Mike Yu, CEO of Vesta, on building a new system of record for the mortgage industry

Today, the existing systems are all very human driven. The primitive is really the loan file. A human looks at the loan file and translates that to, "I need to call the borrower and ask for this. I need these documents. I forgot I needed these documents; I'm going to ask for them too late. I already got them, but they're in my inbox, and I'm going to ask the person again." That’s because the primitive is too big. People are thinking about it as like, "I'm assigned a loan, and that's what I'm going to figure out." There's a lot of human autonomy to figure that out.

For us, the question is, "How do you take that primitive alone and break it down into these smaller, much more atomic primitives, where there's much less autonomy?"

Then, the computer can do the hard work of stitching that new set of smaller primitives together into a loan, as opposed to the human just taking the loan primitive and creating all this confusion. 

In many ways, I think that we're really taking the loan primitive that exists today and breaking it down into a bunch of tasks and a bunch of other different rules and other concepts we're developing, and then figuring out how to stitch those back together into completed loan files without a person's judgment being the messy layer that does it.

The mortgage industry's traditional systems of record are technological walled gardens. What do you think the switching costs look like for a lender to transition from their current system of record to Vesta? And what does a time-to-value look like? When is the "Aha" moment? What objections do you hear from customers?

The objection tends to be that it is a high switching cost project. 

One thing that I do really like about loan origination systems being like this system of record—I always use the analogy that the data in origination is like a river, it's not a lake. What I mean by that is the loan comes in, and you put it on this assembly line, ideally, and you get it manufactured or you get it underwritten, you originate the loan, and then it falls off. Then it goes into servicing, and it goes into investors, and it gets securitized. All that stuff actually happens outside your system of record. What that means is that the data in your loan origination system of record has a lifespan. It comes in, and then 90 days later, the loan's closed, let's say, and then the data's out of the system.

That means that you can actually do a switch by running two systems in parallel. You let the old system ’drain,’ and you let them close out all their old loans in the old system. At the same time, they start all the new things in the new system. 

After 90 days, you're looking at a relatively low percentage of loans in the old system that are still even relevant, and you figure out how to port those over one way or another. 

It's still a big project. It's non-trivial to run two systems at once. It's non-trivial to make the rest of your software play nicely with that. And the biggest objection definitely is the size of the project.

Why is building a new system of record the right wedge for a fintech to break into the mortgage lending space?

To be completely honest with you, it's not a very good wedge, but there are a variety of things that went into our choice. 

One is that we've already been around the space a lot. If we didn't have the existing connections that we have, and the existing knowledge that we have, and you asked me the easiest way into mortgage lending, it certainly would not be to build the core system of record and try to convince these big institutions to switch over. 

The sales cycles are really long. You really have to know the right people just to have the conversations, but no one's buying a loan origination system purely on the strength of the relationship with you. These big organizations can take months or years to navigate. The products are really hard and complex to build. But it doesn't feel like there's a good wedge to go and build and then later transition into building the system of record. 

The reason for that is that if you want to build any wedge, unless you are literally building a new system of record, you're going to have to integrate into the incumbent systems of record. Integrating into the incumbent systems of record is really hard. It's a ton of work, and you put yourself in a position where you're now dependent on these incumbent systems of record for whatever revenue stream you have.

It's not the best way to break in if you're new to the space, but we definitely recognized the need for a new system of record. We figured that probably the only way to do it was to go at it head-on. 

A lot of investors thought we were crazy. They were like, "Why don't you just build some skin on top of it first and then go rip it out from the inside?" The problem is that you end up building something that's very constrained by the architectural constraints of the existing players. If you really want to build something from scratch the right way, you've got to come out with a blank slate.

There's a vision of the future of the mortgage ecosystem where lenders and other financial institutions stitch together barriers best to breed point solutions instead of relying on one end-to-end platform. Do you see that as the future here, and why? What are some of the other companies you think are building towards a similar future?

I 100% think that's the future. One of the big core theses of Vesta was we were

at Blend, we saw all these new startups doing similar things. Polly is building a new pricing engine. There are these new compliance engines. There are all sorts of companies building these different pieces of the puzzle. One thing that was very clear to us was that there's a hole in the middle of all these other point solutions, and it's roughly shaped like the system of record. No one was building that standalone system of record that would be the platform for this modular ecosystem everyone's out there building.

The existing system of records have all played this grow-via-acquisition game, where they've bought additional pieces of the stack, and they try to stitch it all into one end-to-end platform. 

Increasingly, though, I think lenders are looking to differentiate more and more on how they think about technology. For them, it’s really important that they’re not buying end-to-end but that they’re really selecting from various best-in-breed providers. It's really hard to do that around within these existing walled gardens who want you to use your whole end-to-end platform.

In many ways, Vesta is the only system of record that will allow these financial institutions to become technology companies.

A few months ago, Zillow laid off their iBuying team, stating a difficulty in predicting future home prices. To what extent are you exposed or insulated from shocks in the real estate market?

We're definitely not exposed the way Zillow's exposed. A big discount on home prices is going to very directly affect how things go for an iBuyer. 

That said, everyone in the industry is at least a bit exposed because big changes in home prices lead to changes in how many homes get transacted, which leads to changes in the fundamental value that the industry creates each year. 

If the industry does 5 million loans a year versus 10 million loans a year, it's pretty clear that the industry as a whole added more consumer surplus or whatever in net world value in the 10 million case. Everyone's margins and earnings suffer when that number goes down. It's impossible to be in any industry, in my mind, and be totally insulated from cyclicality.

The other thing I will say is that there are some very interesting counterintuitive pieces here when loan volume goes down. Of course, we make less on a per-loan basis, but also, more people are looking to invest in new technology when they're not busy. When loan volume goes down, that actually tends to be an exceptionally good time to win new business. When loan volumes are up, you're making a lot on your existing business, but it's harder to win new business because everyone's too busy originating loans. They're not thinking about doing these big systems projects.

How does Vesta make money? How are you thinking about how that might evolve in the future?

It's all a SaaS subscription, so we charge an annual platform fee, and we charge a per-loan origination fee. 

In the future, we'll likely charge for implementation just because you don’t want this huge cost center around professional services. 

I think this is likely always going to be how we make money. There's some purist component to it in my mind, which is you want to make money by charging customers. 

You don't want to make money by, for example, charging a revenue share to everyone who integrates to your platform. That is what the incumbents do, and that is part of the reason it's so hard for a mortgage lender to have a differentiated technology offering.

I think being very focused on, "Hey, you should deliver a lot of value to your single end customer and then charge to a SaaS subscription," is a very, maybe purist and boring way of looking at it. Long term, I imagine we'll charge people for other kinds of banking, SaaS too potentially, when it comes to evolution in the future. But charging people SaaS revenue is always going to be the plan.

A mortgage can take up to a hundred hours and cost as much as $9,000 produced on the backend. Can you talk about what makes mortgage origination so costly and time consuming, and to what extent can gains be made by digitizing the process?

A lot of this comes down to the fact that the primitive is a loan, and you hand it to a person, and they go and figure it out. They use their brains and some books to figure out what documentation they need, and so on.

Underwriting a mortgage is fundamentally very unnatural if you think about it. You're underwriting someone to make 30 years of payments, which is an insane time horizon to ask somebody to predict the future. As a result, the mortgage itself has become quite unnatural. It's very US-government involved. You've got Fannie Mae and Freddie Mac. They've laid out, based on their long experience, what they think underwriting guidelines should look like with respect to what documentation you need, what you need to prove, what ratios are okay of how much debt you have versus how much income you have.

However, that’s mostly a historical look back, so it's hard to intuit from first principles how you should really underwrite a mortgage. 

Also, because the process is so complicated and un-intuitive, your human is probably going to get it wrong. There's a variety of solutions for this. You can’t get it wrong, so lenders say, “Instead of having one human do the process, we're going to have five humans independently do the process." Now they're all going to get it wrong in the same way only 3% of the time. 

That's really what mortgage shops have done. They have a processor who's the first line of defense, who goes and makes sure they collect all the documents, who cleans up the loan file, who makes sure it all adds up. 

Then, the processor passes it to the underwriter, and the underwriter basically does the same thing—they go look at the documents, make sure all the data lines up, and make sure nothing's fishy. Then, the underwriter will also make a credit decision.

The underwriter normally is not going to send it back to the processor and be like, "You missed these things; go do them again." The processor's going to do them, the underwriter's going to check their work, and this happens three or four times. After that, as you get ready for closing, it's going to move to a closer. And the closer's going to go look at the loan file again, make sure nothing's fishy, make sure all the documents are there, make sure they match the data. After the closer, before you fund the loan, you send it to post-close QC. They do the same thing: look at the data, make sure the documents are all there, and make sure nothing's fishy. What you end up with is different people doing the work five or six times. 

To make it even worse, this makes it almost impossible to automate because it's not like anyone is doing work they don't expect to have already been done. Even if you automate the work at the front end, your people will still do the work five times because they're used to it, because there's no concept of, "This work's already been done, so don't do it." 

As a result, prices have pretty much monotonically gone up. Mortgage gets more complicated every year because there are new rules, there are new things where Fannie Mae's like, "Oh, we lost a bunch of money on this risk. We've got to go patch over this risk in our credit box,” so the process gets more complicated. You can't automate anything or really learn from your existing process. There's no structure. It's five independent humans thinking about it in five different ways. And you paid to do the same origination work many times, and it just adds up.

Historically, it's been difficult for fintechs to build in this space largely because of government regulations and high capital requirements. Is there any regulatory context around the mortgage space improving the odds for fintechs? And if so, how?

The regulatory context is really interesting because it's very administration dependent. For a few years, it looked like the GSEs—so Fannie Mae and Freddie Mac—were leaning much more into fintech programs. There was a lot of talk under Trump around privatizing them and around letting them try new things. 

With the new administration, one of the things that happened is that people realized, "Oh, these entities are probably not going to go private." Fannie Mae's stock trades at like a dollar or something. And the reason for that is that no profits are entitled to the owners of Fannie Mae stock. It all goes into the treasury.

It used to be two years ago, if you were a fintech, you'd get into Fannie Mae's fintech program, sell them loans, and do all that stuff much more quickly. Now, we've seen that pull back dramatically, and it's much harder to start those relationships, and those relationships are really the underpinning of building a mortgage lender.

Let’s say you want to build a new mortgage lender. You're doing relatively conventional stuff but differentiating on where you're getting customers. If you don't buy a mortgage lender with an existing relationship to Fannie Mae, which is what a lot of the big famous fintech names did, building a relationship with those GSEs from scratch is actually much harder than it was just a few years ago. 

We'll see if that changes. I think it’s always going to be in a state of flux, and by and large, the macro long term view is going to be more technology-friendly. But in the last couple of years, you've seen a blip where, due to the very specific administration change around FHFA and Fannie and Freddie, you've seen it actually get a little bit harder for the last couple of years.

How do you think about Vesta's positioning regarding more traditional loan originating software companies like Blend, Floify, TurnKey Lender, SimpleNexus, LendingPad, BNTouch Mortgage, and MeridianLink?

They all do pretty different things. Now it's Blend, Floify, and SimpleNexus are all mostly frontend, consumer-facing.

Over the last 10 years in general, the mortgage technology players that have been built are these point solutions in this ecosystem we talked about earlier, where lenders are going to pick best-in-breed across the board. 

We're very much building the system of record and the workflow engine to drive all the automation and nothing else. We're building an API ecosystem to plug into, but we're not building a consumer-facing experience. We're not building a CRM. LendingPad is probably the closest analogy, where LendingPad is also trying to build a system of record, although they're not really building the workflow engine piece of it. 

Overall, it's just a very different part of the stack. We are building the part of the stack that's probably been most neglected in the recent wave of mortgage tech players. And honestly, that's probably because it's the hardest. It's really hard to start there, and it's really hard to get your first few sales, but I think we'll play very nicely with most of the players that you mentioned in this broader ecosystem that we see evolving.

How do you think about Vesta's positioning regarding other digitally native startups building for mortgage processes like Valon and Polly?

Polly is very much in a similar boat to what we talked about before. They’re building a new pricing engine. There's a lot of room for improvement in pricing engines. They’re, again, building a piece of software that they're going to sell to lenders that will interweave nicely with Vesta and with Blend and with a bunch of these other new-age players that are letting lenders build their own technology stack. 

Valon is interesting in that Valon is actually building a financial services company in some ways. They are very tech-enabled. They're building the tech for themselves, and then they're tackling mortgage servicing, which is a horribly, horribly painful thing.

Broadly, there are two categories of these mortgage tech players. 

There are the SaaS players who are selling to lenders, and we integrate well with those. Those are actually a foundational piece of our story, in that as you see this proliferation of mortgage SaaS players, lenders want to become technology companies and pick and choose who they want to work with.

Then there's also the technology-enabled financial services companies, which Valon falls more into, or better.com falls into, et cetera.—people who are originally doing loans. Those are really good for a couple of reasons. 

One is they’re potential customers for us—people want to build certain parts of their own technology, but they don't want to build all of it. There's a lot of technology inside a mortgage lender, and building all of it is pretty painful.

The second one is even if they don't become customers, those are the players that are really setting the benchmark for the industry in many ways. The proliferation of better.com and many of the other players that want to look like that really contributed to this idea that mortgage lenders have to become technology companies. They’ve helped catalyze this transformation.

Blend expanded from mortgage lending to lending and finance more broadly across fintechs. What are some opportunities for growth that you see for Vesta in terms of new markets or products? Do you think about expanding deeper into the needs of financial institutions or expanding within the mortgage use case to provide customers with a better experience?

The high-level framing I use for people is that the mortgage loan origination system is meant to be a system of record that is cloud-native and API driven with an ecosystem around it. 

The financial institution of the future is a technology company that buys that platform and builds whatever they want around it. That's true in mortgage, and that's going to be true everywhere. 

We could go much deeper in mortgage and end up in this awkward vertical SaaS space, where you end up competing with a lot of your partners. But we see lenders wanting less and less to say, "Hey, we want one vendor to dictate all our decisions for us," and more, "We want to take an active role as a technology partner and decide who we want to work with."

When they find that partner, they're going to say, "Hey, we need a similar problem solved for us in core banking. We need a similar problem solved for us in other lending products." 

I think where else we decide to build this core infrastructure upgrade will be very customer-driven, but if we're successful in mortgage, I think it's most likely that we'll be driven by customers into other verticals of financial services to do similar core infrastructure upgrades there.

This transcript is for information purposes only and does not constitute advice of any type or trade recommendation and should not form the basis of any investment decision. Sacra accepts no liability for the transcript or for any errors, omissions or inaccuracies in respect of it. The views of the experts expressed in the transcript are those of the experts and they are not endorsed by, nor do they represent the opinion of Sacra. Sacra reserves all copyright, intellectual property rights in the transcript. Any modification, copying, displaying, distributing, transmitting, publishing, licensing, creating derivative works from, or selling any transcript is strictly prohibited.