Driving Growth
And Profitability
• Scoring Modeling
• Programming
• Price OptimizationAnd
HERMAN JOPIA | DATA ANALYTICS MANAGER | AMERICAN SAVINGS BANK
PREDICTIVE ANALYTICS WORLD BUSINESS | CHICAGO | JUNE 21, 2017
HAWAII
MARKET
13%
AMERICAN
SAVINGS BANK
44%
Source: Assets’ Growth 2016. FDIC Call Report, item RCONK207.
1.4 M
Source: Estimation based on Census and Credit Bureau Data.
~ 0.6 M
Source: Federal Reserve Bank of St. Louis.
673
691
Source: State of Credit 2016, Experian.
MOST EXPENSIVE
US CITIES1
3RD
DEBT TO
INCOME2
2ND
[ HONOLULU ] [ 2.1 ]
Sources: 1Kiplinger 2017, 2Smart Asset 2017.
Sources: 1McKinsey Global Institute 2017. 2Texas A&M University (Map).
ANALYTICS SKILLS GAP | A CONTINUING SHORTAGE OF TALENT1
Targeting + Optimization
The right offer for the right person
Develop talent and build
the analytics infrastructure
9
Explore new segments
Beyond traditional lending
“Business as usual”
was not going to work
OBJECTIVE: GROWTH + PROFITABILITY
1
Data Management Reporting & Analytics Support Mgmt.
Decisions
Reporting
Proposal
(Negotiation)
Reporting
Engine
Analytics
Tools
Data
(Internal)
Implementation
Insights
Data
(External)
Data
Mart
ETL
Data
(Internal)
Data
Analysis
DATA-DRIVEN CULTURE | MARKETING ANALYTICS
→ INVEST IN PEOPLE AND INFRASTRUCTURE ←
DATA-DRIVEN CULTURE | STARTS AT THE TOP
Source: Hawaii Business Magazine
DATA-DRIVEN CULTURE | TALENT | ANALYTICS INTERNSHIP
Source: American Savings Bank Website (Careers)
ANALYTICS INTERNSHIP | TRAINING | BANKING | DATA MGMT | STATISTICAL TOOLS
AMERICAN SAVINGS BANK | THE BEST PLACE TO WORK
1.0
2.9
44%
Understand
The Market
Find High Value
Opportunities and
compete wisely
Understand Your
Business
What drives
the profitability
of your business
Understand Your
Capabilities
Make sure you can
implement and
execute your ideas
Evaluate
Results / Models
Yes, metrics … But
are you getting
the expected results?
1.4 M
~ 0.6 M
Source: Estimation based on Census and Credit Bureau Data.
Scoring
Models
UNDERSTAND | THE MARKET
+ VALUE
BUSINESS AS USUAL … BUT BETTER
1 digit
2 digits
UNDERSTAND | CAPABILITIES
Raw Data
(1+ Sources)
Targeted
Population
RULES + SCORING MODELS
Response
No
Response
Returned
Mail
Interest Income
Non Interest Income (Fees)
Funding
Maintenance
Servicing
Credit Risk
Printing
Postage
Data
Response Rate
INCOME > EXPENSES
UNDERSTAND | BUSINESS
Loan
SCORING MODEL | WHAT IS IT?
720
CHARACTERISTICS ALGORITHM SCORE
SCORE >= 660
SCORE >= 620
SCORE < 620
“Prime”
“Non Prime”
“Subprime”
SCORING MODEL | SCORECARD
CHARACTERISTIC
BINS
(Attributes)
POINTS
EXAMPLE: RESPONSE MODEL
SCORING MODEL | DEVELOPMENT
Implementation
monitoring
Reporting
recalibration
Data Binning
Correlation
Modeling
Sampling DocumentationBusiness
case
Generation of
predictive characteristics
PERFORMANCE
(ODDS) CHART
SKILLS
SCORING MODEL | EXAMPLE OF BINNING
Is a Credit Score (CS1) associated to respond to a Credit Offer (FResponse)?
CS1: Numeric ; FResponse Binary (1: Response, 0: No Response) ; N = 100,000
SCORING MODEL | BINNING
1 Meaningful groups
2 Assumption about the relationship
3 Measure of association
• IV > 0.3: Strong
• IV < 0.1: Weak
SCORING MODEL | OPTIMAL BINNING
1
2
3
4
OPTIMAL BINNING | R PACKAGE ‘smbinning’
# Once the data is loaded in R ...
> result = smbinning(df=dfpultrain, y=“FResponse”, x=“CS1”, p=0.05)
# Plot Response Rate
> smbinning.plot(result,option=“goodrate”,sub=“Credit Bureau ...”)
# Information Value
> result$iv
[1] 0.4627
www.scoringmodeling.com
PRICE OPTIMIZATION | DIRECT MAIL
Raw Data
(1+ Sources)
Targeted
Population
RULES + SCORING MODELS
Response
No
Response
Returned
Mail
Interest Income
Non Interest Income (Fees)
Funding
Maintenance
Servicing
Credit Risk
Printing
Postage
Data
Response Rate
Profits = Income - Expenses
Loan
Credit Model (s)
Fees
MaintenancePostageData
Profits
Credit Risk Cost of Funds
Interest Income
Printing
Response Model
Term APR Amount
Principal
Reduction
Early
Prepayment
Price
Sensitivity
Low Interest
Rate Environment
Expenses
Income
Key Drivers
External Factors
Skills
PERSONAL LOAN | PROFITS’ DRIVERS
PERSONAL LOAN | INCOME
Loan Amount
$10,000
Int. Rate (APR)
10%
Term
48 Months
Monthly
Payment
$254
Total
Payment
$12,174
x 48
Interest
Income
$2,174
Paid on month 30?
Additional 50 $/month
No payments at all?
$1,831
$1,739
($10,000)
EARLY
PREPAYMENT
PRINCIPAL
REDUCTION
CREDIT
LOSS
PRICE OPTIMIZATION | PERSONAL LOAN
Rate
Response
10%
14%
120 (1.2%) 200 (2%)
Profit ($K)
Response
200
220
120 (1.2%) 200 (2%)
Int. Income
- Credit Risk
- Other Op.
Revenue (%)
Revenue ($)
Offers (#)
Response (%)
Response (#)
Revenues
- Marketing
Profit ($)
10.00%
3.00%
2.00%
14.00%
3.00%
2.00%
5.00%
$1,100
9.00%
$2,000
10,000
2.00%
200
$220K
$20K
Loan $10,000 $10,000
$200K
10,000
1.20%
120
$240K
$20K
$220K
240
12.5%
Price
Sensitivity
PRICE OPTIMIZATION | SEGMENTATION
A
B
C
D
E
TESTCONTROL
Credit
Risk
Price
Sensitivity
VERY HIGH
HIGH
MODERATE
LOW
VERY LOW
SUMMING UP → BUILD YOUR UNIQUE PATH TO SUCCESS
• Find and hire the right people.
• Understand your market, business, and capabilities. Then apply PA.
• Demonstrate the benefits of PA and earn trust → It’s all about results.
• Results are not about response rates or profitability → It’s about profits.
• Embrace complexity:
• Programming saves time, a lot of time (Example: smbinning ).
• Scoring models help to make better decisions (Example: Response).
• Apply “real life” optimization (Moving targets, Dynamic constraints).
• Data (lack of data) is never the issue → Make assumptions, test, repeat.
• Be the first in the market.
• “Business as usual” will never get you where you want/need to be.
Driving Growth
And Profitability
• Scoring Modeling
• Programming
• Price OptimizationAnd
HERMAN JOPIA | DATA ANALYTICS MANAGER | AMERICAN SAVINGS BANK
PREDICTIVE ANALYTICS WORLD BUSINESS | CHICAGO | JUNE 21, 2017

1315 keynote jopia_shareable

  • 1.
    Driving Growth And Profitability •Scoring Modeling • Programming • Price OptimizationAnd HERMAN JOPIA | DATA ANALYTICS MANAGER | AMERICAN SAVINGS BANK PREDICTIVE ANALYTICS WORLD BUSINESS | CHICAGO | JUNE 21, 2017
  • 2.
    HAWAII MARKET 13% AMERICAN SAVINGS BANK 44% Source: Assets’Growth 2016. FDIC Call Report, item RCONK207.
  • 3.
    1.4 M Source: Estimationbased on Census and Credit Bureau Data. ~ 0.6 M
  • 4.
    Source: Federal ReserveBank of St. Louis.
  • 5.
    673 691 Source: State ofCredit 2016, Experian.
  • 6.
    MOST EXPENSIVE US CITIES1 3RD DEBTTO INCOME2 2ND [ HONOLULU ] [ 2.1 ] Sources: 1Kiplinger 2017, 2Smart Asset 2017.
  • 7.
    Sources: 1McKinsey GlobalInstitute 2017. 2Texas A&M University (Map). ANALYTICS SKILLS GAP | A CONTINUING SHORTAGE OF TALENT1
  • 9.
    Targeting + Optimization Theright offer for the right person Develop talent and build the analytics infrastructure 9 Explore new segments Beyond traditional lending “Business as usual” was not going to work OBJECTIVE: GROWTH + PROFITABILITY
  • 10.
    1 Data Management Reporting& Analytics Support Mgmt. Decisions Reporting Proposal (Negotiation) Reporting Engine Analytics Tools Data (Internal) Implementation Insights Data (External) Data Mart ETL Data (Internal) Data Analysis DATA-DRIVEN CULTURE | MARKETING ANALYTICS → INVEST IN PEOPLE AND INFRASTRUCTURE ←
  • 11.
    DATA-DRIVEN CULTURE |STARTS AT THE TOP Source: Hawaii Business Magazine
  • 12.
    DATA-DRIVEN CULTURE |TALENT | ANALYTICS INTERNSHIP Source: American Savings Bank Website (Careers) ANALYTICS INTERNSHIP | TRAINING | BANKING | DATA MGMT | STATISTICAL TOOLS
  • 13.
    AMERICAN SAVINGS BANK| THE BEST PLACE TO WORK
  • 14.
  • 15.
    Understand The Market Find HighValue Opportunities and compete wisely Understand Your Business What drives the profitability of your business Understand Your Capabilities Make sure you can implement and execute your ideas Evaluate Results / Models Yes, metrics … But are you getting the expected results?
  • 16.
    1.4 M ~ 0.6M Source: Estimation based on Census and Credit Bureau Data. Scoring Models UNDERSTAND | THE MARKET + VALUE BUSINESS AS USUAL … BUT BETTER 1 digit 2 digits
  • 17.
  • 18.
    Raw Data (1+ Sources) Targeted Population RULES+ SCORING MODELS Response No Response Returned Mail Interest Income Non Interest Income (Fees) Funding Maintenance Servicing Credit Risk Printing Postage Data Response Rate INCOME > EXPENSES UNDERSTAND | BUSINESS Loan
  • 19.
    SCORING MODEL |WHAT IS IT? 720 CHARACTERISTICS ALGORITHM SCORE SCORE >= 660 SCORE >= 620 SCORE < 620 “Prime” “Non Prime” “Subprime”
  • 20.
    SCORING MODEL |SCORECARD CHARACTERISTIC BINS (Attributes) POINTS EXAMPLE: RESPONSE MODEL
  • 21.
    SCORING MODEL |DEVELOPMENT Implementation monitoring Reporting recalibration Data Binning Correlation Modeling Sampling DocumentationBusiness case Generation of predictive characteristics PERFORMANCE (ODDS) CHART SKILLS
  • 22.
    SCORING MODEL |EXAMPLE OF BINNING Is a Credit Score (CS1) associated to respond to a Credit Offer (FResponse)? CS1: Numeric ; FResponse Binary (1: Response, 0: No Response) ; N = 100,000
  • 23.
    SCORING MODEL |BINNING 1 Meaningful groups 2 Assumption about the relationship 3 Measure of association • IV > 0.3: Strong • IV < 0.1: Weak
  • 24.
    SCORING MODEL |OPTIMAL BINNING 1 2 3 4
  • 25.
    OPTIMAL BINNING |R PACKAGE ‘smbinning’ # Once the data is loaded in R ... > result = smbinning(df=dfpultrain, y=“FResponse”, x=“CS1”, p=0.05) # Plot Response Rate > smbinning.plot(result,option=“goodrate”,sub=“Credit Bureau ...”) # Information Value > result$iv [1] 0.4627 www.scoringmodeling.com
  • 26.
    PRICE OPTIMIZATION |DIRECT MAIL Raw Data (1+ Sources) Targeted Population RULES + SCORING MODELS Response No Response Returned Mail Interest Income Non Interest Income (Fees) Funding Maintenance Servicing Credit Risk Printing Postage Data Response Rate Profits = Income - Expenses Loan
  • 27.
    Credit Model (s) Fees MaintenancePostageData Profits CreditRisk Cost of Funds Interest Income Printing Response Model Term APR Amount Principal Reduction Early Prepayment Price Sensitivity Low Interest Rate Environment Expenses Income Key Drivers External Factors Skills PERSONAL LOAN | PROFITS’ DRIVERS
  • 28.
    PERSONAL LOAN |INCOME Loan Amount $10,000 Int. Rate (APR) 10% Term 48 Months Monthly Payment $254 Total Payment $12,174 x 48 Interest Income $2,174 Paid on month 30? Additional 50 $/month No payments at all? $1,831 $1,739 ($10,000) EARLY PREPAYMENT PRINCIPAL REDUCTION CREDIT LOSS
  • 29.
    PRICE OPTIMIZATION |PERSONAL LOAN Rate Response 10% 14% 120 (1.2%) 200 (2%) Profit ($K) Response 200 220 120 (1.2%) 200 (2%) Int. Income - Credit Risk - Other Op. Revenue (%) Revenue ($) Offers (#) Response (%) Response (#) Revenues - Marketing Profit ($) 10.00% 3.00% 2.00% 14.00% 3.00% 2.00% 5.00% $1,100 9.00% $2,000 10,000 2.00% 200 $220K $20K Loan $10,000 $10,000 $200K 10,000 1.20% 120 $240K $20K $220K 240 12.5% Price Sensitivity
  • 30.
    PRICE OPTIMIZATION |SEGMENTATION A B C D E TESTCONTROL Credit Risk Price Sensitivity VERY HIGH HIGH MODERATE LOW VERY LOW
  • 31.
    SUMMING UP →BUILD YOUR UNIQUE PATH TO SUCCESS • Find and hire the right people. • Understand your market, business, and capabilities. Then apply PA. • Demonstrate the benefits of PA and earn trust → It’s all about results. • Results are not about response rates or profitability → It’s about profits. • Embrace complexity: • Programming saves time, a lot of time (Example: smbinning ). • Scoring models help to make better decisions (Example: Response). • Apply “real life” optimization (Moving targets, Dynamic constraints). • Data (lack of data) is never the issue → Make assumptions, test, repeat. • Be the first in the market. • “Business as usual” will never get you where you want/need to be.
  • 32.
    Driving Growth And Profitability •Scoring Modeling • Programming • Price OptimizationAnd HERMAN JOPIA | DATA ANALYTICS MANAGER | AMERICAN SAVINGS BANK PREDICTIVE ANALYTICS WORLD BUSINESS | CHICAGO | JUNE 21, 2017