Alternative Data in Decisioning Models
1 like311 views
1. The document discusses how small businesses can have difficulties obtaining loans from traditional lenders due to having little credit history. It suggests that using alternative data sources in lending models could help assess risk for these businesses. 2. Two new alternative data approaches are described: alternative lending models that use alternative lending data to build predictive models, and social media scores that analyze social media data about businesses. 3. The models showed improved approval rates and reduced bad rates compared to traditional credit scoring. Around half of businesses analyzed had their risk assessment significantly impacted by social media data.
Alternative Data in Decisioning Models
- 1. 1 © Experian 27/06/2019 Sip and Solve Alternative Data in Decisioning Models Featuring: Emily Garrett Analytics Consultant Experian Business Information
- 2. 2 © Experian “Growing firms need resources, but many small firms may have a hard time obtaining loans because they are young and have little credit history. Lenders may also be reluctant to lend to small firms with innovative products because it might be difficult to collect enough reliable information to correctly estimate the risk for such products. If it’s true that the lending process leaves worthy projects unfunded, some suggest that it would be good to fix this “market failure” with government programs aimed at improving small businesses’ access to credit.” Small businesses can be at a disadvantage when trying to access capital and credit 27/06/2019 Experian Public Sip and Solve: Alternative data in decisioning models Source: Congressional Research Service https://fas.org/sgp/crs/misc/R40985.pdf
- 3. 3 © Experian “Growing firms need resources, but many small firms may have a hard time obtaining loans because they are young and have little credit history. Lenders may also be reluctant to lend to small firms with innovative products because it might be difficult to collect enough reliable information to correctly estimate the risk for such products. If it’s true that the lending process leaves worthy projects unfunded, some suggest that it would be good to fix this “market failure” with government programs aimed at improving small businesses’ access to credit.” Small businesses can be at a disadvantage when trying to access capital and credit 27/06/2019 Experian Public Sip and Solve: Alternative data in decisioning models Source: Congressional Research Service https://fas.org/sgp/crs/misc/R40985.pdf
- 4. 4 © Experian 1. Inability to expand – many businesses rely on people to be successful, especially in our heavily service-based economy 2. Inadequate capital investment in the company 3. Business failure – some new businesses cannot generate enough early stage working capital to build to self-sufficiency What are the implications of inadequate capital and credit for small businesses? 27/06/2019 Experian Public Sip and Solve: Alternative data in decisioning models
- 5. 5 © Experian 1. Alternative Lending Model 2. Social Media Score Two new approaches 27/06/2019 Experian Public Sip and Solve: Alternative data in decisioning models
- 6. 6 © Experian 1. Alternative Lending Model 2. Social Media Score Two new approaches 27/06/2019 Experian Public Sip and Solve: Alternative data in decisioning models
- 7. 7 © Experian New and well-established businesses need working capital to thrive… but many traditional banks decline most loan applications Growth of alternative lending Funding rate in 2018 * Bix2credit.com/small-business-lending-index 2X! 27% 40% 50% 57% Big Banks Credit Unions Small Banks Alternative Lenders Small business loan open rate
- 8. 8 © Experian Business owner risk Consumer score distribution 27/06/2019 Experian Public Sip and Solve: Alternative data in decisioning models 6% 15% 54% 25% Subprime Near Prime Prime Super Prime Alternative Traditional Subprime + Near Prime 21% Traditional Lending Prime + Super Prime 79% Traditional Lending
- 9. 9 © Experian 27/06/2019 Experian Public Sip and Solve: Alternative data in decisioning models 18% 25% 6% 15% 54% 25% Subprime Near Prime Prime Super Prime Alternative Traditional Subprime + Near Prime 21% Traditional Lending 43% Alternative Lending Prime + Super Prime 79% Traditional Lending Business owner risk Consumer score distribution
- 10. 10 © Experian 27/06/2019 Experian Public Sip and Solve: Alternative data in decisioning models 18% 25% 39% 18% 6% 15% 54% 25% Subprime Near Prime Prime Super Prime Alternative Traditional Subprime + Near Prime 21% Traditional Lending 43% Alternative Lending Prime + Super Prime 79% Traditional Lending 57% Alternative Lending Business owner risk Consumer score distribution
- 11. 11 © Experian Using alternative lending data to build new models that predict lending performance yields positive results! 27/06/2019 Experian Public Sip and Solve: Alternative data in decisioning models 18 67 22 36 47 21 43 23 38 50 30 28 44 57 KS KS % increase Bad cap 10% Bad cap 20% Bad cap 30% Consumer Blended Alternative lender
- 12. 12 © Experian 0% 1% 2% 3% 4% 5% 6% 7% 0% 20% 40% 60% 80% 100% Consumer Blended Custom Alternative lending model performance Comparative performance gain Portfoliobadrate Approval Rate At 50% approval rate Bad rate by score: - Consumer: 4.3% - Blended: 4% - Custom: 3%
- 13. 13 © Experian 0% 1% 2% 3% 4% 5% 6% 7% 0% 20% 40% 60% 80% 100% Consumer Blended Custom Alternative lending model performance Comparative performance gain Portfoliobadrate Approval Rate At 4% bad rate Approval rate by score: - Consumer: 22% - Blended: 49% - Custom: 75%
- 14. 14 © Experian 1. Alternative Lending Model 2. Social Media Score Two new approaches 27/06/2019 Experian Public Sip and Solve: Alternative data in decisioning models
- 15. 15 © Experian Social Media development for commercial risk Research findings demonstrated impact of adding social media data Business Information Services credit-only With social media The Monkey Paw Proper Gastropub Total of 22 San Diego gastropubs analyzed: • 5 dramatically better with social media view • 10 confirmed / little difference • 4 dramatically worse with social media view • 3 no hits / firmographic-only with social media confirming business activity Social media data could have made a difference in approximately half of these businesses Note — Proper Gastropub later closed
- 16. 16 © Experian Attributes with major impact on predictive models • Risk associated to the business category in the business trade area • Risk based on business category • Average rating on social media over various timeframes • Number days since first review within last year • Risk based on price level and hours of operation Social media descriptive attributes • Detailed description of business category, more in depth then SIC • Hours of operations, licensing, other Available via batch or API Social media attributes for risk assessment Key social media attributes utilized in model 27/06/2019
- 17. 17 © Experian • Logically businesses that generate social media activity have higher “coverage” or availability of social media data. Higher coverage lines of business include: – Restaurants, stores, entertainment with the highest coverage rates. – Health and professional services (doctors, accountants, plumbers, etc.) with high coverage. • Increasing each year, by noticeable percentage, with growth in social media. Social media coverage rates Significantly higher availability of data on consumer facing businesses 57.38% 59.40% 62.74% 57.14% 68.00% 83.90% 0% 50% 100% merchant processor 1 bankcard 2 merchant processor 2 insurer1 insurer2 satellite tv1 Restaurants 37.17% 43.70% 45.48% 52.86% 58.70% 55.07% 0% 50% 100% merchant processor 1 bankcard 2 merchant processor 2 insurer1 insurer2 satellite tv1 Beauty Shops Average coverage rate is 20%+ with range from 10% to 62%
- 18. 18 © Experian Social media insight score- fully developed 37% boost overall / 177% gain for new/emerging businesses * Baseline is IPV2 model 23% 19% 65% 177% 0% 20% 40% 60% 80% 100% 120% 140% 160% 180% 200% LiftOverBaseline* Segments Performance Boost by Intelliscore v2 Model Segment Thick Derog Thin Emerging
- 19. 19 © Experian 27/06/2019 Thank you! Have questions or want to learn more about alternative data models? • Chat with us: bit.ly/sip-solve-chat Upcoming Sip and Solves: • Join the list to stay in the loop: bit.ly/sip-and-solve
- 20. 20 © Experian 27/06/2019 Private and confidential Presentation Title https://bit.ly/2X7upTi Register Now!