Evolving FICO® Scores from GFS to Mature Market Scores John Hadlow Senior Director FICO © 2014 Fair Isaac Corporation. Confidential. This presentation is provided for the recipient only and cannot be reproduced or shared without Fair Isaac Corporation’s express consent. Agenda ► Evolving Scores—Global Landscape ► FICO® Scores Global Deployment Map ► GFS Approach and Key Benefits ► Empirical Scores and Key Benefits ► Russia and South Africa—Examples ► Traditional Scores-Takeaways ► Improvements from Non-Traditional Tools ► Value from Non-Traditional Data ► Key Takeaways ► FICO® Score for International Markets in Romania ► Implementing the FICO® Score in México 2 © 2014 Fair Isaac Corporation. Confidential. Evolving Scores—Global Landscape ► Positive Credit Data is key to production of good quality credit bureau scores ► For now, Positive Credit Data is now mostly accepted as the most predictive type of information that is acceptable to regulators, lenders and consumers for supporting credit decisions around the world ► Scores ► There & FICO® Scores are now available from a majority of credit bureaus is other ‘big’ data that is predictive, powerful but there are challenges: ► Hard to operationalize ► Can be unpalatable to regulators and/or consumers ► Some data adds insufficient value in current methodologies ► Clearly, models are very important but good, complete data is paramount supplies Global FICO® Scores (GFS), empirical FICO® Scores and has new ‘Big Data’ based offerings ► FICO 3 © 2014 Fair Isaac Corporation. Confidential. FICO® Scores Deployed Globally—2014 GREENLAND ALASKA (USA) SWEDEN ICELAND RUSSIAN FEDERATION FINLAND NORWAY CANADA ESTONIA LATVIA DENMARK LITHUANIA REPULIC OF IRELAND UNITED KINGDOM BELARUS NETHERLANDS GERMANY POLAND BELGIUM CZECH REPUBLIC UKRAINE SLOVAKIA KAZAKHSTAN AUSTRIA MONGOLIA HUNGARY FRANCE SWITZ. ROMANIA ITALY UZBEKISTAN BULGARIA PORTUGAL GEORGIA KYRGYZSTAN SPAIN NORTH KOREA GREECE TURKEY UNITED STATES of AMERICA TURKMENISTAN SYRIA IRAN IRAQ TUNISIA MOROCCO SOUTH KOREA CHINA JAPAN PAKISTAN ALGERIA FICO Score Deployment Legend TAHKISTAN AFGHANISTAN NEPAL LIBYA WESTERN SAHARA EGYPT MEXICO SAUDI ARABIA TAIWAN UAE OMAN INDIA VIETNAM CUBA MYANMAR MAURITANIA LAOS MALI FICO® Score Deployed NIGER GUATEMALAHONDURAS CHAD SUDAN SENEGAL THAILAND YEMEN NICARAGUA PHILIPPINES CAMBODIA BURKINA GUINEA NIGERIA COSTA RICA PANAMA FICO® Score Available at Credit Bureau VENEZUELA LIBERIA GUYANA FRENCH SURINAME GUIANA COLOMBIA ETHIOPIA GHANA COTE D’IVOIRE SRI LANKA CENTRAL AFRICAN REPUBLIC CAMEROON MALAYSIA SOMALIA UGANDA KENYA GABONCONGO ECUADOR DEMOCRATIC REPUBLIC OF CONGO TANZANIA Credit Information Legend * PAPUA NEW GUINEA INDONESIA BRAZIL PERU ANGOLA ZAMBIA BOLIVIA MADAGASCAR ZIMBABWE NAMIBIA PARAGUAY BOTSWANA AUSTRALIA Full/Partial +ve Data Sharing URUGUAY REPUBLIC OF SOUTH AFRICA CHILE Negative-only Data Sharing ARGENTINA NEW ZEALAND No Data Sharing *Source: World Bank. 2013. Doing Business 2014: Understanding Regulations for Small and Medium-Size Enterprises. Washington, DC: World Bank Group. DOI: 10.1596/978-0-8213-9984-2. License: Creative Commons Attribution CC BY 3.0 © 2014 Fair Isaac Corporation. Confidential. Why Global FICO® Score? ► Business Requirements: Demand ► ► ► ► ► ► ► Competitive Stresses and challenges ► ► ► ► ► ‘Everyone’ can build a score these days! FICO is a partner to bureaus and partners like a share of the revenue Short term performance vs long term tolerance and stability Thin files and unique populations and data Unique FICO Advantages in meeting the Requirements and Challenges ► ► ► 5 (Not really for mature markets except for new entrants) Highly predictive credit bureau score with the same features and benefits as FICO ® Score Able to take whatever data is available/usable and score it like a FICO® Score Make it sufficiently transparent and well documented to meet all likely compliance/regs Build in enough unique IP and features to make it valuable and different ….and ideally make it part of an international standard for the long term FICO IP and diligence is still giving extra performance The FICO International Data Consortium and R&D data pools FICO tools (Model Builder) processes (Best Practice, Compliance and delivery mechanisms (Software) © 2014 Fair Isaac Corporation. Confidential. What Does GFS Bring in a New Market or Entrant? pre-built predictive performance ‘out of the box’* ► Data to validate performance of the score is needed but not to build it ► 300–850 plus reason codes ► Data tolerance over time and stability ► Best practice documentation and compliance for most markets ► A platform for further localization and to gain even more performance, segments ► Ease of implementation due to delivery in software ► Excellent Most importantly—Extremely strong ROI through low investment cost which utilizes and leverages high investment and data consortium from FICO over several years *Requires mapping of bureau interface to GFS interface 6 © 2014 Fair Isaac Corporation. Confidential. So What Does a Full Empirical Score Bring? ► An obvious question then follows—so why bother with a full empirical FICO® Score? ► Whilst the GFS framework brings clear benefits, it ultimately restricts the data input utilized and characteristics used in the models—and hence ultimate performance ► A full ground up development will give the best possible performance using all the data available but obviously the investment costs are much higher ► Potentially, complete transparency and ultimate separation of goods from bads with the potential trade off of shorter term durability and stability ► Therefore it is more a matter of ‘horses for courses’—Both options are best at what they are designed to do; ► GFS for new, developing markets, or data constrained markets ► FICO® Score empirical with more stability for larger more mature environments 7 © 2014 Fair Isaac Corporation. Confidential. Russia Score Enhancement—More Data, Better Scores NBKI Russia FICO® Score: Account Management Performance NBKI v1 NBKI v2 NBKI v3 100 90 Cumulative % Bads 80 70 60 50 40 30 20 10 0 0 8 © 2014 Fair Isaac Corporation. Confidential. 20 40 60 Cumulative % Total 80 100 South Africa Score Enhancement—Changing Data, Scores TU South Africa—Empirica Score: Account Management Performance Empirica v1 Empirica v2 Empirica v3 Empirica v4 100% Cumulative % Bads 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 0% 9 © 2014 Fair Isaac Corporation. Confidential. 20% 40% 60% Cumulative % Total 80% 100% Traditional Scores—Takeaways ► There is much to be gained by evolution of Scores through improved tools, data and models ► GFS is a great starting point for some markets and offers great performance and ROI ► FICO® Scores built empirically add even more value in developed markets with good ROI ► It is a matter of matching the approach to the market or client requirements ► Data changes often mean trade offs between ultimate performance and stability ► …but 10 what of Other Data and scores? © 2014 Fair Isaac Corporation. Confidential. Example Improvement from Non-Traditional Tool 11 © 2014 Fair Isaac Corporation. Confidential. Value from Non-Traditional text based Data 12 © 2014 Fair Isaac Corporation. Confidential. Key Takeaways ► There’s still a way to go and great value in using ‘traditional’ credit bureau data in FICO® Scores and other scores in many markets ahead of going for the ‘Big Data’ ► But...there is significant additional value in Big Data if you can operationalize ► Embrace machine learning, but challenge is to generate fully comprehensible models ► Data-driven models typically need to be refined by domain experts before they can be deployed ► New approaches allow you to combine the power of machine learning with the benefits of domain knowledge ► There is power in new data sources, such as unstructured text ► Powerful analytic approaches such as topic modeling and semantic scorecards allow you to comprehend the value and meaning of text data for predicting consumer behavior ► Either 13 way FICO is able to lead and support partners through these changes © 2014 Fair Isaac Corporation. Confidential. FICO® Score for International Markets In Romania (formerly Global FICO® Score) The “Win-Win” Solution During the Crisis in Romania Cristian Racu IT Manager Romanian Credit Bureau © 2014 Fair Isaac Corporation. Confidential. This presentation is provided for the recipient only and cannot be reproduced or shared without Fair Isaac Corporation’s express consent. FICO® Score for International Markets is a key element in supporting Romanian lenders during the economic crisis and an important source of revenue for the Romanian Credit Bureau 15 © 2014 Fair Isaac Corporation. Confidential. Agenda ►Market ►The Elements Romanian Credit Bureau ►FICO® ►Initial Score from Biroul de Credit Observations ►Actions ►Current 16 Situation © 2014 Fair Isaac Corporation. Confidential. Market Elements Romania ► Population: ► GDP: 21.6 million $188 billion ► Number of banks: 40 ► Total household savings: $38,376 million ► Total bank placements to retail: $29,415 million 17 © 2014 Fair Isaac Corporation. Confidential. The Romanian Credit Bureau ► Established in 2004 ► Shareholders: ► Info 23 banks type: negative and positive, on individuals and sole traders ► Market Coverage: 99.8% retail banking sector + major non-banking financial lenders ► Data ► Hit up-dating: daily ratio (2014): 86.48% ► Response 18 time: less than 5 seconds © 2014 Fair Isaac Corporation. Confidential. The Romanian Credit Bureau Total Credit Reports 9 8 7 Millions 6 5 4 3 2 1 0 2004 19 2005 © 2014 Fair Isaac Corporation. Confidential. 2006 2007 2008 2009 2010 2011 2012 2013 2014 (8M) FICO® Score from Biroul de Credit ► Scoring bid main request world class provider FICO ► Launched ► Sales in 2009 (at the start of the crisis) Evolution ► Usage Evolution ► Stability 20 © 2014 Fair Isaac Corporation. Confidential. FICO® Score from Biroul de Credit Sales Evolution FICO® Score from Biroul de Credit Inquiries Number of Participants 25 Portfolio monitoring Other inquiries 4,000,000 20 3,500,000 3,000,000 15 2,500,000 2,000,000 10 1,500,000 1,000,000 5 500,000 0 0 2009 21 2010 2011 © 2014 Fair Isaac Corporation. Confidential. 2012 2013 2014 (8M) 2009 2010 2011 2012 2013 2014 (8M) FICO® Score from Biroul de Credit Usage Evolution ► 2009-2012: Collection ► Fast tool to help collection ► Key element for selling the bad accounts to the external collectors ► Stability element for the de-calibrated socio-demographic scores ► 2013: Account Management ► Portfolio scoring for major lenders ► Early warning system using the FICO® Score ► 2014: Re-start lending ► Big lenders build their marketing strategies on the portfolio analysis ► Market target groups using the FICO® Score 22 © 2014 Fair Isaac Corporation. Confidential. FICO® Score from Biroul de Credit An Effective Score in a Changing Environment ► Despite ► The bad rate of the lowest scoring deciles is consistently 100x that of the lower scoring deciles Overall Population Bad Rate Aug-11 9.5% 23 Sep-12 13.7% Aug-13 18.2% Feb-14 16.4% © 2014 Fair Isaac Corporation. Confidential. Bad Rate by Decile Over Time Aug-11 100% Sep-12 Aug-13 Feb-14 90% 80% Interval Bad Rate the changing economic landscape the FICO® Score from Biroul de Credit nicely rank orders the consumer risk 70% 60% 50% 40% 30% 20% 10% 0% 1 2 3 4 5 FICO® 6 7 Score Decile 8 9 10 Initial Observations ► No scoring alternative—big confidence in socio-demographic scores (before the crisis) and their fast de-calibration (within the crisis) ► Non-homogenous ► The market know-how on bureau scoring increasing “cost issue” due to the crisis ► Analytic services provided by external consultants with no knowledge about the local market 24 © 2014 Fair Isaac Corporation. Confidential. Actions ► Initial marketing campaign (FICO training seminars, face-to-face meetings with the lenders) ► Retro-fit analysis discount offer (first year) ► Volume based discount offer for account management (last 4 years) ► Dedicated software for retro-fit analysis ► Dedicated sales team ► Lenders 25 “education” using analytic team © 2014 Fair Isaac Corporation. Confidential. Current Situation ► A (more) homogenous market know-how ► Constant increase of FICO® Score from Biroul de Credit’ usage ► Usage extension for new purposes (provisioning, Basel III) ► Risk-based 26 pricing marketing campaigns using FICO® Score from Biroul de Credit © 2014 Fair Isaac Corporation. Confidential. Implementing the FICO® Score in México Juan Manuel Ruiz Palmieri Commercial VP Círculo de Crédito © 2014 Fair Isaac Corporation. Confidential. This presentation is provided for the recipient only and cannot be reproduced or shared without Fair Isaac Corporation’s express consent. Agenda ►México ►About ►Our Credit Market Overview Círculo de Crédito Challenges ►New data (Microlending) ►FICO® Score Implementation ►Metrics and Results ►Benefits 28 to Market © 2014 Fair Isaac Corporation. Confidential. México Credit Market Overview ► Consumer ► 50% credit culture still a challenge of population on informal jobs ► Traditional banks and retail companies ► Targeting medium to high incomes ► Focus on credit cards ► Savvy in the use of sophisticated risk tools ► Microfinancial ► Service 29 institutions evolved from ONG’s to Regulated Institutions and Utilities companies starting to use predictive models © 2014 Fair Isaac Corporation. Confidential. México Credit Market Overview Monthly income by family in USD Banks A 0.79% B 1.87% $ 12,000 – $ 100,000 C + 3.39% Cm 5.92% $ 1,450 – $ 11,999 C - 8.13% D + 11.97% Microlending Entities Dm 14.84% $ 400 – $ 1,449 D - 22.06% E 31.03% $ 60 – $ 399 Source: Sigmarket 2010 30 © 2014 Fair Isaac Corporation. Confidential. About Círculo de Crédito ► We attend Banks, Retail, Microlending, Services and Utilities industries ► We bring new data to the market (Microlending) ► 10 years operating in México ► More than 3,000 customers ► Positive ► We and negative information are regulated and supervised by ► Secretaria de Hacienda y Credito Publico (IRS) ► National Banking and Security Commission (SEC) ► Central Bank (FED) 31 © 2014 Fair Isaac Corporation. Confidential. Numbers ► 40% of market share and growing ► Consumer ► 57 database: Millions, reporting 300 millions of credit accounts ► Corporate ► 800K, database: reporting 2.5 million of credit accounts ► Inquiries ► 80 million transactions per year *Cifras en miles 32 © 2014 Fair Isaac Corporation. Confidential. Products *Cifras en miles 33 © 2014 Fair Isaac Corporation. Confidential. Our Challenges ► 2008, first origination score developed by one of the top 3 American Credit Bureaus, did not perform as expected ► FICO´s performance in the traditional banking sector is a sure bet, the challenge was to perform in all sectors ► The goal was: ► Prove the effectiveness in microlending and provide the first Score for this market ► Outperform other scores in all segments 34 © 2014 Fair Isaac Corporation. Confidential. New Kid in Town Microlending ► Growing for the last 10 years ► Different legal frame ► Banks, Regulated Entities (Saving, IPO’s) and Non regulated ► High employee turnover ► High cost of operation ► Increased ► More ► New ► Big 35 competition than 400 entities in risk management market to serve © 2014 Fair Isaac Corporation. Confidential. Microlending Methodology ► Individual Credit ► Mostly for self-employment ► Weekly payments (16 weeks) ► Credit amount (Minimum $115 USD, maximum $12,000 USD and average $1,300 USD) ► Group Credit ► Between 15 and 20 people ► Each of them ask for a specific amount ► All in group are joint guarantees ► They meet weekly to make the payment 36 © 2014 Fair Isaac Corporation. Confidential. FICO® Score Implementation ► Sample ► Model ► First ► IT size 2M files adjusting and customization validation: 6 months Statistic Value K-S 58.24 K-S Score 556 ROC Area 0.817 implementation 2 months ► Results: ► Strong rank ordering of risk demonstrated on the development population ► Existing data elements showed strong predictive capability ► Better results in each and every subsequent yearly validations (3 years now) 37 © 2014 Fair Isaac Corporation. Confidential. Metrics and Results 16.0% 40.0% 14.0% 12.0% % Population 30.0% 10.0% 8.0% 20.0% 6.0% 4.0% 10.0% 2.0% 0.0% 0.0% 300-479 500-519 540-559 580-599 % Distribution Bad = 90+dpd 12 months after first purchase. 38 © 2014 Fair Isaac Corporation. Confidential. 620-639 660-679 % Bad 700-719 740-759 780-850 Benefits to Market ► Reduce time in granting credit ► Automate origination process ► Have a standard risk metric ► More homogeneous analysis of credit report ► No training required ► One 39 score for all the market © 2014 Fair Isaac Corporation. Confidential. Thank You! Juan Manuel Ruiz Palmieri Commercial VP Círculo de Crédito Cristian Racu IT Manager Romanian Credit Bureau © 2014 Fair Isaac Corporation. Confidential. This presentation is provided for the recipient only and cannot be reproduced or shared without Fair Isaac Corporation’s express consent. John Hadlow Senior Director FICO Please rate this session online! Juan Manuel Ruiz Palmieri Commercial VP Círculo de Crédito 41 © 2014 Fair Isaac Corporation. Confidential. Cristian Racu IT Manager Romanian Credit Bureau John Hadlow Senior Director FICO