Editor’s note: This feature originally appeared in the March issue of DS News
There’s no doubt that we are living in times of accelerated change. If you look at how any industry has evolved over the past 100 years—whether it’s travel, communications, or entertainment—you will see a sharp contrast between how slowly things changed a century ago compared to how quickly they are changing today.
As just one example, vinyl long-playing records, or LPs, were the standard for decades until the mid-1960s, when 8-track tapes and cassettes took over. Then came CDs in the 1980s, followed by digital music files in the early 2000s. Now most people stream their favorite music through online apps. LPs still exist, but their primary value is based on nostalgia.
Eventually, everything changes, but some things take longer than others. Innovation such as artificial intelligence, machine learning and data capture technologies have taken hold in loan manufacturing but have been slower in adoption on the secondary marketing and servicing end of the market. Here they could be enabling faster, more efficient MSR acquisitions and onboarding of loans. Their potential to transform the secondary market is undeniable.
The Costs of Yesterday’s Technologies
An abundance of diligence questions along with legacy systems with wildly differing customizations has led to a lack of any type of consistency or uniformity in how institutions behave in the secondary market today. As a result, an institution that sells loans of the exact same asset class to both Buyer A and Buyer B is going to have two totally different experiences. This lack of continuity really makes it cumbersome for all parties involved.
There are advantages to using trading platforms, of course. They are, on the whole, a much more efficient way to manage capital market price, time of trade, and reduce trade fails. But these platforms do not address the validation of loan documents and data to identify any data inconsistencies. In a single file, you could have a piece of data that says one thing, and documents that say something else entirely—and no means to reconcile the two.
Data ingestion is another factor that contributes to variability because of the fact some institutions have modified their systems so they are capable of boarding a sizable amount of data, while other institutions haven’t. Even though there is more loan data available than ever before, those that are slow to adopt have no place to store it—so they stick with spreadsheets or archive the trade tape, figuring that as long as they have the data somewhere, it’s okay.
Most secondary market trading platforms have also failed to solve two of the biggest issues MSR traders face, the first of which is timing. Given how quickly prices change in the capital markets, spending days or weeks on MSR acquisitions can be costly and result in the loss of better opportunities.
The second issue is the functional fulfillment of the loan and making sure that you bought what you thought you did. This is where things can get clumsy, given the overall lack of data quality and the high degree of variation in document ordering and naming from sellers. In the past, the spreadsheet became the great equalizer, using macros to map data and manual processes to determine if data was missing and to enter it into the buyer’s systems. That approach is neither accurate nor scalable.
On the origination side of the business, consumer demand for a simpler, faster, digital experience has motivated lenders to adopt increasingly higher levels of automation. It’s quite a different story in the secondary market, where institutions have been slower to implement automated tools and seem to remain loyal to embedded manual practices. In fact, capabilities already exist that could enable institutions to conduct MSR trades and onboard loans in a fraction of the time it currently takes. With the increasing adoption of these new technologies, however, things are starting to change.
Understanding AI’s Impact
One of the most exciting things happening in the secondary market today is the emergence of AI technology, as well as machine learning tools, which are a subset of AI. AI describes technologies that analyze data and make decisions based on data patterns, whereas machine learning describes technologies that learn to distinguish patterns in data from human instruction and self-learning algorithms.
In truth, AI and machine learning, while often talked about, are not as commonplace as many are led to believe. They can create the uniformity these institutions need through the data normalization they provide across sellers’ loan files, making it much easier to ingest documents and data accurately. This then can lead to the use of more sophisticated applications that can significantly help servicers and investors gain greater insights on their portfolios and their trading decisions, especially when determining where risk lies and which loans to sell versus which loans to retain.
AI and machine learning technology also enable MSR traders to capture a greater amount of data off loan files and then automatically run business rules around that data. This eliminates the historic “stare and compare” methods of checking data and loan quality and greatly reduces the time and overall cost of due diligence. By green-lighting the vast majority of files in which the data can be trusted, companies can focus on exception-based processing, which allows them to move loans forward much more efficiently.
A particular benefit of these technologies lies in the fact that different investors look for different data elements when determining risk. There’s a huge variation among loan purchasers in terms of the data elements that they are concerned about. Certain buyers require checks on 40 different data fields, while others may want to look at 30 or 50—there’s really no consistency among them. Their legal agreements all differ as well. If an investor needs to check 30 unique data elements when buying a pool of loans, they could do so in seconds using machine learning tools to aid the process, rather than spend several days poring over them by hand.
These tools are also great for filtering loans. For example, if I’m buying MSRs and I don’t want a large concentration of loans in a specific state or zip code because of the perceived risk involved, I can get that information with a click of a button. I could also build a view of key data elements that are important to me for the loans I’m acquiring and have this information presented in nanoseconds.
New technologies also give sellers a huge advantage when it comes to market timing as well, which is a big deal in a rising rate environment. When you’ve made a commitment to sell a pool of loans and there is an agreement in place and a price locked in 10 or 15 days, and you don’t deliver the files in time, you’re at risk of having your loans repriced at today’s rate, which cuts into the premium you were expecting to receive. Because new technologies enable faster trades, the timing issue virtually disappears.
When it comes to onboarding loans onto one’s servicing platform, buyers want to make sure there are no data defects and make sure they reach out quickly to borrowers to let them know they are their new servicer. Automated technologies allow them to make sure they have all the correct information, send out borrower letters and have staff reach out to borrowers much faster—servicers can even set up these processes with automated dialers and messaging. This saves an enormous amount of time and improves loan retention.
How Momentum Is Growing
Over the past two years, we’ve seen a tremendous amount of pickup in machine learning and data extraction tools in the secondary market. Ultimately, I believe this will bring greater consistency to how the secondary market operates, as well as new best practices. Similarly, secondary market participants that invest in these new technologies won’t simply be able to absorb a great number of loan file types faster and more efficiently, they will also be able to build stronger, more efficient organizations that are better able to compete going forward.
Yet there is a major obstacle that lies in the way that I haven’t talked about. Inertia. There’s an entire generation of secondary market professionals who grew up doing things a certain way and remain loyal to those processes—and the people who perform them.
There is a different way to think about this, as new technologies do not necessarily mean these jobs will go away. If institutions that both originate and service loans no longer need 30 acquisition team members to onboard newly acquired loans, but instead only need five, they can move their staff to the origination side of the business and have more flexibility to scale on either side as market volume ebbs and flows. It’s really a matter of reallocating your human resources to best utilize their skills to improve your business.
Sometime in the future, there will be an inflection point at which participants will either adopt these new emerging technologies or put their companies’ futures at risk. Throughout the business world, there are countless examples of companies that are no longer around because they couldn’t keep up with the pace of accelerating change. It is naïve to think that it can’t happen in the secondary market.
At the end of the day, it’s not a question of if the secondary market will embrace new AI and machine learning technologies, but when. The reduction in time, the cost savings, and the higher quality assets that will result from these changes are too great to ignore much longer. In any event, there will come a time when the predominant technologies used in the secondary market today will be looked at like yesterday’s vinyl records—minus the nostalgia.