Analytics Solutions For Retail Banking - Marketelligent

Analytics Solutions for
Retail Banking
1
Confidential & proprietary information. Property of Ashley Marketelligent Pvt Ltd.
www.marketelligent.com
Marketelligent: Managing Risk & Reward across Retail Banking
Customer
Spend Lend Transact Invest Protect
Card Loan Mortgage CASA Deposit Investments Insurance

•Credit Card •Revolving •Revolving •Current Acct •Term deposit •Unit Trust •Credit
•Charge Card •Installment •Installment •Savings •Unfixed •S. Notes •General
•Revolving •Secured •Bonds •Life
•Installment •Portfolio Finance •Equities
•Insured
Customer Acquisitions Customer Segmentation Marketing Investment Optimization Cross-sell/One-sell

Branch location placement Growing profitable balances MIS
Risk Management Collections & Recoveries Flow of Funds Executive Dashboards
2
And across the Customer Lifecycle
Identify Customers most Identify Customers most likely

Decide Loan pricing likely to default so as to to attrite so as to take
and amount take corrective action proactive actions for retention
Delinquency Scorecards
First pay Default Scorecard
4th or 5th cycle Risk Scorecard
Pricing and Retention
Loan Amount
Approval Scorecards
Conversion Scorecards
Application Fraud Re-Activation
Revenue Scorecards Winback
Collection Scorecards
Profitability Scorecards Self-cures
Decision on who to approve Maximize Collections Target Inactive Customers for

based on expected profitability Efficiencies repeat loans
3
Our Expertise in Risk & Marketing Analytics
Credit Risk
Marketing Analytics Management & Credit Risk Analytics
Training
1. Profit-based Customer 1. To understand existing data / 1. Credit Delinquency Models
Acquisition Strategy reports and present a top-line 2. Other Delinquency Models; eg.
2. Revenue Models; eg. Total 180 "what additional analytics to 5+ cycles bad
days revenue read" 3. Customer Approval and
3. Campaign Management 2. Prepare and deliver the Conversion Models
additional analytics and
4. Cross-sell 4. Optimal Loan Amount, Pricing
highlight key concerns on
5. Retention & Activation and loan duration
policies and processes
6. Loyalty and Winback 5. Forecasting
3. Present credit policy
7. Pricing Analytics changes, collections strategies 6. Collections Analytics
and product program changes 7. Mortgage Portfolio Optimization
to deliver required 8. Fraud Analytics
management deliverables 9. Basel II Analytics
4. Credit Policy Training
4
Marketelligent: A strong P&L discipline to all analytics
Eg. Credit Cards
Bank P&L
Reduce Customer Attrition New Accounts Acquired Acquire New Customers
- Voluntary / Involuntary Accounts Closed - Segments X Products X Channel
- Retention Strategies - Mailbase Expansion
- Winback - Pricing
Account Activation rate
Payment Rate
Improve profitability of Total Ending Receivables
Assets
- Balance Transfer Increasing activation rates
- Credit Line Strategies Interest - Deepening Engagement
- Pricing Cost of Funds - Inactive Customer Treatment
Net Interest Margin
REVENUES
Maximizing Fee Revenue Maximizing Interest Revenue
- Over Credit limit
Risk-based Fees - Product Pricing
- Delinquency Interchange - Customer Behavior –
Affinity Rebates Revolvers, transactors, etc
- Bad Check
Cross-Sell
Annual Fees
Net Credit Losses
EXPENSES
Reduce Net Credit Net Credit Margin Increasing Cross-sell Revenues

Losses - Revenue Enhancing Products
- Credit Line strategies Operating Expenses - Breadth of relationships
- Pricing strategies Loan Loss reserve
- Collections Net Income
Top-down approach
Analytics that impact all line items of the P&L
5
Marketing Analytics
Credit Risk Analytics
6
Profit-based Acquisition
Acquisition strategies that balance Risk & Reward
Objective
Implement a Customer-level profit-based Acquisition Strategy based on
segmentation, predictive models and joint scores
Approval Model
Acquisition
Conversion Model
First Pay Default Model • Individual Scores Flexible

Risk • Strategy Matrix Acquisition
5+ cycle Default Model • Joint Scores Strategy
180 day Revenues

Revenue
Reactivation Model Illustrative
7
Customer Segmentation
Segment Customers to better understand their needs & wants
•Very High utilization •Very high access to •Very high access to •Defined as balance
of revolving credit limits limits
<$500
•Lower access to •Very high balances •High balances, but •Low access to limits
credit as a group •Above average low utilization
•Lowest FICO utilization and risk •Lower risk •Low risk
Key Bureau Attributes

High Moderate
Moderate High Moderate
Over Access, Access, All
Access, Access, Access, Low User
Extended Controlled Controlled Accounts
Heavy User Heavy User Sloppy User
User User
Number of Accounts 2,364,300 4,617,100 2,314,000 4,812,800 2,477,700 12,206,300 10,158,100 38,950,300
Percent of Accounts 6.1% 11.9% 5.9% 12.4% 6.4% 31.3% 26.1% 100.0%
Total Bank Revolving Limits $33,874 $51,621 $200,685 $46,990 $191,763 $59,760 $29,617 $64,554
Total Balance on Revolving Trades $22,150 $18,378 $105,309 $6,308 $21,788 $4,597 $122 $13,417
Utilization of Revolving Trades 65.1% 36.8% 26.8% 14.0% 8.3% 7.2% 0.9% 14.9%
% of Bankcards > 50% utilized 78.0% 32.1% 23.5% 5.7% 5.1% 2.6% 0.5% 12.1%
FICO Score 684 712 725 702 764 782 772 751
•Moderate limits •Moderate limits •Moderate limits

•High balances and •Below average •Low balances and
utilization balances utilization
•Increased risk •Higher risk •Lowest risk segment
Segmentation using SAS PROC FASTCLUS
8
Building Profitable Assets
Balance transfer strategies to build profitable assets
Universe Management Performance Tracking

Eligibility, Risk Universe Testing discipline, MIS
Tracking
Offer Strategy
Segmentation
Offer Strategy Universe Segmentation

Pricing / Duration / Fees Customized Marketing
9
Pricing Analytics
Customer-level pricing to build Deposits
TD Elasticity
Term Deposits Curve Curve
Pricing Sensitivity
200000
Retail Banking 150000

100000
Balance Change
Rate hunter
50000
Moderate Mover
0
Loyal depositor
-50 -40 -30 -20 -10 0 10 20 30 40 50 -50000
Lazy Depositor
-100000
-150000
-200000
deviation From market(bps)
Pricing is one of the most sensitive lever to improve profitability. We can build tools to establish
the price sensitivity of various customer segments. Based on this, pricing strategies can be
developed for different segments to maximise profitability through better margins and/or better
volumes.
10
Cross-selling
Deepen engagement with existing Customers
Installment
Loans
Mortgages
- Which Customer to target
Existing - What Product to Offer

Retail Banking HELOC/FRHEL
Customer - Impact of new product on
existing Product Profitability
Credit Cards
- Overall Profitability
Wealth
Management
11
Retention & Activation
Manage Customers across their Lifecycle
Retention
Value
Activation
Acquisition Usage & Loyalty Time
RETENTION ACTIVATION
• Identify Customers at Risk of Disengagement via • Segment Inactive Customers across various
predictive modeling or activity-based segmentation dimensions
• Take proactive actions via targeted offers • Implement targeted activation campaigns
12
Marketing Analytics
13
Predictive Scorecards to optimize Decisions
1. Approval Scorecards
Customer-level score to decision on which
Eg. Delinquency Scorecard
Customer to approve and which to decline for
100%
New Products (loans, cards, etc) based on
90%
information provided – application data, bureau
80%
data, etc.
Cumulative Default Rate

70%
60%
2. Delinquency Scorecards 50%
Customer-level score to decision on which 40%
Random
Customer is going to default on their loans so as New Model
30%
to enable the business to take proactive actions
20% Existing Model
to minimize losses
10%
0%
3. Collections Scorecards 0 1 2
Score deciles
3 4 5 6 7 8 9 10
Customer-level score to decision on which Predictive Models using SAS PROC LOGISTIC
delinquent Customer has a higher likelihood of
paying back balances; and which Customer is
likely to self-cure; so as to enable business to
optimize Collections activities
14
Eg. First Pay Default Scorecard
High
# Marginal Marginal
Risk Score Range Cuml % FPD Cuml % Non-FPD Cuml % KS
Accounts rate rate
789 887 512 10% 368 72% 23% 144 28% 4% 28.03
756 788 518 20% 271 52% 39% 247 48% 11% 39.36 Model captures
731 755 506 30% 209 41% 52% 297 59% 20% 41.32
710 730 493 40% 177 36% 63% 316 64% 29% 41.88 63% of First Pay
693
680
709
692
583
480
51%
60%
170
111
29%
23%
74%
81%
413
369
71%
77%
40%
51%
39.74
34.65
Defaulters in
659 679 493 70% 103 21% 87% 390 79% 62% 28.33 40% of
615 658 512 80% 86 17% 92% 426 83% 74% 21.12
413 614 521 90% 65 12% 96% 456 88% 87% 11.02
Accounts
180 412 511 100% 59 12% 100% 452 88% 100% 0
Low . . 5129 . 1619 32% . 3510 68% . 41.88
Risk
High # Bad Loans in # Inquiries in

Risk Score # Customers # Bad Loans # Loans given
past 30 days past 60 days
0 512 1.30 1.15 8.85 1.76
1 518 0.55 0.41 10.61 1.41
High Risk
2 506 0.10 0.03 7.43 0.54 Customers have
3 493 0.03 0.01 5.01 0.34
4 583 0.02 0.00 2.55 0.18
a significantly
5 480 0.04 0.01 3.61 0.59 higher # of loan
6 493 0.02 0.01 3.45 0.75
7 512 0.03 0.01 3.45 1.08
inquiries in the
8 521 0.04 0.00 3.10 1.45 past 60 days
9 511 0.05 0.01 3.00 4.26
Low Grand Total 5129 0.22 0.17 5.09 1.23
Risk 15
Control Exposure with right appropriate Lines of Credit
Revenue and Cost
Behavioral Models Drivers Optimal Line Determination Optimal Line Drivers
Revenue
Balance
Model
CMV Actual Predicted V/s Actual Inactivity
300.00
Revolve
250.00
Model LOC
200.00
Predicted Inactivity
150.00
Ideal
100.00
Cost
50.00
Other
0.00
Models Optimal LOC

0 50 100 150 200 250 300 350 400
LOC Predicted
Risk Model
LOC
Illustrative process for assigning
Optimal Line of Credit (LOC)
16
Collections Analytics
Prioritize Customers to action on to optimize Collections efforts
OBJECTIVE
Collect more $ efficiently thereby reducing cost/dollar collected
Unique strategies across various stages: early-stage; late-stage; charge-off/recovery
Profiling & Segmentation Behavior Scorecards Multi-dimensional Selecting Optimal

Analysis Strategies
• Mine Customer & • Rank order accounts • Assess & create smaller • Test and Evaluate
Operational data on a dimension of segments across actions under
• Create broad profiles and interest multiple scores business constraints
segments • Event probabilities : • Assess trade-offs • Typically ‘black-box’,
self sure; charge-off, between strategies heuristically driven
etc models
• Expected value of
Collections
17
Mortgage Portfolio Optimization
OBJECTIVE
Optimal treatment for each Customer so as to maximize NPV given
business constraints
PREDICT OPTIMIZE AUTOMATE

NPV calculations Optimal Treatment Business Rules for
for all possible for each Customer Decisioning
Outcomes
identify likely outcomes of Optimization analytics using Build rules that can be
different actions for action-effect models to deployed through your
different customer select the best action for business rules management
profiles, and the overall each customer, given system
effect on the NPV of business objectives and
portfolio constraints
18
Fraud Analytics
Manage for Fraud while ensuring a positive Customer experience
OBJECTIVE
Minimize Fraud-related Losses and Fraud-related expenses while ensuring a positive Customer experience
More Information Break Conventions Dig Deeper Look for Inconsistencies

Traditional credit scoring Online verification of Verification processes
Responsible use of data is a
and underwriting information beyond a should check for
powerful weapon
procedures do not identify Social Security number is consistencies in address and
against fraud
fraudulent applications needed credit bureau information
LAST
SSN ADDRESS
NAME Neural Networks
Rules-based
FIRST HOME
NAME FRAUD PHONE
NO MATCH
SAME
Illustrative for Insurance Claims

E-MAIL WORK
DOB
ID PHONE 19
Forecasting
Eg. Portfolio-level Charge-off forecasting
A model was built to predict the charge-off rate in the US economy. It performs well in both
observation and validation windows except for two peaks that cannot be attributed to
macroeconomic factors
OVERALL USA CHARGE-OFF RATE (%)
Independent Sign
5.00 Variables
Due to Post Due to change in Total Consumer +
9/11 Bankruptcy law
4.00 Borrowing
Federal Rate +
3.00
Houses for Sale -
2.00 Disposable -
Personal
Income
1.00
Average Weekly -
Earnings
0.00
Financial +
1986Q4 1989Q4 1992Q4 1995Q4 1998Q4 2001Q4 2004Q4 2007Q4
Obligations
Actual Model built Validation Forecast
20
Basel II Analytics
Pillar 1 – Credit Risk & Operations Risk
Our experience in Basel II
Pillar I
Credit Risk Market Risk Operations Risk
Minimum Capital
Same approach as Basel I

Standardised Local, Small Banks
Internal-ratings based
IRB*- Foundation PD inputs provided by bank, rest by Regulator
Multi-line National Banks
Internal-ratings based
IRB*- Advanced PD, EAD, LGD based on inputs provided by bank
Large Global Banks
* IRB - Internal Ratings Based Approach

21
Case Study -
Predictive Model for Identifying High Risk Customers
BUSINESS SITUATION APPROACH

•The client is a leading issuer of • Data from multiple sources(DP
payday loans Bureau, Lexisnexis,Innovis,Teletrack)were leveraged.
•They are experiencing a high default • The Model data was split into Development data and
validation data
rate and need to create a framework
to customize the lending process in • CART analysis was conducted to identify the best
order to reduce the risk exposure predicting variables. ex89_SCR_l 794 - 760 - 734 - 713 - 693 –
MIScore
684 - 667 - 630 - 434 - 182 -
Overall
894 793 759 733 712 692 683 666 629 433
• Correlation analysis for the variables was performed 701 - 718

696 - 700
693 - 695
0.05
0.11
0.13
0.08
0.13
0.06
0.14
0.19
0.09
0.24
0.19
0.16
0.23
0.26
0.14
0.14
0.20
0.17
0.35
0.25
0.30
0.24
0.36
0.33
0.43
0.61
0.52
0.77
0.76
0.80
0.22
0.28
0.25
and variables with values between -0.01 and +0.01 690 - 692
687 - 689
0.09
0.12
0.05
0.10
0.16
0.18
0.14
0.33
0.26
0.27
0.23
0.26
0.34
0.33
0.34
0.30
0.47
0.50
0.64
0.73
0.27
0.30
KEY OBJECTIVES removed 684 - 686

682 - 683
0.03
0.16
0.09
0.21
0.17
0.13
0.28
0.16
0.33
0.17
0.45
0.41
0.40
0.27
0.42
0.44
0.50
0.41
0.71
0.72
0.35
0.32
• Develop a Risk model for first pay • A stepwise Logistic regression was performed and
defaults least significant variables with flipping signs were
dropped
• To score each customer based on their
probability of default
RESULT
METHODOLOGY • Positive coefficients for “No. of inquiries in past 60
days” and “Delinquencies on loans in past 30 days”
• The model was built using the stepwise
showed that number of defaults increases with Lorenz Curve for Bad Lorenz Curve for Good
logistic regression, to estimate the 100%

100%
higher risk exposure 90%

90%
probability of a first pay day default 80% 80%
% Cummulated Goods
% Cummulated Bads
70% 70%
• Through the stepwise logistic regression model, we 60%
50%
60%
50%
40% 40%
were able to identify the explanatory impact of 30%

MI Model
30%
MI Model
r r 1 20%
Random 20%
Random
p = P( Event / x) = P( Bad / x ) = n significant variables on the number of defaults. Thus 10%
0%
10%
0%
0 2 4 6 8 10
1 + exp(−∑α i x )
0 2 4 6 8 10
i
Marketelligent enabled the client to reduce default Percentile of Population ( Highest Risk to Lowe st Risk) Population Perce ntile (Lowe st Risk to Highe st Risk)
i =0
losses by 19%
22
Case Study -
Customer Retention Strategy for a Credit Card Portfolio
BUSINESS SITUATION APPROACH

•The Client was experiencing a • Historical bank internal performance data, and external
significant runoff in its Card credit bureau data was leveraged for the analysis
receivables. • All accounts present in the books as on 1st May 2009
were considered for the analysis
• Customer payment rates had been
climbing up over the past 12 • Account level analysis was conducted to identify the set
months leading to balance erosion. of attrited accounts
• Detailed investigations on the accounts were
•Moreover the balances lost were performed to identify the factors(Balance
high-priced balances contributing in Transfer, Cash Out
a greater proportion to overall Refinancing, etc) contributing to the voluntary attrition
portfolio profitability. • The external macroeconomic environment was studied Decision Tree- Attrition Rate
to understand the customer behavior
• Impact of interest rates on mortgage behavior
KEY QUESTIONS •A predictive model was built to identify the customers BUSINESS RESULTS
most likely to pay off their card balances in the future • Recommendations were implemented in a robust test
• Define and scope the balance erosion
and its impact on Profitability control mode and results were tracked
• Understand internal and external • Lower Card pricing resulted in a 30% drop in balances
factors causing the balance erosion
RECOMMENDATIONS • Results from offering alternative products showed that
• Lower card pricing to accounts identified as most likely test cells had 45% greater response to these products
• Implement strategies to stem the
to pay- off their card balances in the future • Although balances were still lost from cards, the client
erosion and bring the portfolio back
to its ‘normal’ state • Offer HELOC and cash-out refinance to “likely to attrite” was able to save the balances on a different but
accounts with existing mortgage balance. internal product
23
Marketelligent: A strong P&L discipline to all analytics
Eg. Credit Cards
Bank P&L
Reduce Customer Attrition New Accounts Acquired Acquire New Customers
- Voluntary / Involuntary Accounts Closed - Segments X Products X Channel
- Retention Strategies - Mailbase Expansion
- Winback - Pricing
Account Activation rate
Payment Rate
Improve profitability of Total Ending Receivables
Assets
- Balance Transfer Increasing activation rates
- Credit Line Strategies Interest - Deepening Engagement
- Pricing Cost of Funds - Inactive Customer Treatment
Net Interest Margin
REVENUES
Maximizing Fee Revenue Maximizing Interest Revenue
- Over Credit limit
Risk-based Fees - Product Pricing
- Delinquency Interchange - Customer Behavior –
Affinity Rebates Revolvers, transactors, etc
- Bad Check
Cross-Sell
Annual Fees
Net Credit Losses
EXPENSES
Reduce Net Credit Net Credit Margin Increasing Cross-sell Revenues

Losses - Revenue Enhancing Products
- Credit Line strategies Operating Expenses - Breadth of relationships
- Pricing strategies Loan Loss reserve
- Collections Net Income
Top-down approach
Analytics that impact all line items of the P&L
24
ASHLEY MARKETELLIGENT PVT LTD
+91-80-26642802 (India)
1-408-834-8822 (USA)
Thank You info@marketelligent.com


Analytics Solutions For Retail Banking - Marketelligent

Uploaded by

Copyright:

Available Formats

Analytics Solutions For Retail Banking - Marketelligent

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Analytics Solutions For Retail Banking - Marketelligent

Uploaded by

Copyright:

Available Formats

What are the different products and services offered by retail banks?

What are the different products and services offered by retail banks?

How can analytics help retail banks acquire new customers and reduce customer attrition?

How can analytics help retail banks acquire new customers and reduce customer attrition?

Analytics Solutions for

Spend Lend Transact Invest Protect

Card Loan Mortgage CASA Deposit Investments Insurance

Customer Acquisitions Customer Segmentation Marketing Investment Optimization Cross-sell/One-sell

Identify Customers most Identify Customers most likely

Decision on who to approve Maximize Collections Target Inactive Customers for

Reduce Net Credit Net Credit Margin Increasing Cross-sell Revenues

Credit Risk Analytics

First Pay Default Model • Individual Scores Flexible

180 day Revenues

Key Bureau Attributes

•Moderate limits •Moderate limits •Moderate limits

Universe Management Performance Tracking

Offer Strategy Universe Segmentation

Retail Banking 150000

Existing - What Product to Offer

Acquisition Usage & Loyalty Time

Credit Risk Analytics

Cumulative Default Rate

High # Bad Loans in # Inquiries in

CMV Actual Predicted V/s Actual Inactivity

Models Optimal LOC

Unique strategies across various stages: early-stage; late-stage; charge-off/recovery

Profiling & Segmentation Behavior Scorecards Multi-dimensional Selecting Optimal

PREDICT OPTIMIZE AUTOMATE

More Information Break Conventions Dig Deeper Look for Inconsistencies

Illustrative for Insurance Claims

Our experience in Basel II

Same approach as Basel I

* IRB - Internal Ratings Based Approach

BUSINESS SITUATION APPROACH

• Correlation analysis for the variables was performed 701 - 718

KEY OBJECTIVES removed 684 - 686

logistic regression, to estimate the 100%

higher risk exposure 90%

probability of a first pay day default 80% 80%

• Through the stepwise logistic regression model, we 60%

were able to identify the explanatory impact of 30%

p = P( Event / x) = P( Bad / x ) = n significant variables on the number of defaults. Thus 10%

BUSINESS SITUATION APPROACH

Reduce Net Credit Net Credit Margin Increasing Cross-sell Revenues

Confidential & proprietary information. Property of Ashley Marketelligent Pvt Ltd.

You might also like