There are plenty of valid criticisms to be made of BNPL, but I’m pretty sure the scenario you described above with the DL tacos is not accurate. I have not used BNPL myself, but I believe they actually do ask for a lot of identifying information, run credit checks and cut off credit to delinquent borrowers. This doesn’t exonerate their business model, which certainly could fail, but I don’t think you can credibly accuse the Klarnas/Affirms/Afterpays of the world of not attempting to develop extensive credit/risk systems.
I looked into this a bit and wasn't able to find much indicating that they had good risk systems / people couldn't just default regularly. I did however find a lot indicating that they weren't using existing credit risk systems like credit scores...Do you have links?
Sure -- I'm definitely not an expert but their F-1 (https://www.sec.gov/Archives/edgar/data/2003292/000200329225000024/klarnagroupplcf-1a3.htm) has a bunch of information about their underwriting/risk. Search for the "Consumer Credit Underwriting" section, which talks about how they assess consumers using "transactional data, consumer data (internal purchase and payment history) and credit bureau data".
The way I've heard BNPL pitched is that because they can do per-transaction underwriting (as opposed to per-individual, which is basically how credit cards work), they're actually able to be much more sophisticated. On top of that, these companies are relatively nimble and tech-saavy, at least in comparison to Visa/MC.
Whether you actually believe this narrative, of course, is another story. I personally find it somewhat credible. Although the past 20 years are full of startups claiming to disrupt consumer credit through AI and ML-powered underwriting and not really doing anything of the kind.
My favorite example of this was from a coworker who previously worked at one of these companies. Despite claiming to have sophisticated ML systems evaluating risk, the biggest signal they got was from a slider they put on their homepage asking people how much money they wanted to borrow. If someone landed on the homepage and immediately maxed the slider out to the right, they learned, don't lend to them. Innovation!
Ok read through in more depth. I think the F1 has a fair bit of fluff (which makes sense for an F1) so its quite difficult to figure out how much they are able to price risk differently from other creditors. I think they certainly have this hypothesis of some automated per-transaction dynamic risk model, but they also admit in their F1 that they are looking at people who have a 'range of credit histories' that may not be serviced by the existing industry. So I more or less stand by my original criticism, that they end up in a part of the market that other creditors don't service because said creditors do not trust the potential debtors. (That may not be a ding against their ability to build a good business. Maybe they are pricing the risk effectively.) Maybe the more extreme version of the claim -- that it is easy for a random person to get a free taco -- is incorrect...though I do think basically everyone in the country could get a free taco at least one time.
One other line that I thought was funny was that they "accelerated the placement of overdue accounts with debt collection agencies." which I took to mean 'they werent even sending overdue accounts to collections'
I think this is also all still basically just credit? I can't figure out the difference between BNPL and a standard credit company except that they were dodging regulations
Yep, makes sense. My high level take on these types of businesses (again, not knowing very much about the specifics of BNPL) is that finding risk that’s mispriced by traditional credit agencies != finding risk that you should build a business around. I.e. you’d probably prefer to have a portfolio of mispriced, uncorrelated risk than correctly priced, correlated risk.
I work as a Taco repo specialist and you would be amazed at the techniques we can use to recover the loss…
Deeply concerned 👀
There are plenty of valid criticisms to be made of BNPL, but I’m pretty sure the scenario you described above with the DL tacos is not accurate. I have not used BNPL myself, but I believe they actually do ask for a lot of identifying information, run credit checks and cut off credit to delinquent borrowers. This doesn’t exonerate their business model, which certainly could fail, but I don’t think you can credibly accuse the Klarnas/Affirms/Afterpays of the world of not attempting to develop extensive credit/risk systems.
I looked into this a bit and wasn't able to find much indicating that they had good risk systems / people couldn't just default regularly. I did however find a lot indicating that they weren't using existing credit risk systems like credit scores...Do you have links?
Sure -- I'm definitely not an expert but their F-1 (https://www.sec.gov/Archives/edgar/data/2003292/000200329225000024/klarnagroupplcf-1a3.htm) has a bunch of information about their underwriting/risk. Search for the "Consumer Credit Underwriting" section, which talks about how they assess consumers using "transactional data, consumer data (internal purchase and payment history) and credit bureau data".
The way I've heard BNPL pitched is that because they can do per-transaction underwriting (as opposed to per-individual, which is basically how credit cards work), they're actually able to be much more sophisticated. On top of that, these companies are relatively nimble and tech-saavy, at least in comparison to Visa/MC.
Whether you actually believe this narrative, of course, is another story. I personally find it somewhat credible. Although the past 20 years are full of startups claiming to disrupt consumer credit through AI and ML-powered underwriting and not really doing anything of the kind.
My favorite example of this was from a coworker who previously worked at one of these companies. Despite claiming to have sophisticated ML systems evaluating risk, the biggest signal they got was from a slider they put on their homepage asking people how much money they wanted to borrow. If someone landed on the homepage and immediately maxed the slider out to the right, they learned, don't lend to them. Innovation!
Amazing, thanks for the receipts. I'll update the article to account for this (though I do want to read through the thing after/at some point)
Ok read through in more depth. I think the F1 has a fair bit of fluff (which makes sense for an F1) so its quite difficult to figure out how much they are able to price risk differently from other creditors. I think they certainly have this hypothesis of some automated per-transaction dynamic risk model, but they also admit in their F1 that they are looking at people who have a 'range of credit histories' that may not be serviced by the existing industry. So I more or less stand by my original criticism, that they end up in a part of the market that other creditors don't service because said creditors do not trust the potential debtors. (That may not be a ding against their ability to build a good business. Maybe they are pricing the risk effectively.) Maybe the more extreme version of the claim -- that it is easy for a random person to get a free taco -- is incorrect...though I do think basically everyone in the country could get a free taco at least one time.
One other line that I thought was funny was that they "accelerated the placement of overdue accounts with debt collection agencies." which I took to mean 'they werent even sending overdue accounts to collections'
I think this is also all still basically just credit? I can't figure out the difference between BNPL and a standard credit company except that they were dodging regulations
Yep, makes sense. My high level take on these types of businesses (again, not knowing very much about the specifics of BNPL) is that finding risk that’s mispriced by traditional credit agencies != finding risk that you should build a business around. I.e. you’d probably prefer to have a portfolio of mispriced, uncorrelated risk than correctly priced, correlated risk.