20. DATA
MINING
Process of discovering meaningful correlations, patterns and
trends by sifting through large amounts of data stored in
repositories.
Data mining employs pattern recognition technologies, as
well as statistical and mathematical techniques
21. • Evolution of data analysts
• Amalgamation of Mathematics,
Computer Science, Applied
Statistics and Business
Knowledge
• Imagine driving…
DATA SCIENCE
22. APPLICATIONS
Acquire customers:
• Understand who your best
customers are
• Connect with them in the right
ways
• Take the best action maximize
what you sell to them
PREDICTIVE CUSTOMER ANALYTICS
23. Grow customers:
• Understand the best mix of things
needed by your customers and
channels
• Maximize the revenue received
from your customers and channels
• Take the best action every time to
interact
APPLICATIONS
PREDICTIVE CUSTOMER ANALYTICS
24. Retain customers:
• Understand what makes your
customers leave and what
makes them stay
• Keep your best customers
happy
• Take action to prevent them
from leaving
PREDICTIVE CUSTOMER ANALYTICS
APPLICATIONS
25. APPLICATIONS
PREDICTIVE OPERATIONAL ANALYTICS
Manage operations:
• Maximize the usage of your assets
• Make sure inventory and resources are in the
right place at the right time
• Identify the impact of investment
27. Maximize capital efficiency:
• Improve the efficiency and
effectiveness of your assets
• Reduce operational costs
• Drive operational excellence
in all phases: procurement,
development, availability and
distribution
APPLICATIONS
PREDICTIVE OPERATIONAL ANALYTICS
29. Detect suspicious activity:
• Identify fraudulent patterns
• Reduce false positives
• Identity collusive and fraudulent
merchants and employees
• Identify unanticipated transaction
patterns
APPLICATIONS
PREDICTIVE THREAT AND FRAUD ANALYTICSPREDICTIVE THREAT AND FRAUD ANALYTICS
30. Control outcomes:
• Take action in real-time to
prevent abuse
• Reduce Claims Handling Time
• Alert clients of transaction fraud
APPLICATIONS
PREDICTIVE THREAT AND FRAUD ANALYTICSPREDICTIVE THREAT AND FRAUD ANALYTICS
32. SUCCESS STORIES
• Telco Customer Segmentation and Churn
Prediction
– Zong China Mobile
• Predictive Maintenance and Quality – internal
IBM project
• Injury Prevention and Talent Identification
– Auckland Blues Rugby Franchise
33. CUSTOMER BEHAVIOR SEGMENTATION AND
CHURN PREDICTION ZONG CHINA MOBILE
Zong China Mobile’s key focus
was to build it’s advanced
analytics capability which could
enable them explore more
revenue channels, reduce the
increasing churn and monetise
their existing assets
34. Business challenge
• The client was facing a high customer
churn rate and the success of this project
was critical for client’s market dominance
• Churn models with 70% accuracy for
predicting timely churn were required
• The data provided for modeling was limited
and noisy
CUSTOMER BEHAVIOR SEGMENTATION AND
CHURN PREDICTION ZONG CHINA MOBILE
35. Solution
• Churn Prediction models for prepaid, post-paid and early
churn subscribers
• Customer Segmentation for identifying Behavioral
(Revenue, Usage, Dormancy) clusters
• Social network analysis – churn based on influencers
• Identification of Leaders and distinct User Groups
• Revenue Forecasting for zero balance subscribers and the
monthly value Migration model for complete subscriber
base.
• Best Bundle Prediction
CUSTOMER BEHAVIOR SEGMENTATION AND
CHURN PREDICTION ZONG CHINA MOBILE
36. Benefits
• Identify indicators of customer churn using CDRs
and call center data for tailoring marketing efforts
as a result
• 50% improvement in churn detection and an
initial reduction in average gross churn per month
of 0.43% (approx. 1 million customers).
• Churn reduction resulted saving on potential
revenue loss of half a million USD.
Components
• IBM SPSS Modeler
50%
CUSTOMER BEHAVIOR SEGMENTATION AND
CHURN PREDICTION ZONG CHINA MOBILE
37. IT OPERATIONAL ANALYTICS
INTERNAL IBM PROJECT
Enable data-based decisions to
guide automation and
optimisation decisions to improve
operations, quality and customer
satisfaction
38. Business challenge
• Prevent outages from prescribing balanced
configuration of IT assets
• Intercept outages from detecting operational
patterns
• Optimise client experience by prioritisation of
requests
• Root cause isolate problems rapidly and identify
remediation
IT OPERATIONAL ANALYTICS
INTERNAL IBM PROJECT
39. Solution
• Incident Reduction: Analyse structured and
unstructured tickets data to link incidents to top drivers by
usage type for remediation actions
• Defect Prevention: Analyse ticket data of various types
(i.e., change and incident) to identify patterns and triggers
that can be proactively fixed to prevent similar defects to
propagate
• Voice Operation Issues: Analyse Voice operations ticket
to identify issues from current operations.
IT OPERATIONAL ANALYTICS
INTERNAL IBM PROJECT
46. AUCKLAND BLUES
INJURY PREVENTION MODEL
Using sports performance tracking
devices, coupled with medical data
and well-being, Auckland Blues are
predicting the likelihood of a player
getting injured during the game and
practicing off-season. This helps Blues
to modify the training loads and
overcoming performance thresholds of
players.
47. Business challenge
• Decrease soft tissue injuries
• Study the thresholds of
performance for each player and
positions they play at
• Root cause analysis for known
injuries
AUCKLAND BLUES
INJURY PREVENTION MODEL
48. Solution
• Used rfid based tracking modules to
calculate the acceleration, distance
traveled, intensity and other factors to
track performance along with medical
data, past injuries, wellbeing and
demographic data to predict the likelihood
of injuries in players.
• Root cause analysis was done on the
known injuries to ensure future prevention
of injuries.
Components
• IBM SPSS Modeler – Batch Mode
AUCKLAND BLUES
INJURY PREVENTION MODEL
53. Why an advanced analytics capability?
• Border transactions and passenger
movements continue to climb exponentially
• Do not have infinite resources at the frontline
• Must facilitate trade and travel as well as
“protect the border”
54. Increase in trade transactions
Imports: from approx $40b in 2007 to estimated $60b in 2018
55. Travel volumes are growing
0
50,000
100,000
150,000
200,000
250,000
2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 2013/14
Total arriving cruise ship passengers and crew processed by Customs
Cruise passengers: from approx 60,000 in 2006/07 to 235,000 in 2013/14
56. …and are forecast to increase
0
500,000
1,000,000
1,500,000
2,000,000
2,500,000
3,000,000
3,500,000
20002001200220032004200520062007200820092010201120122013201420152016201720182019
Other
US
UK
Korea
Japan
Germany
China
Canada
Australia
Passenger arrival volumes (historical and forecast) by nationalities (excluding New
Zealand passport holders)
Air Arrivals: from 1.7m in 2000 to estimated 3.2m in 2019
57. Current risk assessment process
Current system:
• Ahead of its time but now challenged by
volumes of data
• Relies on experience and knowledge of
officers to create alerts and profiles for
further examination purposes
58. Potential benefits?
Opportunity with big data tools to:
• Enlist tools to do the work human brains
cannot (e.g. correlate and interpret big data)
• Gain insights not currently available
• Use information strategically to prioritise work
and ensure effective interventions
59. Lessons we are learning along the way…
Understanding and preparing Customs for:
• How analytics is going to change the way we work, and even the
way we organise ourselves
• Making sure key resources are available to support analytics:
• Factor in resource for data wrangling/data infrastructure
• Right mix of staff capabilities for this organisation
• Staff are properly trained – on our data
Data data data!
• Change processes to improve the quality of our data – e.g. use
mandatory fields
• Data culture: not just in the backroom
62. • Investment up-front software
• Solution architecture and services fees
• Hardware cost
• On-going maintenance and updates
• Data Scientist headcount
• Investment between 250k – 750k*
TRADITIONAL
APPROACH
BUYING AND DEPLOYING SOLUTION
63. • Industry specific solutions
• Customer Analytics, Predictive Maintenance, Operational
Analytics and more
• Automated solutions (batched and real time) and available as
web service
• Embedded into existing infrastructure of your business
MI PREDICT
CLOUD HOSTED AND MANAGED SOLUTION
Quick Intro to what Predictive Analytics is all about.
https://www.youtube.com/watch?v=w1-hbFOytNg&index=6&list=PLd3MHD3LUqCw6YOpAw-iOw3xknNWaWM-4
Check this out
https://youtu.be/iiv4X1K7iX4
What are the factors that determine their likelihood to do so?
It’s not just grades…
Distance
Multiple jobs- night shift
One with other responsibilities
Who can predict the intentions of a storm cloud?
Power companies have to.
Seven out of ten power outages in the US are caused by weather.
Utilities can’t send crews to chase every storm
So they’re working with IBM to combine micro weather forecasts with detailed local data
from sensors analyzing
Topography
soil saturation
even the number of trees.
So they can predict
within a few city blocks
where an outage is most likely to occur
And send crews exactly where they’re needed when they’re needed.
https://youtu.be/cnaAL7wUzPw
Two thousand sensors detecting changes in
temperature,
Vibration,
alignment.
https://youtu.be/hDFe2uWxFO8
A single cracked wheel can cause a derailment
Out of more than one and a half million wheels,
how do you find the right one?
By listening
You may not be able to hear it
But analytics can.
a railroad analyzes one hundred thousand data points a day
To help stop trouble
before it starts.
MRI’s growing by 10% year on Year
More and more, data is visual.
In fact, the number of MRIs has increased by ten percent a year.
And a radiologist might view a thousand images to find one tiny abnormality
in shape, contrast or movement.
Because it’s so challenging,
Analytics is being used to help clinicians spot key patterns quickly and precisely.
https://youtu.be/hH2pCP5fBEk
https://youtu.be/Mi4ZPtKSbMs
In Africa 1 in 5 malaria pills are counterfeit. Predictive analytics is being used to detect which ones are.
We’ll here more about this sort of thing in Gillian’s presentation.
Handing over Farhad. Hear is a perspective on how you can use predictive to reshape customer experiences.
https://www.youtube.com/watch?v=b82Quinl7ac
Setting the scene:
Data analysts vs data scientists, how they are different and what that means for organisations that want to drive actionable insights out of their data using predictive analytics.