Big Data: Changing the Mortgage Business
From online shopping to social media posts, the data trail consumers leave behind these days only grows longer and more telling. In fact, data scientists estimate that 90 percent of the world’s data was generated over the last two years alone, according to a 2018 article published by Forbes. What does this mean for mortgage lenders?
While the era of big data is still taking shape, no one can dispute that it’s arrived. It’s expected to transform the mortgage industry because of the insights it reveals about borrowers who are difficult to underwrite. This technology can accelerate the loan production cycle, improve risk management, render mortgage servicing more efficient, and extend the reach of marketing campaigns among other benefits.
Big Data, Big Deal?
Big data can mean many things, but technologists generally define it by the four Vs—massive in volume; extremely varied in format; high speed in velocity; and complex in terms of veracity. For mortgage lenders, it can include customer records (loan files, bank statements, brokerage accounts) and third-party information (credit scores, tax returns, credit card statements, or even cell-phone payment histories).
Related terms, such as big data technology or big data analytics, refer to the way this information is captured, organized, and made searchable. Machine learning—a subset of artificial intelligence— uses algorithms to analyze big data, draw conclusions from it and update these conclusions as new data becomes available.
By using the insights gleaned from big data, lenders can learn a lot more about borrowers with thin credit files—meaning people who haven’t tapped enough credit to be judged on just a generic credit score. For example, a lot of millennials don’t take out car loans, use credit cards, or work as salaried employees the way their parents did. However, they do have bank accounts, pay cell phone bills, and often use mobile payment apps—all of which can help lenders to decide whether they’re a good credit risk.
In the same way that lenders can build alternative credit profiles for millennials, they can also do this when evaluating mortgage applicants from underserved communities, many of whom lack definitive credit histories.
A digital lending platform isn’t just an online mortgage application form, but rather an interactive experience that meets or exceeds consumers’ expectations. Would-be borrowers will quickly leave a lender’s website if it asks them too many questions only to recommend generic loan products. Big data usually draws on internal analytics and third-party information to come up with the best options almost immediately. If prospective customers begin shopping for loans on a lender’s web-based platform, this initial interaction can accelerate the sales process, with loan officers or sales representatives following up with a phone call.
In the face higher interest rates, declining volume, and high production costs, lenders need to control costs and preserve profit margins. They can integrate big data analytics into existing systems to digitally process applications, speed up underwriting, and better onboard customers. For example, with a customer’s consent, a lender can more efficiently gain a broader picture of borrower financials with third-party data providers, such as employers, banks, brokerage firms, and credit bureaus.
In addition to enhancing data integrity, machine-learning models can help lenders prevent last-minute delays by flagging a data point that requires further investigation. For example, if the system uncovers a large deposit or withdrawal in a borrower’s bank account, the processor or underwriter can request clarification via an account status alert, with the customer’s answer feeding into the analytics application assessing her overall credit risk.
The speed and efficiency of approving and closing a loan directly affect production and underwriting costs. With more comprehensive, better-organized, and easily searchable data loan processors can deliver higher-quality files more quickly to underwriters. Underwriters can focus on automated, flagged exceptions rather than having to discover them with “stare and compare” work. This can help shave days off the lending cycle timeline.
Maximizing Servicing Returns
Servicing is a highly commoditized, volume-driven, low-margin business. With big data analytics, servicers stand to maximize collections and better control costs by helping them to identify borrowers on the cusp of missing an upcoming payment and optimize customer outreach to reduce delinquencies.
Within this new analytical paradigm, servicers can supplement existing customer data with borrowers’ current credit card balances or payment status on student debt or car loans. Machine-learning models rate these variables against credit risk scoring methodologies to determine the most effective, cost-efficient means of working with at-risk borrowers.
When it comes to predicting actual delinquencies and defaults in a servicer’s portfolio, a model is only as good as the data supporting it. Today, the volume of relevant data exceeds the storage capacity of traditional warehousing systems. However, with big data technology, servicers can run machine learning driven models in a quicker span of time without consuming additional computing power.
By tapping into big data, lenders can gain additional insights into borrowers beyond what they learn from credit scores and tax returns. They can get a better, more recent picture of borrowers with credit card transactions and income fluctuations reflected in their bank accounts. Once they aggregate this data, it’s easier to more precisely segment customers and market products best suited to their needs at any point in their lives. Models powered by machine-learning algorithms identify correlations or reveal hidden trends to help a lender extend the reach of marketing campaigns to generate repeat business or win new customers.
With their shift to digital services, lenders must stay on top of fraud, especially since the mortgage industry is the most frequently targeted sector in financial services. At the same time, they don’t want to lose legitimate business or run afoul of regulators by rejecting applications too aggressively. Fortunately, big data analytics can help to balance these competing goals.
Fintech vendors, lenders, and third-party data suppliers are moving beyond traditional detection measures dependent on siloed data and manual processes to distinguish actual fraud from anything flagged as suspicious activity. Their technology can help lenders to minimize false positives and spot questionable transactions within seconds, and then score risk against thousands of variables.
Embracing this technology often requires a new managerial mindset toward risk, and such analytics are not an overnight fix. However, they can yield significant benefits within a reasonable time period and reduce the costs of solely relying on legacy systems.
Managing compliance risk is a herculean task due to the matrix of mortgage lending regulations and housing finance programs that the government put in place after the 2008 financial crisis. Lenders now answer to a lot more government entities that want information sliced and diced differently depending on what part of the business they scrutinize. With the help of big data, lenders can integrate systems and models—which have been siloed in different departments—to speed up compliance and reporting and upgrade reporting quality.
To promote regulatory compliance, lenders have traditionally managed their day-today business by focusing on one loan application at a time, as it’s processed through the mortgage cycle. Operations departments are tasked with feeding application data into systems that are then maintained by risk management. Machine-learning models can draw upon massive volumes of data delivered in real time to test a portfolio against a broad range of compliance criteria, and more accurately predict problems before regulators point them out.
Analytics and Big Data in Action
Every lender knows that selling to existing customers is less costly than acquiring new ones, but success depends on improving the borrower experience. In early 2018, a Minnesota credit union used big data analytics to target only 1,400 members after calculating the amount of money they’d save by converting into short-term mortgages. By refining its target market, the lender maximized its marketing dollars and wrote nearly $30 million in new loans.
Using third-party data for income and asset verification, instead of requiring documents from borrowers directly, is increasingly common to ease the documentation demands on borrowers. In March, a Georgia-based nonbank mortgage lender integrated an application through which it receives employment status, income, and W2s from a major credit bureau into its loan origination system. It reported that within six months, it had automatically validated borrower financials for about three-quarters of some 25,000 loan applications, valued at $6.5 billion. In doing this, the lender said it sped up its pipeline by about a third and cut closing times by about five days.
Cost containment is the biggest challenge in the servicing business. An industry leader headquartered in Texas is using big data analytics to boost the productivity of its customer call centers. With more robust cost-benefit analyses, it’s helping employees make the most of the time they spend talking to customers to prevent delinquencies and defaults. Using output generated by machine learning models, the servicer has determined the best day and time to call customers whose payment activity deviates from prior months. For example, employees might wait a week to call a customer who has a high FICO score and regularly makes a direct deposit on the 12th of the month for payment due on the 15th. However, they might phone a borrower a day after her balance is past due if she has a history of late payments.