Why You Can't Get a One-Click Mortgage Refi
44:21 Share

Wells Fargo seeks broad impact in their communities. They're focused on building a sustainable, inclusive future for all by supporting housing affordability, small business growth, financial health, and other community needs. That's why they've donated nearly $2 billion to strengthen local communities over the last five years. Wells Fargo. The Bank of Doing.

See how at wellsfargo.com slash say do. Wells Fargo's philanthropic support includes contributions from Wells Fargo and Company, Wells Fargo Bank N.A., and the Wells Fargo Foundation. I'm Rachel Fultman, and I host a podcast from Popular Science called The Weirdest Thing I Learned This Week. Every other week,

I circle up with guests like Bill Nye, Josh Gondelman, Mary Roach, and many more to prove that the lofty and noble pursuit of science can also be profoundly weird. From flying Ford Pintos to the world's most illegal cheese, The Weirdest Thing I Learned This Week is the ultimate source for all things interesting, informative, and, most importantly, frickin' weird. Check out The Weirdest Thing I Learned This Week wherever you get your podcasts. Come on over whenever you're ready to get weird.

Bloomberg Audio Studios. Podcasts. Radio. News. Hello and welcome to another episode of the All Thoughts Podcast. I'm Traci Alloway. And I'm Joe Weisenthal. Joe, do you remember a month or so ago we recorded an episode all about why mortgage rates were going up even though the Fed has cut?

Yes, this was a big one at the time, and I think it still is really important. Basically, there's this intuition that people have that the Fed affects policy by cutting rates. And one thing that happens when rate cuts is that borrowing costs go down. And one form of borrowing that's very popular is mortgage rates.

But we're in the middle of a rate cut cycle. The Fed cut 50 in September, then kind of another 25 at its subsequent meeting. But mortgage rates have generally not moved down at all and, in fact, moved up after that 50 basis point cut.

By the way, we're recording this November 25th. We haven't seen much improvement at all. And so there is this, I don't know if it's really a mystery, but there is certainly a story about the Fed in the middle of a rate cut cycle and yet it not really feeding through to a lot of kinds of borrowing. I don't think it's a mystery. We did a whole episode.

No, you're right. You're right. You're right. We explained it all. You're right. But what I was going to say is, as part of that conversation, you asked a really interesting question. Thank you. Which, for once. For once. Which is, why can't we have a one-click mortgage refi? Do you remember that? Yes. So, you know, like, I've refied a mortgage in my life.

It's kind of annoying. You know, if you're in the right situation, you can save money and it's probably worth it. But it involves a lot of paperwork, et cetera. And, you know, I think we're so used to one-click financial transactions or maybe two clicks or whatever.

One of the things that came up is that there are often a lot of mortgages out there that theoretically are sort of in the money where the borrower or the homeowner could take advantage of lower rates, but they don't for whatever reason. Perhaps one reason is they don't know that rates have gone down. Perhaps another reason is they can't be bothered to do all the paperwork and stuff. And so there are these lag effects. And so I've always sort of wondered, why can't you just have a one-click refi? Right.

Well, as someone who lives in fear of paperwork, I think this is an interesting question and we should talk about it. And it turns out we actually have the perfect guest. We're going to be speaking with someone who was involved in a one-click mortgage lender, Mike Yu, co-founder and CEO of Vesta. Mike, thank you so much for coming on All Thoughts.

Yeah, thanks for having me. So why don't you give us a very quick career summary? Why are we talking to you? Yeah, so I've spent my entire career in the mortgage industry on the tech side. Purely, I like to joke that everyone who ends up in mortgage origination stumbles into it by accident. So I worked at Blend. Kids don't dream of like one day I'm going to be a mortgage originator. You didn't dream of that when you were in elementary school. Anyway, keep going. Sorry.

Actually, when I'm recruiting engineers here at Vesta, I tell them, I'm like, you know, lots of founders will give you some weird story about how they've dreamed about doing this thing since they were 12. I could tell you a story like that and you wouldn't believe me anyways. So let's be honest, we stumbled into it. I started at...

a mortgage tech startup blend in 2016. So the company was about 50 people back then. And we built a ton of the bar we're facing experience for big banks, big mortgage lenders, et cetera, with the goal really being how you make the process more digital. Like it was all paper forms back then, even for the borrower to fill out.

And then how do you eventually consolidate that down into one click or one tap if you're on mobile? That company went public in 21, but I left in 2020 to go and build a different startup, Vesta, where I think a lot of what we struggled with that blend was the core infrastructure in the back end of the system made it really hard to consolidate the mortgage process, accelerate it, cut costs for lenders, save time and make it easier for borrowers.

And so kind of working on the backend system of record now where I think a lot of the other technology problems are. And I think as this is about one-click mortgages, I think there are, of course, some technology limitations, which is why we started the whole company, but there's also a fun variety of regulatory implications that I'm sure we'll dive into today too. Great. You know, what's funny is I think not only on the podcast did we talk about why were there no one-click refis available. I think we specifically put out a call. We were like, if you ever have

been in a startup, some Y Combinator thing, which is attempted to do one click reach out to us.

And you were one of the, I don't know if Blend was ever a Y Combinator thing, but that's not really that important. But you answered the call, literally, and you heard it. Answered the oddlots call. Answered the oddlots call. So you're the perfect guest. Obviously, we want to get into what you're doing at Vesta and just how it all works. Talk to us a little bit more. What did Blend do to attempt to solve the problem of, I don't know if it really gets to one click, but simplifying or streamlining the mortgage application process?

Yeah, it's really funny because it's only been 10 years. But 10 years ago, actually, if you wanted to get a mortgage, you couldn't even go and apply on the internet. I would actually say in the early days of Blend, you know, 2016, 17, we talked to lenders and they would be like, you don't understand. People applying for a mortgage, they don't want to do it online.

which just blew our mind. And there were a lot of old school loan officers who were like, I call my borrower and I interview them. And I literally take the paper that's called the 1003 or the Uniform Residential Loan Application. I take this paper form and I fill it out with a pencil. I pull over to the side of the road. My borrower's talking to me on the phone. I take them through my process and I fill it out with pen and pencil. Then I give it to my assistant and they go like type it into the backend system. And we were getting this kind of pushback left, right, and center in the early days, really just,

around the idea that people wouldn't want to apply online. And then Rockets big Super Bowl ad came out, push button, get mortgage. And then all the banks were like, oh, well, if Rockets gonna do it, we probably need something competitive with this. And I would say that was really an inflection point for that company. But so much of it was like, if you want to get to a one-click mortgage, well, first of all, you need people applying for the mortgage on the internet, not via physical paper.

And then it really becomes about how do you start to pull in the data from all these various sources? So Blend was one of the first to work with some of the GSEs on getting asset data direct from the banks and pulling that into the mortgage application so you can get faster underwriting and not needing to upload a whole bunch of paperwork and bank statements.

similar things for like income data and then people are making big pushes around property data and avms it's a whole variety of data sources you've kind of got to stitch together in order to really save the borrower from sending in a whole bunch of paperwork because one thing that i think is a little less obvious when you send the lender your pay stub you're not just proving that you have the income they actually take a bunch of the numbers on that pay stub that they didn't you know ask you to fill out anywhere and fill that out in a spreadsheet or something to calculate your income based on their models and what the gses tell them to do and things like that

And so, so much of it was just, if you can digitize the process and get that data in a structured format from a whole bunch of sources at the beginning, the belief was that's really gonna drive you towards a faster, more efficient process and eventually one click.

So you mentioned the GSEs a couple of times there. I imagine if you're doing a mortgage, at some point you're going to have to get the guarantee from Fannie and Freddie. And so you're going to have to go through them. What are their systems like? You mentioned you worked with them. How did you plug into the GSE systems? Yeah. So as you might expect from a couple of large financial institutions that are also now under conservatorship,

Their systems are definitely of varying degrees of maturity. One thing that I will say that I actually really appreciate about both Fannie and Freddie is they have invested a lot in technology over the last decade. I think when you get down to it, financial products are all, the nice thing is there's no physical commodity, right? They're all, it's all money and numbers in a ledger. And so I think they've definitely started to embrace their role more as needing to provide technology to the ecosystem that is modern, that is effective. But very honestly, like the core piece of technology they built that

that gives you guidance on whether your loan is going to qualify or not for sell to Fannie or Freddie. It's called Desktop Underwriter in the Fannie case and Loan Product Advisor in the Freddie case. The original versions were built in the 90s. And so definitely some older systems, it's all still XML system to system conversation if you can integrate to them at all. Some of the systems don't have

you know, any capability for system to system integration. And so if you want to actually sell the loan to Fannie or Freddie, someone has to go to their existing loan origination system, download an XML file, log into their website and upload it. So there's definitely, I would say, varying degrees of modernization and capability across those technology systems. And that is, I would say, no more true at Fannie and Freddie than it is at most of the major financial institutions you think about, even most of the smaller lenders you think about. It's all kind of on the spectrum of

Everything was built between one and 30 years ago, and everyone's got to kind of move. 30 years is long enough in technology. Everyone's got to kind of move and rebuild a new version of this, a new version of that. Oh, yeah. Joe, I remember, you know, those charts that show like all the acquisitions that a JP Morgan or Bank of America has done. It's sort of like a flow chart. Yeah.

Every single one of those probably has a different IT system. So I always hear that one of the big difficulties in building a giant bank is basically sorting out the IT.

Totally. Big institutions, you know, it's easy to sort of assume that sort of quasi-government institutions are going to be worse on so forth. And maybe sometimes that's correct and sometimes that's not. But big, gigantic institutions, particularly ones that had all kinds of mergers and roll-ups, et cetera, they all have this. And this has been something that's come up a little bit in the past.

We've done some episodes on bank software in general, and so I'm not particularly surprised to hear that Fannie and Freddie have a lot of still work to do, even if they have invested. Why is it hard? Maybe from your perspective, from the perspective of...

of either a Fannie or Freddie or just any other gigantic financial institution, how would you describe why it's challenging to update these systems so that they resemble the type of software we're used to in 2024?

Definitely. Maybe I'll start even way back, like 50 years ago. Are we going to talk about COBOL? I hope so. We can talk about COBOL if you'd like. But I think it actually, when I was at Blenda, I was fortunate enough to work with Tim Myopolis, who was before Blenda CEO at Fannie Mae. And he has this line which really stuck with me, which is that everyone says that banks are like slow adopters of technology. But the problem with banks actually is that they were very early adopters.

of technology, right? Going back to everything really is just a number inside a spreadsheet at a bank. There's no corn that you're shipping or gold bars or whatnot. The financial services industry was really an early adopter of technology. What that means is they installed a lot of technology very, very early on that then became harder and harder to rip out.

And one thing that we find, for example, at Vesta, where we're replacing one of these core systems, when I talk to other founders in the technology space, it's much easier to install a new system to replace a spreadsheet that just like so obviously doesn't work. The enterprise security is terrible. The controls are terrible, etc.,

than to get someone to upgrade a system that kind of works for them. It's clunky, it's inefficient, it's slow, but it isn't a burning pain where they're like, "Oh, if I don't modernize, I'm going to lose the business or lose my job or something." On the other hand, I would say there's a very strong incentive in all these big institutions. If you try and do a huge modernization project of a big existing system of record,

and it doesn't work, you're basically putting your job on the line. And if it does work as a CIO or a line-level CIO at a bank, you're getting a small promotion. So the trade-off is pretty bad. I was just going to ask, like, how much is it tech, Quay Tech, versus...

institutional inertia and incentives that really create that problem of why it's harder to upgrade? I think it's mostly institutional inertia and incentives. There certainly is a lot of work that actually goes into it, right? And so you've got to get budget, et cetera. But it's all very tractable. I will say in financial services, there are relatively few technology problems that are like fundamentally hard technology problems. Like we're not launching rockets over here. They all tend to be people problems, organizational problems, inertia problems that get in the way. ♪

Wells Fargo seeks broad impact in their communities. They're focused on building a sustainable, inclusive future for all by supporting housing affordability, small business growth, financial health, and other community needs. That's why they've donated nearly $2 billion to strengthen local communities over the last five years. Wells Fargo. The Bank of Doing.

See how at wellsfargo.com slash say do. Wells Fargo's philanthropic support includes contributions from Wells Fargo and Company, Wells Fargo Bank N.A., and the Wells Fargo Foundation. Even if you think it's a bit overhyped, AI is suddenly everywhere, from self-driving cars to molecular medicine to business efficiency. If it's not in your industry yet, it is coming fast. But

But AI needs a lot of speed and computing power. So how do you compete without costs spiraling out of control? Time to upgrade to the next generation of the cloud, Oracle Cloud Infrastructure, or OCI.

OCI is a blazing fast and secure platform for your infrastructure, database, and application development, plus all your AI and machine learning workloads. OCI costs 50% less for compute and 80% less for networking, so you're saving a pile of money. Thousands of businesses have already upgraded to OCI, including Amazon.

Thank you.

That's oracle.com slash strategic. Can you talk to us about the sort of life cycle of a mortgage in terms of technology? So like what's the first thing that happens? What system is it put into and then where does it go next?

Sure. So we're talking one-click refis today. So we'll start with the refi. In, let's call it a relatively idealized case. Let's talk the real ideal, which is like the borrower is going to get an email from their servicer, right? Which says, hey, you're in the money. Like we service your loan. We know what you pay. We know what rates are. We know, you know, roughly your credit profile. We can tell you that you probably want to refinance.

So the consumer is going to click on that link and they're normally going to go and fill out an online application. Today, they're going to go type in a whole bunch of their data again. I think you probably know your servicer has a ton of data on you. Do you really need to type it in? And this varies kind of depending on how tech forward your servicer is. But often people are still typing in their whole application again. They're uploading a whole bunch of documents. And this is kind of sitting in the consumer facing system, which today is, you know, they call it a point of sale. This is the space where really Blend is the category leader now.

And so the borrower is going to kind of type all that information in. They're going to hit submit. That's going to push it to what's called a loan origination system on the back end. And you can think of that as both the system of record and it's going to do all the compliance checks. It's where the people are going to do all the processing and the underwriting. It's where any automated underwriting might happen. And it's also the system that's going to be integrated to like

15 other systems. So one really annoying thing about the mortgage ecosystem is to produce your loan from front to back, you're probably hitting at least 15 different technology vendors. - Wow. Can you run some, what are some of these and what different parts of the stack are they serving? You don't have to list all 15, but give us an example of like the various things that need to be hit and who's doing them. - Yeah, so I think of it as there's a whole bunch of stuff around the property, right? So you've got to go to a title company, you've got to go to an appraisal management company,

kind of get the appraisal. So there's a whole bunch of stuff around the property. Someone's got to check what flood zone it's in. And we'll get into why some of the rules are really hard to change. But if you want to know what flood zone a property is in, I mean, you can go on House Canary or Zillow and kind of figure that out pretty quickly. If you want to sell a mortgage to the GSEs, you've actually got to hit one of their four or five designated flood certificate providers for an official flood certificate, quote unquote, which is really just

Those are the providers that signed a deal with the GSEs where they get the FEMA maps that everyone else gets and they produce a piece of paper that's official enough and the GSEs trust them. So that's a provider you basically have to hit. And so there's a whole variety of property vendors you've got to hit around those categories. There's a bunch of borrower vendors you've got to hit around pulling credit is the obvious one, but you're going to want to verify their income. You're going to want to look them up in fraud databases.

So there's a whole set of those. And then there's a bunch of compliance stuff to do. So generally, you know, there are entire companies that are dedicated to, I have all the data in the mortgage, I'm going to prepare the disclosures for you. Like the disclosures are so complicated, that's less a technology problem. That's like those companies have an army of lawyers who basically read all the regulatory updates, all the updates in each of the 3,800 counties in the US.

any state updates, any investor updates around exactly what you have to tell the consumer before they can kind of sign a lien on their property, which as you know, like some states are very onerous about that.

And so between compliance and property and kind of checking the borrower, there's just this whole constellation of stuff that has to be done. A lot of data sources and a lot of rules. Talk to us a bit more about the rules then. Like, I'm curious how these rules come into place, what sort of factors they're being based on, and then how often they actually change.

Yeah, I kind of think of rules in two buckets. They're, of course, the regulatory rules. So you can imagine a ton of those regulatory rules were driven by 07-08 and a lot of the things that we saw during the great financial crisis or that kind of led up to the great financial crisis. So a lot of regulatory stuff around what you disclose to consumers around, you know, you have to qualify their ability to repay in order to have a compliant loan. So there's lots of regulatory stuff. And then there's investor rules, which overwhelmingly, you know, come from Fannie or Freddie. There is a small private label market and some other stuff.

One thing that is true about all of these rules, right, the investor rules and the regulatory rules, is they're both kind of set, again, talking about organizational inertia, by pseudo government institutions that have really been burned by mortgages in the last two decades. And so they're pretty nervous about that.

And there's very little incentive to simplify the process or remove rules. And so the rules change, I would say every month or two, you get a few new rules from the investors, but they very rarely subtract rules, which tends to be a reason that you end up with. I want to say the Fannie Mae selling guide is now 1200 pages basically of rules that the loan has to satisfy. And these rules can vary from relatively straightforward things like you can't refinance an FHA loan within a certain amount of time after the loan was originally

originated. So there's like a seasoning requirement. You know, there's a lot of documentation rules. Like if you're going to provide an income to Fannie Mae, you normally have to have a pay stub and a W-2 attached to it to kind of

verify that income. And you get into like really complex and arcane rules as well once you get into 1200 pages, like, are you allowed to have a 10% increase in income year over year and use that new income? Well, you have to document that a certain way. If it's about 30%, you have to document it another way. So a whole kind of slew of rules, which is why you have this huge body of people basically that have to work on every single loan because they have to learn all the rules.

So one thing that would be really nice and it sort of gets to the one clickness of what we're trying to get at is if it were really easy to pull in data quickly from all these disparate providers. So I go to your website and I want you to know my income and maybe I want you to know my assets that I have, but I'm not sure if that's as important in a mortgage.

And I want you to know the location of my property so that you can do various things, including see what the floodplain looks like. The various providers of this stuff. So one of the providers might be my payroll provider. Another provider might be the bank that I use, the bank online, etc. How forthcoming are they in making these systems easy to

for a third party, say yours or a blend or some other FinTech or a rocket, et cetera, to just go access them such that I don't have to download PDFs and then re-upload them somewhere else? Yeah. So I imagine you're asking because you know the answer is going to be there. Very mixed results. Okay. So when it comes to banks, for example, there have been a whole advent of new kind of players that help you connect your banking data. Plaid or something. Plaid is the big one.

Exactly. And then the banks, I would say, clearly had mixed feelings about it. People are worried about security of people typing their bank password on something that's not the bank's website. They're very notably over the last 10 years or so, but a number of times when Chase just shut off Plaid's access. And so there certainly was some complexity in that relationship earlier on.

With some of what's going on in open banking in the US, I think the idea that you'll have access to your own asset information is definitely a place the regulators are pushing as well. And that gets easier and easier every year.

Payroll is a particularly interesting one. It's a very hot topic in the mortgage world now because everyone basically uses a product offered by Equifax called the work number, which you may have heard of. So the work number, it basically, they have a partnership with ADP where ADP charges them a very large amount of money actually to use the borrower social to look up the data. And then the work number turns around and marks that up a whole bunch. So they also charge a ton of money to the lender. So I've heard of lenders spending, you know, like

four or five, $600 a loan, like to close one loan to get that income and employment data in this digitized way via the work number, which is I think obviously ridiculous.

And so there certainly is some struggle going on with the data ecosystem. There are a bunch of startups now trying to basically do what Plaid did where the borrower can log into their payroll provider. One problem you might imagine with that is like, do you know your payroll password? Because I don't even know who my provider is. No idea. And so there certainly is some difficulty in getting the data together. I would say the ecosystem's made a lot of progress on that in the last 10 years.

But payroll and income tends to be a lot harder because it's much more fragmented. And with banking and open banking and every bank has some kind of electronic system of record, that's a problem where I'd say they've made a lot more progress. How much does mortgage financing depend on just your sort of basic mail? I want to be able to say the housing market is powered by FedEx or something.

So actually, every closing, I wouldn't say every closing package, but the vast majority of closing packages in this country, like you go to your lender or title office and you sign the closing doc and the notes and whatnot. Actually, the legally binding piece of paper that says there is a lien on my property now, like I have to pay this loan back.

And that gets FedExed back to the lender and the lender then scans it, they upload it to their electronic system and they turn around and they FedEx that to a doc custodian. And some lenders are more efficient with their FedEx schemes than others on like it just goes straight to the doc custodian and the doc custodian scans it and sends it to them. But I would say that very much all of the actual debt and recording and like all of the legally binding stuff

Probably 90% plus is still physical pieces of paper that are getting mailed around. There was a big trend, especially in 2020, around how do you digitize those notes? How do you do e-closing? Then it's a matter of you've got to get 3,800 counties to accept it. All the title companies have to accept it, et cetera. So it's a big network problem. Again, you're probably hearing a theme of it's just like organizational inertia. But you very much can say that home financing is still powered by FedEx.

because pretty much everyone, I mean, we have a field in our system where people are like, we need a FedEx tracking number field for the note. Like, not even kidding. And that could be really cool if you could integrate that to FedEx to like automate, like looking at the tracking. And it's like, of course you can do that technology wise. But sometimes you ask yourself like, what problem are we really solving here, guys? Like we should just move it to the cloud.

So I think the last time I applied for a mortgage was late 2017. And I just remember like documents and documents and checkboxes. And I didn't read any of those documents. I just signed the checkbox and I assumed it was all OK. What were all those checkboxes I was or signature boxes that I was putting a digital signature into?

A lot of those signature boxes are basically people disclosing your rights. So very similar to when you have a checking account and you want a bunch of disclosure. Yeah, it's like you sign something that gives them authorization to pull your credit in many cases. And then you sign something that, you know, it's like, hey, here's your credit score and here's how it was calculated in the state of California. Or you might sign something which is like, if you are getting an FHA loan with lead paint, which I hope you didn't, there's a disclosure that says, hey, like we determined that the house has some old lead paint, like sign here to acknowledge that we disclosed that to you.

And then the other thing that happens is because each of these disclosures are legally mandated and it's really hard to make sure that you signed all of them in the one go, they'll just say, hey, when you first apply for the loan and we disclose to you the terms, we're going to stick all of those disclosures in there. And then when you get to the closing table, we're going to put them in there again, just so we've built a standard that, you know, you've signed it like you're at the closing table.

you're not going to walk away now. Like, let's just make you sign it one more time. So probably if I had to guess, you got 50 or something disclosures, depending on the state you were in. There's a whole bunch of, you know, some states are more onerous than others. But you probably also signed each one an average of like two and a half times. I did. Just real quickly, how different was that experience for me in 2017 than it would have been in 2007 before the mortgage crisis?

Well, in 2007, you can imagine there were way fewer disclosures. Actually, in 2015, they passed what's called TRID or Tila Respa Integrated Disclosures, which is actually the main reason you can't have a one-click mortgage today. So TRID puts a minimum timeline as well, where you have to give people, you know, within three days of getting what's called a full application, you have to provide them an estimate of all the fees that like really clearly in a very standardized format discloses all the fees. That comes with a bunch of disclosures.

And then you have to give the borrower seven business days from giving them that loan estimate to close the loan. And so we actually talk in mortgage now a lot about the 10-day mortgage because you actually can't have a one-click mortgage purely by virtue of the fact you need that seven-day waiting period. There's some other timelines in there. Even if you got rid of that seven-day waiting period, you'd still have to...

remove a whole bunch of other regulatory timelines to really get it down to one day. But in 2015, they released this new regulation, which I would say made it a lot more onerous, a lot more documents to sign. And I mean, it's good, right? Like pre-2015, people were getting loans and they were getting bait and switched. And, you know, people were having new fees pop up that they didn't know about. And now all that stuff is really strictly regulated, but it definitely adds to the paperwork work.

Wells Fargo seeks broad impact in their communities. They're focused on building a sustainable, inclusive future for all by supporting housing affordability, small business growth, financial health, and other community needs. That's why they've donated nearly $2 billion to strengthen local communities over the last five years. Wells Fargo. The bank of doing.

There are two kinds of people in the world. People who think about climate change and people who are doing something about it. On the Xero podcast, we talk to both kinds of people. People you've heard of, like Bill Gates. I'm looking at what the world has to do to get to zero, not...

using climate as a moral crusade. And Justin Trudeau. There are still people who are hell-bent on reversing our approach on fighting climate change. And the creative minds you haven't heard of yet. Really don't need to have a tomato in December. It's going to taste like nothing anyway. Just don't do it.

What we've made here is inspired by shark skin. It is much more simplified than actual shark skin. Drilling industry has come up with some of the most creative job titles. Yeah. Tell me more. You can imagine tool pusher. No. Driller, motorman, mudlogger. It is serious stuff, but never doom and gloom. I am Akshat Rati. Listen to Zero every Thursday from Bloomberg Podcasts on Apple, Spotify, or anywhere else you get your podcasts.

So how did Blend actually try to solve all of this? Because when I listen to you talk about all these sort of challenges in the mortgage market, it just sounds like an unsolvable kind of spider web of requirements.

Yeah, it certainly is very challenging. I wouldn't call it, you know, nothing is really unsolvable, except the regulatory timeline is going to be what it is. But for Bund, a lot of it was, hey, can you really get to a one click and tell the borrower that they're clear to close? And what that means is we fully underwritten everything, we know that you're going to close, the only thing we're really waiting on is the compliance clock.

And so you kind of split it up into the various things that get underwritten in the loan. So in property, for example, the most clear thing you have to do in order to say, I can instantly underwrite your property is they have to be able to get no appraisal and get instant title. And today title insurance is this whole other thing that I'm sure you could do 10 episodes on. - Oh yeah. - Some people have asked for them. - Yes, I would say every two years, some Silicon Valley person tweets that like title insurance is a racket and someone should go take it out. And I always get that tweet texted to me like 10 times.

But there's a complexity around, well, the problem with title insurance is actually someone does have to go to the county office, still in a bunch of counties, and go downstairs into the basement of the courthouse and get the key and unlock it and go look up the records for that house or something like that.

So you have to figure out how you're going to make title insurance instant, which there are a whole bunch of startups that have worked on, are working on, have had some success in digitizing that process in some counties. You've got to make the appraisal instant, which basically means you have to get a message from Fannie or Freddie that for this particular property, they've written a loan on it recently enough that you don't have to appraise it again. And so that's the property side. Those two things you can imagine combined already, like you're taking the 100% of properties in the US and you're shrinking your hitbox to like

20 or 25% or something like that. And then you've got to look at the borrower and for Blend, a lot of the approach was, well, we partner with a lot of big banks. And so can you get the banking data directly from those banks and use that to either figure out the income or use something like the work number to instantly get income. You can verify assets income

and the property, and then you can, of course, pulling credit is the easiest one because we've been able to pull credit digitally for decades and decades in this country. If you can kind of check all four of those boxes, then you're quite a bit further towards a instant clear to close. You're still not fully there. There's a bunch of stuff around the margins you've got to go and sort out. But it really is a matter of blocking and tackling, executing detail by detail. And it's like,

There are 1,200 pages of rules. I've probably read those 1,200 pages of Fannie rules three or four times. And you've just got to systematically tick them off one by one. What are you doing now at Vesta that you weren't doing at Blend? Yeah, so a lot of the struggle that we had at Blend was you really, because you own the front end of the process, if you could get to fully automated, you could pull it all in and be done. The problem was, you know, you heard how I talked about appraisal title. You kind of shrink the hitbox for what you can do fully automated.

And so what we found was that if you fully automated, like, let's say you really could fully automate 1% of a lender's loans, that would be great. But they're still, you know, spending a ton of manual dollars, a ton of, you know, operational people on 99% of their loans.

And the big problem was all of the data that you got at the front, it was really hard to use that to drive efficiencies at the back of the process or for any of the loans that did have even one manual touch. Like I ticked through all these rules. You can imagine if only 10 rules had to be done by a human, well, now it's got to go through this manual process.

And what it does today is it goes through this old manual process where they basically have to underwrite the whole loan manually because the system doesn't have an understanding of what's already been done. It's not, you know, task or workflow oriented. And people basically have muscle memory. So the underwriter is going to look at everything, order the appraisal, whatnot, even if they don't have to.

And so a lot of what we realized was the back end of the process was making it really difficult to realize any efficiency from the good work you're doing at the front end, because the change management and organizational inertia of, you know, you've got 3000 people on your mortgage manufacturing line, so to speak, doing exactly what they've always done. And the software isn't really guiding them to do anything different. Like it's not, it's not a piece of software. Like you might be used to working in today, like Slack gives you notifications, for example, it's really almost like a

spreadsheet with a different UI layer on top of it, and you've got to figure out exactly what you're going to do. So a lot of it was how do you change the way the operation works so that people are doing a lot less? And then the other big thing was with the existing loan origination systems being so difficult to integrate to, that was one of the biggest hindrances in actually getting all of the data and making that process one click was that you couldn't actually do all of the jobs that needed to be done by that old system. Like the old system...

has all of the integrations I mentioned, and they have hundreds of integrations to all these data providers, like coordinating the appraisal. So you ended up having to build around the old system instead of through the old system to achieve a lot of this stuff. And that just seems like so clearly the wrong way to do it. Now, the downside is you have to go and modernize the old system, which is a really hard problem.

But by kind of modernizing the old system, you unlock A, the operational efficiency that you actually get from all this data, and then B, a much easier platform for everyone who wants to build a front end to get that data through your integrations and through your processes that Blender already has that exists manually today, instead of having to recreate it on the side to try and automate it, if that makes sense. I have a slightly random question, which is, given that we're

talking about technicalities. How easy is it to commit some sort of mortgage fraud nowadays? Totally random, not out of personal interest. Yeah, I did listen to your recent episode about government fraud, where Joe was the one I think was very interested. Yeah, I was the one entering to start doing fraud. Yeah, mortgage fraud, I think, is actually quite difficult these days, mostly because there are so many human eyeballs that look at the loan.

And so let's take something really simple. Like you wanted to like doctor a document, like probably the most straightforward thing, because it's not like mass scale fraud. It's like somebody is like, I want a mortgage on my primary residence. I can't afford it. And I'm just going to doctor the documents to make my income look bigger. Well, first you have to hope the lender doesn't check some third party verified data source or they don't reach out to the employer, which they often do. And then you have to hope that like your document makes it through the processor looking at it and the underwriter looking at it and the closer looking at it. And the underwriters, especially they're looking for things that don't add up.

And so I would say mortgage fraud is probably really pretty to very difficult to actually accomplish today. It sounds a lot harder to achieve than like, you know, figuring out how to get some Medicare dollars. So it's probably not worth the squeeze. Now we'll like generative AI, make it way easier to make fake profiles and all that stuff. Maybe that's something that I think lots of people worry about. But today I would say it's definitely the mortgage industry has done a pretty good job of

I'd say the regulators have done a good job of making it really hard, just given everything that happened two decades ago.

Back to the question of refis. So you mentioned that theoretically, if you're a homeowner and you're in the money on your mortgage, that is to say where it would make economic sense for you to refi, you might get an email or something. It's like, hey, you should refi and you can save this much. But as we're talking about, it's going to be a lot of paperwork and all this stuff. After our episode came out several weeks ago, someone on Twitter, they said, why can't we have a mortgage product that

that you pay a higher premium upfront, but it's a floating rate mortgage that only resets downward. In other words-

basically, if rates drop lower, your mortgage mechanically drops with it. And again, obviously, if you're going to have that, theoretically, that's a more valuable option and you pay some premium upfront. But then in theory, you save all of this effort and time and document checking and human hours that go into this. In your mind, does that seem like a plausible financial product that could exist? Seems like a totally reasonable financial product. I think that

There may even be somebody doing it in like the private label securities market. There is a small market of basically hedge funds that will like underwrite non-QM is what they're called mortgage products and offer those to lenders and lenders can originate them. One thing I will say is it seems unlikely to come from the GSEs just because so much of the GSEs mission these days is affordability and democratizing homeownership. And I can't really think of a marginal person that that product would get into a home. Right. So even if it makes sense, that's just now what moves the dial.

Yeah, I think it makes sense from a single person financial instrument perspective. I would love to have one of those, for example. But I think that from the kind of stated policy goals of the biggest investors in the market, it's just not really something that aligns with their policy goals. And so I can't see that being a big area of where we're going to see a bunch of those in a decade.

So a lot of mortgages get bundled together into mortgage backed securities. I'm curious, like how much of that granular detail about pay and, you know, lead paint in the house and things like that gets ported over to the securitization aspect of it? So it's really not a lot. I actually I have capital markets people reaching out to me all the time being like, if you have a modern loan origination system, you can solve my problem if I can't actually get any of the data

or much of the data that underlines these instruments when I am going and securitizing them or trading them or whatnot.

What generally comes out, as much banking technology still is today, is you export a big CSV of some of the data fields. You take a bunch of docs and you send them off. And then that CSV, which they fancily call it tape, but really it's a spreadsheet, just gets ingested. And now you've taken a process that had 3,000 fields and hundreds of pages of docs, and you've boiled that down into like 50 or 100 fields that describe the mortgage.

which maybe is for the best. Like, I'm not really sure that people buying MBS should be thinking about, you know, the specific credit profile of the thousand different mortgages that are chopped in there and put in. And so it's nice that there's some standardization, but it's definitely very lossy. And I will tell you, it is something that mortgage traders complain to me about a lot.

So what's realistic? It doesn't sound like you'd ever get like true one click because at a minimum, you're probably going to have to tell the front end who your payroll is and who your bank is and a few other things. What is a plausible version of if there's continual coordination among different banks, if Fannie and Freddie continue to update their technology so that

It's a little easier. What could it look like if I were to say in 10 years I'm applying for a mortgage again? I think it's very reasonable to strive for a world where it is, to your point, as close to one click as possible on the very front end. Maybe you're talking about a 10-minute application max where you connect some accounts.

Once that's done, I think that what you should get instantly is one of basically three decisions. Hey, you are definitely clear to close. You're going to get your disclosures and then we're just going to wait 10 days and we'll close you. Option two is, hey, you as a borrower are definitely clear to close, but we need some additional information on the property. So you're going to wait 10 to 14 days. You're going to pay for the appraisal and we're going to close you.

Or three is basically, unfortunately, you know, we're not able to close you here. Some things you can do to improve your stance. I think that property is probably going to be the thing that even if you ask me 10 years from now, it's like it's really a question of there will always be those corner cases. It feels like where the GSEs are going to want to see an appraisal unless they're the big credit profile change or if they get privatized. You know, there's all sorts of things that can happen in 10 years. But I think those three outcomes being 10 minutes away fingertips wise from the borrower and you close in 10 days.

is very much attainable and it's really what everyone in the industry is and ought to be working towards. Does blockchain solve this?

And I mean that somewhat seriously. No, yeah, yeah, it's a good question. Because like I've often thought like one of the few real world applications of blockchain technology could be in the mortgage assignation space where you have that sort of chain of title moving around constantly. But also I'm thinking like from a wallet perspective, if you could have like a personal profile that carried with it, you know, your pay and how much you're worth and etc.,

you could use that too. Yes. So those are two very interesting use cases. On the wallet perspective, the way that I think about blockchain helping here is it almost lets you build like a more encompassing decentralized credit bureau. And decentralized is actually really important because I don't think the banks are super excited about the idea of helping build like a fourth credit bureau. Like they built three and now they pay the three for their own data, which I think is a

a little bit difficult for them. And then, you know, the regulators are not super excited about these centralized credit bureaus, et cetera. And so I think the idea that each consumer could have their own key that unlocks access to all of their data on this decentralized credit bureau that all the payroll providers and financial institutions, et cetera, are writing to is very much an idea that has legs. It is really hard to implement for a lot of similar organizational inertial reasons.

But I do think that is a real use case for blockchain because it solves the incentives problem where people are basically like, I don't want there to be one middleman with every consumer in America's financial data. And so blockchain lets you kind of like decentralize and split that up. So I think there's some really interesting avenues there that people can go down. And there are some companies, I think, actually looking at that.

on the title front what i usually tell people on could title be on blockchain absolutely is it an interesting use case absolutely the hard part of digitizing title is getting 3 800 counties to even move to like putting you know capturing the records digitally and not in the courthouse basement getting 3 800 counties to move the blockchain seems further away than that not closer i don't know if i've ever mentioned it on the show before i once had a gig right after college

in which there was some company out in California doing some asbestos lawsuits. And they needed names of all these people who had been party to some suit. I don't know. They were maybe putting together a database for lawyers. And part of my job was to go to various county courthouses all around rural Texas, in central Texas, and go to the basement and just literally pull out files and ask for names. I'm just going to ask one last question. Since Tracy hit one tech buzzword, which is blockchain,

Generative AI, whether it's in the field of scanning documents or understanding documents quickly, in your work right now, is there a substantive use that you're getting out of this technology?

Yes, it is definitely to your point, it's scanning and understanding documents. And so you can think of mortgage the way that I think of it high level is it's a whole bunch of data and a bunch of rules. And the rules are well defined by investors, the government, etc. And the data is just data. And so data and rules to a lot of this conversation should be a one click experience. Like we've known for decades how to run rules on data.

And a lot of the problems come about because the rules are written in this 1200 page PDF and there's a little gray and someone has to learn them. And the data comes in a bunch of documents and a bunch of disparate places from the borrower. And so it's not structured. And so if you can bring structure to the data and structure to the rules, both of which generative AI is really good at, right? It can read the Fannie Mae selling guide and turn that into code rules. It can read a document and turn that into data points.

If you can use generative AI to structure those two things, then you still have structured data, structured rules, and whatever rules the GSE set. You don't have these compliance things with always AI underwriting the loan, but we're seeing a lot of really promising results taking the most cutting-edge large language models and applying them to these documents both to write the rules for us and to lift the data off the documents. All right, Mike Yu, thank you so much for coming on OddLots and explaining to us why we can't have a nice thing.

- Yeah, thanks for having me. That was fun. - That was amazing, Mike. You were the perfect guest. Thank you so much for coming on. - Yeah, that was fun. Thanks for having me. - Joe, that was really fun. - That was really fun. I thought Mike was exceptionally clear at explaining how all this works.

And although it's still annoying, the process of getting a mortgage and lots of documents that I didn't read and attached my signature to, like, I guess I understand a little bit more why now. Do you remember after the financial crisis, there were all these problems with loan documentation? And I remember like there's a big thing about assigning mortgages in blank that all turned into like court cases. Yeah. I kind of wonder.

We did a great episode on that with David. Yeah. David Dyan of the American Prospect, like in 2015. What was the name of his book? David Chain of Title. Oh, I totally forgot about that. We did an episode with David Dyan, Chain of Title, and how crazy that was. It was just the state of disarray in documentation after the mortgage crisis. But you understand why it is when you still have, you know, so much at the county level and the county level.

I guess for obvious reasons, not feeling any particular pressure to update or digitize or modernize or coordinate all of their systems. Yeah, that's really it, isn't it? It's sort of like a hodgepodge of state and county law. I feel like we're going to be waiting a while for a solution to this. Yes. And I've sort of hinted at it before on the podcast.

But for very arcane reasons that I'm not going to get into, I do have a loan that will need to be refinanced at some point in the next couple of years. You're very optimistic, Joe, about interest rates. No, I'm very pessimistic and I'm very anxious about it. But if I'm not optimistic about the path of interest rates, maybe I'll be optimistic that in a couple of years, the process is at least a little bit better than it was the last time I applied for a mortgage. Yeah.

We'll see. You'll have to tell me how many documents like get mailed out and stuff. Although I guess most of that is on the sort of like lender and servicer side. But I'm not looking forward to it. Yeah. All right. Shall we leave it there? Let's leave it there. This has been another episode of the All Thoughts Podcast. I'm Traci Allaway. You can follow me at Traci Allaway.

And I'm Joe Weisenthal. You can follow me at The Stalwart. Follow our guest, Mike Yu. He's at Michael underscore Yu. Follow our producers, Carmen Rodriguez at Carmen Arman, Dashiell Bennett at Dashbot, and Kale Brooks at Kale Brooks. Thank you to our producer, Moses Andam. For more Odd Lots content, go to Bloomberg.com slash Odd Lots, where we have transcripts, a blog, and a daily newsletter. And you can chat about all of these topics 24-7 in our Discord, discord.gg slash Odd Lots.

And if you enjoy OddLots, if you like it when we dive deep into the reasons we can't have one-click mortgage refis, then please leave us a positive review on your favorite podcast platform. And remember, if you are a Bloomberg subscriber, you can listen to all of our episodes absolutely ad-free. All you need to do is find the Bloomberg channel on Apple Podcasts and follow the instructions. Thanks for listening. ♪

What could you do if your data was working for you and not against you? With Bloomberg delivering enterprise data directly to your systems, you get easy access to the details you want, optimized for higher-level analysis, and financial data experts committed to helping you maximize your every move. Our data is made for more, so you can show the world what you're made of. Visit Bloomberg.com slash enterprise data to learn more.