This presentation described Big Data concept. Then it shows example of applications in Banking. The presenter is Dr. Tuangtong Wattarujeekrit in Big Data Analytics Day event.
RABBIT: A CLI tool for identifying bots based on their GitHub events.
Apply (Big) Data Analytics & Predictive Analytics to Business Application
1. Apply (Big) Data Analytics
& Predictive Analytics
to Business Application
Tuangthong Wattarujeekrit, Ph.D.
23/Sep/2017
2. Who am I ?
Working Experience
2014-current: Senior Customer Data Analytics & Intelligence Specialist
TMB Bank Public Company Limited, Bangkok, Thailand
2008-2013: Manager, Data Mining Department
Total Access Communication (DTAC), Bangkok, Thailand
2006-2007: Business Analyst/System Integration Engineer
(AIS Thailand and DiGis Telecommunication Malaysia projects)
WEDO Consulting, Co. Ltd., Bangkok, Thailand
1996-1999: Computer Engineer,
Head of System Management Control Division
Hoya Glass Disk (Thailand) Ltd., Lamphun Factory, Thailand
Education
Doctor of Philosophy (Informatics), September, 2005
Department of Informatics, The Graduate University for Advanced Studies, Tokyo, Japan
Master of Computer Engineering, October, 2002
Department of Computer Engineering, Kasetsart University, Bangkok, Thailan
Bachelor of Computer Engineering, March, 1996 (First Class Honour)
Department of Computer Engineering, Chiangmai University, Chiangmai, Thailand
5. Big Data = Data in high scale
or high complexity
Big Data = Data in different forms
Video
Photo
Voice
Natural Language
Tuangthong W. (23/Sep/2017)
6. Big Data = Data in motion
Time–to-Value of Data
Big Data = Data in doubt
Language
Ambiguity
Data
Incomplete
Data
Deception
Tuangthong W. (23/Sep/2017)
7. Tuangthong W. (23/Sep/2017)
Source of Information
“Think about your Customer Journey”
Customer
Input
Financial
Transaction
Touch
Point
Social
Media
16. Unsupervised Leaning to separate the data items into
subsets
Data points in one cluster are more similar to one another
Data points in separate clusters are less similar to one
another
Clustering/ Segmentation
Tuangthong W. (23/Sep/2017)
19. Supervised-Learning to analyze current data &
historical facts to determine patterns, then predict
• Classify unknown
• What might happen in the future
• Predict potential opportunities
Classification/ Prediction
Tuangthong W. (23/Sep/2017)
20. Help to Find the Right Target!
Reduce Cost of Execution/Marketing
Increase Revenue/Market Share
Uses of Predictive Analytics
Tuangthong W. (23/Sep/2017)
24. X-Sell
Tuangthong W. (23/Sep/2017)
Start with thinking about mission
• Right Product
• Right Time
• Right Channel
Up-Sell Retention
Need Revenue Growth, with maintaining
Good Customer Experience
Analysis to know “What Offer is Relevant to the customer?”
Saving
High Yield Saving
Term-Deposit
Bank Assurance
Saving
Mutual Fund
Credit-Card
Personal Loan
25. Tuangthong W. (23/Sep/2017)
Know your customer
• Who’s your customer ?
• What’s your customer life-style & preference ?
• What’s your customer network ?
• What’s your customer location ?
• What’s your customer personality ?
• What’s your customer financial?
26. Tuangthong W. (23/Sep/2017)
Derive your 360-view of customer
“Getting Attributes or Predictable Fields to do Analytics”
• Identification
• Age
• Gender
• Income
• Residence
• Financial Plan
Demographics
• Inflow, Outflow
• Outstanding
• CC spending
• Loan Payment
• Product holding
• Networking
Financial Usage
• Branch
• ATM/ADM
• Mobile Banking
• Internet Banking
• Web-Site, Facebook
• Contact Center
Channel Behavior
• Preference e.g. Hang-Out place
• Traveler?, Shopper?
• Wealth Level
• Interest Event
• Event-of-Life
• Personality
Life-Style
27. Tuangthong W. (23/Sep/2017)
Use-Case “MASS Promotion Target”
Business Return
Key Idea
• Precision vs. Coverage
Operational Possibility
• Propensity Model
vs.
• Market Basket Analysis
28. Tuangthong W. (23/Sep/2017)
Use-Case “Propensity to Buy Product”
Sub-Target
Key Idea
• Channel
• Customer Segment
• Execution Segment
Feature Engineering
• SNA (Word-of-Mouth)
29. Tuangthong W. (23/Sep/2017)
Use-Case “Propensity to Churn”
Churner = Customers who stop using your service
Pre-Analysis to define Target Definition
Key Idea
• Who you can win-back
• Account Level or Customer Level
30. Tuangthong W. (23/Sep/2017)
Use-Case “Behavior Segment”
Product Holding
Define Segment Universe
Key Idea
• Expected Strategic Outcome
Grouping/ Re-Grouping
• Direct Machine-Learning outcome
need to be refined by value-chain
33. Morgan Stanley
• Real-time predictive analytics from Web-Log and DB-Log to understand who did
what, how, when and what caused the market issue
• Detect market freak out
• Better recommendations for their investments in stocks
Tuangthong W. (23/Sep/2017)
34. Bank of America
• Transactional data of 50 million customers
• Raise the bar from sampling-analysis to the full customer set (all channel and
interaction) by using Big Data technology
• Propensity to buy model to appeal offers to well-defined customer segments
The largest bank
in US
Tuangthong W. (23/Sep/2017)
35. Commonwealth Bank
• 9 millions transactions per day (40% of card transactions in Australia); 12
million account profiles
• Real-time analytics to create personalized services to customers both in
person and online
Tuangthong W. (23/Sep/2017)
36. • Analytics customer basic profiles, their services used, their business,
market trend for personalized financial advice for each customers
• Less frequent that customers have to meet-up with the financial advisor
Tuangthong W. (23/Sep/2017)
37. • Analysis on average
spending habits of people in
that demographic (such as
monthly shopping, housing,
communication costs) both
from UBank transactions and
customers’ input
• [PeopleLikeU] application
(which is not survey-based,
but it’s real transactional
data) to compare and
benchmark the spending
habits of different types of
people
Tuangthong W. (23/Sep/2017)
38. • a wide range of sources to decide loan approval
• For example, whether a customer has GitHub account
Tuangthong W. (23/Sep/2017)
39. • Use bulk of data other than simple credit report
• For example, how borrower uses smartphones and social network
• This reduces 40% default rate.
Tuangthong W. (23/Sep/2017)
40. • Solve the limited traditional historical credit bureau data
• 4.5 billions of “credit invisibles”
• Predictive algorithms to customer scoring from some self-declared data and
other reliable sources, such as LinkedIn
• Be able to increase acceptance by 14%
Tuangthong W. (23/Sep/2017)
43. Special Thanks
Customer Dynamic Marketing team @TMB, especially
1. Boontarika Maythayodom
2. Tanaporn Tunyaset
3. Pornnareumol Kaewyok
who make TMB case-studies to go-live.