In an article about Invoice finance I have read recently, the author asks an interesting question about AI models built with data going from 2018 to 2008. What happens when you build a model based only on data from the longest expansion cycle? How can this data predict potential defaults if no recession data from the previous cycle is included?
The Author emphasizes 3 important aspects to keep in mind when building credit default predictive models:
- Models need to be built with significant credit cycle information in order to incorporate depression and expansion times
- Models that do not include diverse credit lifecycles may work for a period of time, but will fail when comes the next recession
- When it comes to invoice finance, these models need more information than just a buyer and their sellers’ transaction history — that is necessary but not sufficient credit information.
The topic is timely, with all the talks about a looming recession, or the advent of a downturn.
- How will model-based credit scoring financial companies react? The question remains the same for algorithmic trading and all algorithms based financial services companies.
- Any company that built models using financial data that did not include the past 2 recessions, 2008 and 2000 is doomed to see strong surprises in their model ability to predict credit defaults or a decrease in market returns.
That’s why it’s good news to see the regulator now ready to work with companies using algorithmic-based credit scoring.
Anytime the regulator gets interested in a topic it means it’s getting bigger. That’s the case with the CFPB and the update on their non-action letter on Alternative Data.
- Alternative Data is already used by credit bureaus more to allow for a better analysis of creditworthiness. Transunion, Fico, Experian, all have web pages explaining how they use alternative data in building their credit scores.
- Companies are now using social media activity, credit card data, utility invoices to assess creditworthiness.
These companies claim that going past the credit score is allowing them to lend to more customers and have a lower rate of defaults.
In its recent letter, the CFPB tends to agree.
For some consumers, the use of unconventional sources of information, or “alternative data,” to evaluate creditworthiness may be a way to increase access to credit or decrease the cost of credit. Alternative data includes information not typically found in core credit files of nationwide consumer reporting agencies and may indicate a likelihood of meeting obligations on time that traditional credit history may not reflect.
The results of their study are extremely interesting.
- Acceptance rates increased
- APRs decreased
- Approval for Low Income (below $50K) increased
- Approval for younger folks increased
- Approval for “Fair” scores increased
Here is how the CFPB presented its results.
- The tested model approves 27% more applicants than the traditional model, and yields 16% lower average APRs for approved loans.
- The tested model increased acceptance rates by 23–29% and decreasing average APRs by 15–17%.
- The tested model significantly expands access to credit compared to the traditional model.
- “Near prime” consumers with FICO scores from 620 to 660 are approved approximately twice as frequently.
- Applicants under 25 years of age are 32% more likely to be approved.
- Consumers with incomes under $50,000 are 13% more likely to be approved.
In other news in Financial technology
Jack Dorsey: Mass crypto adoption will transform Square’sbusiness (CoinDesk)
Apple reveals that it is finally…interested in Crypto (Forbes)
A bitcoin holder transferred $1bn of the cryptocurrency in a single transaction. As crypto-legend Nick Szabo points out, this degree of ‘confidence in Bitcoin is splendid but tempts fate.’
European banking app Monese partners with deposits marketplace Raisin (TechCrunch)
Chime now has 5 million customers (TechCrunch)
No one-trick pony: How Ally got to $100B in deposits (Bank Innovation)
Research: Challenger banks on track to treble customers next year but profits remain elusive (Fintextra)
Will online banks see slower growth as rates fall? (Bloomberg)
PNC Bank launches in-house fintech startup numo (deBanked)
Credit Unions go on bank buying spree (WSJ)
‘Have a great day!’ — robohelp leaves banking fraud victims cold(Reuters)
Is Uber steering more and more towards fintech? (AltFi)
Franklin Templeton to track money fund shares on Stellar blockchain (CoinDesk)
Stripe Capital will lend $10,000 to $20,000 to Stripe clients (Tearsheet)
PrimaryBid, a UK-based new share platform for investors, raised £7 million co-led by Pentech and Outward VC. http://axios.link/AfLb
IceKredit, credit risk and credit management platform with offices in China and Los Angeles, raised $47 million in new funding co-led by Guohe Capital and Yunqi Partners. http://axios.link/m3MT
Spendesk, a three-year-old, Paris, France-based spend management platform for small to mid-sized businesses, has raised €35 million in Series B funding led by Index Ventures. The round brings the total the company has raised to €45 million. The outlet UKTN has more here.
Root Insurance, a 4.5-year-old, Columbus, Oh.-based car insurance startup that uses smartphone technology to understand individual driver behavior, has officially raised $350 million in funding at a valuation of $3.65 billion following a report last month in Axios that this deal was afoot. DST Global and Coatue Management co-led the round, joined by Drive Capital, Redpoint Ventures, Ribbit Capital, Scale Venture Partners, and Tiger Global Management. More here.
AppZen, a seven-year-old, San Jose, Ca.-based maker of AI tools for corporate expensing and finance teams, has raised $50 million in Series C funding. Coatue Management led the round, joined by earlier investors Redpoint Ventures and Lightspeed Venture Partners. TechCrunch has more here.
Happy Money secures $70M to help people break up with their credit cards (Crowdfund Insider)
Credit Sesame raises $43M in advance of IPO (Finovate)
European open banking platform Railsbank raises $10M to expand to the US (TechCrunch)
Sage Group (LSE: SGE) said it is considering a sale process for its payment processing unit. http://axios.link/LQNN
Railsbank, the open banking and compliance platform co-founded by CEO Nigel Verdon, who previously founded Currencycloud, has raised $10 million in Series A funding. (Techcrunch)
Source: Tearsheet, Strictly VC, Termsheet
This story is part of my series on How Artificial Intelligence transforms the world. The previous articles are
– How Artificial Intelligence is transforming Asset Management.
– How Artificial Intelligence is radically transforming your TV experience