Data Modeling & Bureau Scoring Experian for CreditChex Karachi Nov. 29 th 2007
Experian Decision Analytics Credit Services Help clients with data and services to make business critical decisions in credit and fraud Sales ($ M) 1,520 % of Sales 44% Growth 4% People 6,500 Decision Analytics Help clients lend profitably to businesses and consumers, maximising revenue, minimising risk, controlling fraud Sales ($ M) 392 % of Sales 12% Growth 16% People 1,500 Part of the Experian Group Marketing Solutions Help clients to acquire new customers and develop and manage relationships with existing customers Sales ($ M) 728 % of Sales 21% Growth 13% People 3,500 Interactive Help consumers connect with companies over the Internet to sell products from our clients or from Experian Sales ($ 784 M) 23% % of Sales 40% Growth 1000 People Annual sales in excess of $3.6 billion 12,500 people worldwide Offices in 36 countries and clients in over 60 countries Products and services in more countries than any competitor Group results for 2007
Worldwide presence, local focus and. Offices in 30 countries Turkey Russia Norway Japan Germany Denmark Canada Brazil USA UK Spain South Africa Singapore Romania Poland Netherlands Monaco Italy India Ireland Hong Kong Greece France Finland China Chile Bulgaria Austria Australia Argentina Clients in over 60 countries Argentina Australia Austria Azerbaijan Belgium Bolivia Brazil Bulgaria Canada Chile China Colombia Costa Rica Croatia Cyprus Czech Republic Denmark Egypt Finland France Germany Greece Honduras Hong Kong Hungary India Indonesia Iran Ireland Italy Japan Jordan Kazakhstan Lebanon Malaysia Malta Mauritius Mexico Morocco Netherlands New Zealand Nigeria Norway Panama Peru Philippines Poland Polynesia Portugal Puerto Rico Romania Russia Saudi Arabia Senegal Singapore Slovakia Slovenia South Africa South Korea Spain Sweden Switzerland Taiwan Thailand Turkey UK Ukraine USA Venezuela
.client relationships Over 700 clients in more than 60 countries Partnerships with multi-national organisations
Who is Experian Decision Analytics? Delivering leading practices in Credit Risk Management processes and Fraud prevention 1. Worldwide presence, local focus and client relationships The worldwide company delivers experience and expertise developed from working with clients around the globe. Local offices build a long term partnership with clients, delivering and supporting solutions that create long-term value 2. Leaders in the field Specialists in risk and fraud management in financial services and media services with an unparalleled insight creating a unique combination of industry experience, worldwide presence, software expertise and analytical rigour 3. Delivering value Solutions that cover the complete customer relationship, for a range of industries and business size. The approach is always value driven, with solutions that allow organisations to focus on what really adds value to the business
Leaders in the field A trusted partner with leading organisations worldwide Working with more than 700 clients using decision analytics technology Helping clients to make more than 20 billion decisions every year Delivering value and real business benefits Building decision analytics solutions since 1988 Dedicated to decision analytics, a single provider integrating data, analytics, consulting & technology for a complete solution
Delivering value Across the customer relationship Origination Take on the right customers with the right terms to maximise return, minimise bad debt and preventing fraud Customer Management Manage each customer according to their value and potential to maximise customer relationships Revenue management and collections Manage each customer appropriately to maximise collections effectiveness and minimise write-offs Across industries unsecured lending, telecommunications and media services, retail banking, SME, automotive, mortgage Across business size and market stage Start-up Single country Emerging and dynamic No bureaux capability to to to to Established Multi-national and global Mature, sophisticated Data rich
Experian Credit Bureaux Canada USA Brazil Norway Denmark Netherlands Germany -CEG UK Ireland France Spain Italy South Africa Ghana Nigeria Estonia Russia Romania Bulgaria Turkey - KKB Kuwait - Ci-Net Kenya India Iran ITFIEG Saudi Arabia Simah Saudi Arabia Alijsr Pakistan CreditChex Oman Japan-CCB Australia
Agenda Our experience Credit scoring Credit Bureau Score Development process
Our Expertise Experian have developed Bureau Scores in the following countries Norway UK Denmark Russia The Netherlands Italy Turkey USA South Africa Colombia Different levels of credit data available Data sources unique to that country Bureau Score developments in the pipeline South Korea Japan Saudi Arabia India Bulgaria
Models deployment Credit bureau service Generic or bespoke Scorecard embedded with CB software New Business, Customer Management, Collection Individual user deployment Generic Scorecard deployed at user level Potential integration with SM software
Agenda Our experience Credit scoring Credit Bureau Score Development process
Credit scoring objectives Helps make decisions about new customers, Helps identify profitable customers, Allow the application of relevant terms of business for different segments, to reduce operational costs to reduce risk costs to reduce Capital requirement ( Basel II )
Scoring methodology It s a statistic algorithm that use the available information at a decision point to forecast a future event. Decision point Event measure T - 1 Available information T Event to forecast
Scoring methodology It s a statistic algorithm that use the available information at a decision point to forecast a future event. Decision point Event measure T - 1 T Personal loan application No default (1) BINOMIAL EVENT Default (0)
Decision System: Score Application Scorecard Example Data typologies Bureau Score Occupation Other Database (Central Bank) Limit Balance Age Initial quotations: 850 Delphi holder 0-700 -300 701-950 -50 Over 950 +20 Occupation Self-employed +5 Retired -15 Gov employee +30 Blue Collar +15 Use percentage 1-60 +30 60-100 0 Over 101-60 Balance/ turnover Credit balance +20 0-5 +20 Over 5-20 Minimum age subscribers 18-30 -10 31-40 0 40-60 +10 Over 60 0 Calculation method Score calculation: 850 + 20 15 60 20 + 10 = 785 score distribution Risk category Relation G/B % Population Fino a 500 1:1 3% 501-550 2:1 4% 551-600 5:1 5% 601-650 10:1 8% 651-700 15:1 10% 701-750 30:1 10% 751-800 50:1 15% 801-850 75:1 15% 851-900 100:1 15% 901 + 150:1 15% The score shows the Good/Bad ratio foreseen for each level Example of scorecard
Good bad definition For an application score it is important to include accepts and rejects in the development sample Good/Bad definition uses bureau variables at outcome Goods all accounts up to date at outcome & never more than 2 payments in arrears Bads any write off in the outcome period or any account 3+payments in arrears Indeterminates all other cases For a customer management score the development sample consists of accounts held at a particular point in time Good/Bad definition uses account specific information at outcome
Predictive power can be measured through GINI index GINI index may vary from 0% to 100%. It consists in the level of discrimination, derived from scorecard, Between good and bad population 100% G Ideal Model G Real Model % population % of clients which will potentially go into default % good sacrified Random model G/B Real model B Ideal model B 0% 100% High Risk Population scored by PD Low risk
Advantages High score Low risk AND ACCEPTED BY THE SCORING SYSTEM MANUALLY ACCEPTED SCORE Cut off MANUALLY REJECTED AND REJECTED BY THE SCORING SYSTEM Low score High risk REJECTED ACCEPTED
Scorecard GENERIC: developed on the Experian portfolio data. It s a scorecard system developed in different step to allow the usage apart from the available information BESPOKE: developed on the client historical portfolio data. Retrospective Sampling Observation point 24 months 12 months Now
Bespoke modelling Benchmark Portfolio Generic A/R G/B Client Portfolio By following adjustment (A/R analysis & G/B analysis), Experian will modify the generic scorecard in a specific bespoke scorecard.
Agenda Our experience Credit scoring Credit Bureau Score Development process
Potential Credit Bureau Scores Scorecard Objectives New Business Customer Management Prospect Screening Creditworthiness Indebtedness Insolvency Collections Fraud Detection Id Authentication Cross-sell & Up-sell Company Failure Limited Cos Company Failure Non-Limiteds
Data Assets Required -Examples Shared negative data Shared positive data Public data (e.g. bankruptcies) Previous search data Id confirmation data (e.g. Telephone file) Government data (e.g. employment/income details) PLUS Information supplied as part of a credit search GeoDemographic information
Credit Bureau Score It is a statistical model to predict the Probability of Default (PD) based on the information available on the Credit Bureau at the moment of processing a new application
Reference Data Models Summarise credit reference data into a score Provide simple and predictive view of data Assist automation of the credit decision process Encourage objective decision making Available in the majority of countries where significant data is being shared
Credit Bureau Score deployment In a stand alone mode Score to be considered with application scoring Segmentation parameter to define new business strategy Characteristic of a scorecard
Stand alone mode Low CB Score Cut-Off High
Score integrated within application scoring Low Final Score Same acceptance rate & Bad Rate Decrease Cut-Off High Low CB Score High
To define strategy for New Business Good Bad Odds 1 2 3 4 5 Score 1 2 3 4 5 Reject Max loan amount 1000 Duration max 6 months Max loan amount 1500 Duration max 12 months Max loan amount 2000 Duration max 18 months Max loan amount 3000 Duration max 24 months
Characteristic of a Application Scorecard Address Type Occupation Owned +27 Student -20 Rent -15 Blue Collar +40 With parents 0 Manager +100 Age Credit Bureau Score Up to 20-19 1-3 +100 21-25 -10 4-6 +70 26-40 0 7-8 -50 41-65 +23 9-10 -90
SCORECARD AND DEV DATA Application data Negative CB data Positive CB data Bad rate reduction 16 %
SCORECARD AND DEV DATA Application data Negative CB data Positive CB data Bad rate reduction 16 % 24 %
SCORECARD AND DEV DATA Application data Negative CB data Positive CB data Bad rate reduction 16 % 24 % 41 %
SCORECARD AND DEV DATA Application data Negative CB data Positive CB data Bad rate reduction 16 % 24 % 41 % 35 %
Credit Bureau Users: portfolio quality Contract status CB Score Users Non CB Score Users Current 95.8% 89.0% 1/2 installments in arrears 3+ installments in arrears 2.2% 5.9% 0.5% 2.1% Defaults 1.5% 3.0%
A Generic Bureau Scorecard Development 1. Familiarisation Understand data available, its reliability and any legal constraints
A Generic Bureau Scorecard Development 1. Familiarisation 2. Definition of Bureau Interface Define key bureau characteristics to be included in the generic CB Score
A Generic Bureau Scorecard Development 1. Familiarisation 2. Definition of Bureau Interface 3. Scorecard Definition Simulate expected target population and assign scorecard points accordingly
A Generic Bureau Scorecard Development 1. Familiarisation 2. Definition of Bureau Interface 3. Scorecard Definition 4. Expert Validation Score out a cross-section of consumers and sense check the scores Panel of Experts to include representatives from EDA, the local CB and local (potential) users
A Generic Bureau Scorecard Development 1. Familiarisation 2. Definition of Bureau Interface 3. Scorecard Definition 4. Expert Validation 5. Data Validation Score out a large sample of consumers and split the final score distribution into 10 score deciles. These will then represent each individual s credit rating
Development process Step 1: deliver a generic model Step 2: deliver bespoke model based on performance data and weight socio demographic characteristic according to perfomance information On going: monitor Credit Bureau scorecard to capture changes in the market and ensure local staff support Potential: develop niche scorecards and customer management scoring
Credit Bureau Score and data modelling Experian for CreditChex Karachi Nov. 29 th 2007