Mohammad Rashid: Use of Big Data, AI in Mortgage is Nowhere Near its Full Potential

0

PERSON OF THE WEEK: Big data, automation and artificial intelligence (AI) are revolutionizing the mortgage industry – but it would seem that mortgage lenders have barely cracked the surface in terms of what these technologies can do for business.

Big data, in particular, is now starting to play a key role in helping mortgage lenders predict to consumer behaviors, enabling them to, in effect, “know” when a consumer is ready to purchase a home or refinance an existing one.

But there are a myriad of other ways that big data and AI can potentially improve mortgage lenders operationally – not just in terms of “winning customers for life” but also in terms of gaining new back office efficiencies.

To learn more about the potential that big data and AI hold for the industry, MortgageOrb recently interviewed Mohammad Rashid, vice president, head of consumer lending and capital markets practice at financial services technology firm Tavant Technologies.

Q: Artificial intelligence (AI) is a hot topic in the mortgage lending market currently. Describe how AI can be leveraged in consumer lending and how it can be used to address some of the classic mortgage challenges?

Rashid: Artificial intelligence (AI) and machine learning (ML) are having a significant influence across all industries. From “intelligent” robotic process automation and speech recognition to virtual agents and driverless cars, the extent of its impact has moved us from a mobile-first to an AI- first world.

The promise of AI/ML to transform the lending industry is beginning to unfold as key players across the consumer lending landscape are leveraging the technology to streamline operations, increase efficiencies, enhance the borrower experience and improve loan quality and risk alignment. 

Consumer lending, at its core, is a data-rich environment. For example, in a typical mortgage lending scenario more than five thousand data attributes are captured including the borrower’s credit history, assets and liabilities, employment, income, tax, and insurance information as well as detailed information on the property or collateral in question.

This is a time consuming and expensive process – and, in the case of many lenders, an extremely manual and labor-intensive process. Once the data is captured, it is difficult to predict (without using AI/ML techniques) how much of this data is relevant to the mortgage or loan product in question or how useful is the data in predicting borrower behavior during the application processing, closing, funding, post funding and servicing stages.

As AI/ML come into use more frequently it will influence the end-to-end borrower journey from home shopping to home buying to home ownership. AI/ML can be used to score and prioritize leads, to match prospective borrowers with the right-size home and affordable mortgage product, to predict propensity to close and convert, to highlight and detect anomalies for underwriters and eventually be able to predict borrower default propensities within servicing. 

Take, for example, the classic mortgage problem of servicing portfolio retention and how lenders approach it today. According to a recent report from Black Knight, servicers retained less than 18% of the loans in their portfolio that refinanced in the first quarter of 2019. This is the lowest retention rate since the company began tracking rates in 2005.

This is not surprising when one considers that some lenders still use credit triggers as the primary way to identify borrowers who are going to refinance away. It’s like trying to put a saddle on a horse without realizing that the horse is not only not in the barn but on a different farm altogether.

AI/ML techniques applied to the portfolio retention problem allows servicers to not only predict specific customer behavior three to six months before they take the first step but also couple it with the most attractive product and program offering that turns them into a customer-for-life.  

Some of the more common examples of how AI is addressing the traditionally classic mortgage challenges include better end-to-end borrower journey and increase in customer satisfaction; prioritization and scoring of leads in the application pipeline; propensity modeling (to close or to fallout) of borrower application; and better matching of borrower to right-sized home to affordable loan product and program.

Other benefits include faster, more accurate and more efficient underwriting; fewer losses from fraud; better automated valuation models (AVMs) for the property in question;  more accurate pipeline management and measure of fallout risk; and reduced credit losses by more accurate prediction of EPDs and general loan defaults.

Yet more benefits include better servicing portfolio retention and leakage defense; decreased loan repurchases by detecting defects and anomalies upfront; and better hedging in secondary marketing through a more accurate pipeline management.

Q: Rising consumer demand for digital mortgages presents a significant opportunity to lenders. What advice would you offer to lenders who want to ensure they take advantage of this opportunity and remain relevant for years to come?

Rashid: Borrowers want speed and ease, which requires moving from paper-based processes to digital origination and servicing capabilities. Digital Transformation is drastically impacting the mortgage process, and it is imperative for lenders to stay updated with these changes and adopt them proactively. Technology is no longer a roadblock and today’s customers are very receptive to digital offerings across the financial board. Consumers want convenient, secure solutions that meet their lending needs.

To remain relevant, improving the customer experience is paramount. Positive customer experiences can profoundly impact a lender’s growth. Providing a superior customer experience is a long-term competitive advantage that lenders can and should leverage to differentiate themselves from the competition. This can be daunting, but revitalizing processes to provide a modern, enhanced customer experience is perhaps the most important piece of advice one can offer today’s lender.

Borrowers perceive home-buying as a single transaction and anticipate that all stakeholders will work together effectively and seamlessly. Lenders that effectively serve as the central point to orchestrate the overall transaction can position themselves as trusted advisers and improve the overall customer experience.

For example, I recently completed a refinance with a lender who had implemented the front-end digital journey very well but fell short on the back-end operational fulfilment. I ended up having to hunt down multiple documents that I would have to upload during the processing and underwriting cycle. My app to funding time was still above 30 days because they could not substitute an AVM for the manual appraisal process. The appraisal process itself was disjointed from the rest of the journey. I still had to pay upwards of $1500 for a title search and insurance services when the home was built in the last decade and I was the first and sole owner of the property.

For the mortgage world as it stands today, to be able to do straight-through processing, all data sources that are required for the loan manufacturing process (and specifically the underwriting process) need to go through a digital transformation, which means they are freed from the trappings of paper and are accessible through APIs.

The appraisal and title worlds need to be disrupted in order for cycle times to come down and originations costs to decrease considerably for both lenders and borrowers. Without all these pockets of inefficiency getting a dose of constructive disruption, the industry will not see complete innovation in the overall home shopping, buying and owning experience.

Q: What steps can lenders take to ensure they are able to advance in the new digital era?

Rashid: Compared with other industries, the lending industry is slow in the process of transitioning from legacy platforms to data-driven, digital environments. However, there are a few immediate steps that Lenders can take to get them started on the road to a transformative journey. 

First, Lenders need to step back and understand their customer markets and segmentations better. This includes the end-borrower as a customer but also the broker community as customer for wholesale lenders and the correspondent seller as customer for the correspondent lenders.

They need to uncover the personas that appear repeatedly within their customer groups (the digital native Dorothy vs. the first-time Frank vs. the self-reliant Susan) and define the personalized journeys and offerings that they would present to them. This is important for lenders to truly understand their competitive landscape across all channels and to apply the principles of digital transformation with the goal to stand out and not blend in with the rest of the field. 

Second, Lenders need to analyze their channels of operation. For example, in the retail channel, they need to address loan officer compensation to dramatically reduce the cost of the loan (and continue to be a viable firm) but not necessarily by replacing the loan officers but by increasing their efficiency and productivity by supplying them with digital tools. The ability to score leads using AI/ML approaches or inject timely advice to a prospective borrower during the application lifecycle or focus in on the borrowers most likely to fallout post-lock are important characteristics of a true digital retail lender.  

Third, Lenders need to review their technical platforms for all channels; retail, wholesale and correspondent. The review should identify the ability of the platform to be flexible and extensible in offering-up personalized journeys and experiences to customers, not only in the application intake process but also in underwriting, closing and settlement services as well. In addition, the platform should support data-driven capabilities that allow for automation across the loan lifecycle. 

Fourth, Lenders need to implement a consistent and holistic data strategy. They will not be able to compete in the new world order if they do not have strategies and approaches to collect (and obviously secure) detailed data related to their borrowers (credit and capacity) and the journeys they take with the lender, the collateral (the property characteristics) and the eventual disposition of the application, and the subsequent journey of the loan through secondary marketing and the sale to investors. This data if mined and utilized correctly can be the guiding factor to becoming a true digital lender with a profitable balance sheet. 

Finally, Lenders need to unify all of the above elements (customers, channels, platforms and data) together into a cohesive well-oiled machine that is able to churn out loans with the highest of velocity (seven days to a close), lowest of manufacturing costs (can we get back to the $2000 level?) but the highest of customer satisfaction ratings.   

In one perspective, “Winter is coming” to the mortgage industry and lenders must be prepared to quickly innovate across all spheres of influence to survive. The innovation must happen across internal-facing processes, external-facing borrower experiences – even the nature of mortgage products and the type of technology (big data, AI/ML, RPA, blockchain and microservices) they deploy in order to appeal to the next generation of borrowers (the digital natives) and possibly fend-off the large internet giants that are waiting  to come into this market with their mountains of customer behavior data and their AI/ML weapons in hand.  

Q: What is the biggest factor impacting originations right now? Is it market-based, technology, regulatory, or something else?

Rashid: The main factor impacting the ebb and flow of originations is the interest-rate environment, so we saw a sharp decrease in refinances and low volume of purchases in the last couple of quarters as the healthy economy inched the interest rates higher and the housing supply and demand equation led home prices higher. Then we saw spikes of reversals in this trend as interest rates dipped and hovered around the 4.0% level. We also see technology investment in innovation being impacted with the regulatory environment as lenders and LOS vendors start focusing on migrating their application intake process, their storage schemas and data communication formats to the new URLA this year.

However, technology is a significant factor in today’s market.  According to Fannie Mae, 63% of lenders are familiar with some form of AI or ML technology. Recently, we’ve experienced a surge in the number of financial organizations moving towards AI and ML methodologies to automate a wide array of functions within the mortgage lifecycle.

Q: Looking into your crystal ball, what do you see being the next big technological advancement in the mortgage industry? Where do you see the industry in five years?

Rashid: The nature of the mortgage space today doesn’t lend itself to large paradigm shifts and big bangs in innovation. Although, there are green shoots in the private label market, the agencies continue to dominate originations and there are guidelines that must be followed to satisfy loans originated for them. There have been incremental improvements through Fannie’s Day 1 Certainty® program and Freddie’s’ Edge program and the development of a digital borrower journey in the application intake process, but lenders still require a lot of data post-application submission that are still trapped in the world of paper and the world of manual human intervention. 

After the initial push to innovate by digitizing the borrower journey, the application intake process and the sourcing of asset, income and employment data, we see a wider acceptance of e-mortgages taking place in the next few years with its accompanying cornucopia of e-closing, e-notes, e-notarization, e-recording, e-registry, e-vaults and e-delivery leading to a more highly streamlined closing and settlement process. 

Eventually, we see the silos and walled gardens of the lender’s origination world, servicing world and secondary marketing world come down. Today, they are based on separate systems of record, with data being manually uploaded from one system to another with all their attendant errors and latencies. Borrowers and loan officers update data on the origination system on the front-end of the process that directly affect loan pricing and eligibility, but the secondary marketing department doesn’t get to know about it until much later. Stale or “ghost” loans in the origination pipeline end up being actively hedged on the secondary side with attendant leakage or losses in profitability.

The same applies to the ebb and flow of loans and MSRs in the servicing system that is distinct from the secondary system that hedges both. These points of inefficiency will need to be addressed by real time updates and tighter integrations and potentially a holistic loan manufacturing platform that addresses processes from lead and borrower acquisition to the assembly of the loan to the sale of the loan to investors and investor satisfaction.     

Some of these changes are happening incrementally today. We see interesting innovation paths emerging regarding which lenders should increasingly be aware. Figure’s HELOC product is an interesting example that brings together digitally sourced loan data at origination with a blockchain backbone and judicious changes in the decisioning criteria of the product itself such as substituting an AVM for an actual appraisal that streamlines the overall process to minutes for decisioning and days for funding. All stakeholders including investors are able to access real-time data on the same platform at the same time leading to customer satisfaction across the board. 

I believe these types of innovations will incrementally push the mortgage industry towards true end-to end automation. That and straight-through processing are what thrusts us towards a pragmatic consumer-oriented digital future.

Subscribe
Notify of
guest
0 Comments
newest
oldest most voted
Inline Feedbacks
View all comments