CRAFTED.

Lending + AI: How Branch Lends to People Banks Won’t Touch | Matt Flannery (Founder and CEO of Branch; Founder of Kiva)


Listen Later

Matt Flannery and Branch have done something the banks have not: learned how to profitably lend to people who have little to no credit history. Matt is the founder and CEO of Branch, which issues small loans to millions of people in India, Nigeria, Kenya, and Tanzania. 


He's also the founder of Kiva, a microfinance pioneer that skyrocketed from a small project into a worldwide nonprofit that Oprah and President Clinton loved talking about. 


On this episode of CRAFTED., we learn how Branch uses data from people's phones to confidently make loans to people who don't have traditional credit scores. Plus, how they prevent fraud and avoid bias. 


We'll also explore the wild ways that Branch is experimenting with Generative AI, including how they are creating “future synthetic data” that they believe will predict how users will save and spend in the future. 


Takeaways:

  • Branch uses AI to confidently lend to people without traditional credit scores
  • Branch was built on traditional machine learning models – the name “Branch” derives in part from the “random forest” approach – and now is adding Generative AI approaches to the mix
  • Branch is using GenAI to create “future synthetic data” that predicts how people will spend and save in the years to come. As Matt says, “it’s kind of a wild idea” and it’ll take a few years to see how predictive the approach is
  • To avoid bias, lend to lots of people no matter what the data says. It will teach you what the “natural loss rate” is and prevent you from training your model on customers you’ve already selected as creditworthy.
  • Preventing fraud is the biggest challenge. And you can go from zero fraud to massive fraud over a weekend if fraudsters discover a vulnerability.
  • Branch is hugely successful in India, because of the approach it developed in Africa: lend very small amounts to lots of people and, as people repay, offer them larger loans. 
  • Branch failed in Mexico, because the user experience of repaying the loan (visiting a local shop) was too difficult; meanwhile another reason for success in India is the country’s recent rollout of a nationwide mobile payments system (UPI). Plus, willingness to repay in India is naturally very high. 

Key Moments:

  • (02:33) - Founding Kiva and its rise from a side project to a worldwide non-profit
  • (06:31) - Founding Branch and the impact of lending to people that banks won’t
  • (10:03) - Scaling Branch, and why it can grow better as a for-profit than Kiva could as a non-profit
  • (14:28) - Why fraud is the biggest challenge to Branch and how they prevent it
  • (16:17) - How Branch got really good at making quick lending decisions and why it’s critical to approve lots of people
  • (18:41) - Why Branch failed in Mexico, and how those lessons led to their outsize success in India
  • (21:34) - How Branch uses AI to make lending decisions and how it’s experimenting with GenAI to create “future synthetic data”
  • (25:40) - How to prevent bias – and why Branch automatically approves lots of people for loans no matter what the data says about them
  • (29:04) - What’s next for branch
  • (29:57) - How being in a rock band helped Matt gain confidence — and how that served him when he’d later appear on Oprah
  • (30:59) - Outro

  • CRAFTED. is brought to you in partnership with Docker, which helps developers build, share, run and verify applications anywhere – without environment confirmation or management. More than 20 million developers worldwide use Docker's suite of development tools, services and automations to accelerate the delivery of secure applications. 


    CRAFTED. is produced by Modern Product Minds, where CRAFTED. host Dan Blumberg and team can help you take a new product from zero to one... and beyond. We specialize in early stage product discovery, growth, and experimentation. Learn more at modernproductminds.com 

    Subscribe to CRAFTED., follow the show, and sign up for the newsletter 👉 crafted.fm

    ...more
    View all episodesView all episodes
    Download on the App Store

    CRAFTED.By Dan Blumberg

    • 4.1
    • 4.1
    • 4.1
    • 4.1
    • 4.1

    4.1

    47 ratings


    More shows like CRAFTED.

    View all
    This Week in Startups by Jason Calacanis

    This Week in Startups

    1,270 Listeners

    The Changelog: Software Development, Open Source by Changelog Media

    The Changelog: Software Development, Open Source

    283 Listeners

    Freakonomics Radio by Freakonomics Radio + Stitcher

    Freakonomics Radio

    32,202 Listeners

    The Martin Lewis Podcast by BBC Radio 5 Live

    The Martin Lewis Podcast

    54 Listeners

    Fintech One•On•One by Peter Renton

    Fintech One•On•One

    55 Listeners

    a16z Podcast by Andreessen Horowitz

    a16z Podcast

    1,030 Listeners

    The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch by Harry Stebbings

    The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch

    517 Listeners

    Investing in Real Estate with Clayton Morris | Investing for Beginners by Clayton Morris

    Investing in Real Estate with Clayton Morris | Investing for Beginners

    1,064 Listeners

    Y Combinator Startup Podcast by Y Combinator

    Y Combinator Startup Podcast

    216 Listeners

    The Indicator from Planet Money by NPR

    The Indicator from Planet Money

    9,553 Listeners

    MarTech Podcast ™ // Marketing + Technology = Business Growth by I Hear Everything

    MarTech Podcast ™ // Marketing + Technology = Business Growth

    162 Listeners

    Pattern Breakers by Floodgate

    Pattern Breakers

    379 Listeners

    This Is Purdue by Purdue University

    This Is Purdue

    118 Listeners

    Postgres FM by Nikolay Samokhvalov and Michael Christofides

    Postgres FM

    20 Listeners

    The AI Daily Brief (Formerly The AI Breakdown): Artificial Intelligence News and Analysis by Nathaniel Whittemore

    The AI Daily Brief (Formerly The AI Breakdown): Artificial Intelligence News and Analysis

    456 Listeners