrated into
application
Patient Side
Video
PIP Option –
Doctor side
Video
Telemedicine analytics framework
1
2
3
4
5
PATIENT DIAGNOSIS
REPOSITORY
Pulse/Height/Weight
-Temperature/Heart rate
-BP/ ECG / SpO2
-- Diagnostics
Rural patient data capture
Satellite transmission
Data reception
Patient repository
Info syndication biz model
11
13
9
Insurance
14
6
Doctors
observations
text miner
Pharmaceutical
Diagnostic text mining
12
Central early warning control
center unit
8
Remote Diagnostics
10
Govt
Hospitals
7
Field health officer
Diagnostic
scanner
( chi square / t
tester )
Secure hosting platform for Rural Patient Diagnostics data
Diagnostic scanning
process
Theme-12 : Feel good insights vs. the one Transformative insight !
3 Real world examples of Transformative insights
1. Outlet analytics : Outlet recommendations resulting in crores of
ruppees of additional revenue
2. People scoring model : Person who did last interview and TAT
were more important predictors than salary
3. Uplift modeling : Sister brand cannibalization
Summary of 12 key analytical trends & themes
 Information Monetization
 Industry specific data models –
ARTS/SCOR/ACCORD
 Solving problems customers never thought
existed
 Analytics penetrates unconventional industries
and processes – Health care, Education
 Handling unstructured information
 Skills to triangulate predictors
 Digital consumer behavior analytics
 Data narrative story telling & “T” shaped
 Value at the intersect - Multi Channel – Online
individuals will
offline
 Reverse Innovation in Analytics – “
 Ensemble models : Segmentation + text mining
 Need for transformative analytics vs “feel good”
analytics
Slide 35
Big Data will explode & Distilling significant few patterns is key
The “Data” refinery
Oil refineries & “Data” refineries
•
Oil fuelled the industrial revolution . Data will fuel the services revolution
•
Raw or unprocessed crude oil is not generally useful. Raw or unprocessed data is also not useful
•
Companies like Google and face book are building large “data refineries” to distill consumer insights
Analytics is going to be an interesting ride for next 5 years !
Key is to Enjoy the journey !!!