How First American uses automated valuation models to reduce costs and save time

As resistant as the mortgage industry can sometimes seem when it comes to innovation, its growing acceptance of data analytics as a tool for enhancing performance in almost every facet of the business, from marketing to product development, is happening at a rapid clip.

The logical step after a lender or servicer has secured access to both robust data and a crack team to find meaning in it is to start using it to automate key processes, which often leads to better decisions being made and the freeing up invaluable human capital. Through the use of automated valuation models, that’s the direction Jon Wierks is taking First American Financial.

Wierks, who was recently named vice president of data and analytics at First American, took some time to answer MPA’s questions about AVMs, their value, and the role they will play in the next phase of the insurer’s evolution. The following interview has been edited for length and clarity.

Mortgage Professional America: Congrats on the new position. How well will your prior work experiences transfer to FirstAm? What are your thoughts on your new position?
Jon Wierks: Thank you very much! I am very excited about this new opportunity. I have been creating automated valuation models (AVMs) and related analytics products for over 25 years, so this role is right in my wheelhouse. My goal is to provide our customers with a line of top-tier, cutting-edge AVM and collateral risk products. Working with First American’s incredible data science teams and using our databases, we have an opportunity to continue enhancing products already in the market, as well as deliver some exciting offerings in the pipeline.

MPA: Tell us a bit about how automated valuation models work and why you’ve been so interested in working with them over the years.
JW: Property valuation is so interesting because it’s always a moving target and there are a huge number of variables involved. For example, there are the property’s physical characteristics, plus its current condition, geospatial factors that influence value, local/national/global economic indicators, and, at the transactional level, the motivations of the buyer and seller. 

Creating AVMs to estimate the property value never gets old because there are always new modeling techniques being invented or becoming feasible because of advances in technology. AVMs today need to use multiple valuation methodologies to reach competitive performance. These include rules-based emulation, standard regression, and the latest machine-learning techniques. AVMs continue to evolve due to the ongoing enhancements in modeling, as well as the ever-increasing amounts and types of data available. Advances in deriving geospatial, condition, neighborhood and other data will continue to fuel the next generation of AVMs. 

MPA: How will AVMs integrate with the human component at FirstAm? Is there a threat that they might overtake the human component in terms of importance? Have they already?
JW: Appraisers and, for that matter, origination lenders, are already using some form of valuation analytics or AVMs in their analysis of properties today, and this will no doubt continue to expand. AVMs are already an important component of the appraisal and valuation process. The use of AVMs complements the appraisers’ evaluation and can only help in delivering a more accurate and robust assessment of value. As the analysis of public record data, evaluation of recent sales, and the overall logic of the AVMs continue to advance, we’ll continue to see more precise, calculated estimates of value that the industry can use in certain transactions in certain geographical areas with little or no human involvement. 

I don’t foresee AVMs overtaking the human component of the industry. Rather, they can and will be a complement to help people be more productive. We are seeing continued increases in appraisal waiver scenarios and AVMs can play a role as a solution, or part of hybrid products, to help maintain solid collateral risk protocols.

MPA: How big a presence do AVMs have in the mortgage insurance industry right now? Are they something all players in the space have access to or are they more of a growing trend?
JW: AVMs have a natural place in the mortgage insurance industry due to the need for fast and cost-effective ongoing property value monitoring. AVMs provide mortgage insurance firms with the opportunity add another tool for quality control on a one-off basis, but the cost and efficiency of AVMs also provide an opportunity for large portfolio-level analysis and review.

For mortgage insurance companies especially, the ability to quickly, economically and accurately assess the value of collateral they are taking on liabilities for is huge and AVMs have come to play a key role there.

MPA: How have they evolved over the years?
AVMs have evolved into accurate, easy-to-use tools. The first AVMs were often a comp search and calculator tool using public record data only. Today’s AVM has become a sophisticated model using multiple valuation methodologies. Advances in computing technology allow AVMs to deliver virtually instantaneous results using an ever-expanding data set. Geospatial, MLS, appraisal, replacement cost, and derived information, such as condition and quality, are now in play. All competitive AVM offerings are using multiple sub-methodologies based on machine learning, hedonic, statistical and rules techniques.

MPA: What’s the business case for more automation in the mortgage insurance space? What kind of an investment should companies who want to implement AVMs be preparing for, and what exactly will they need to invest in?
JW: I would contend that the drive and business case for more automation extends beyond just the mortgage insurance space and includes all components of the mortgage lifecycle. Industries are always trying to reduce cost and save time, so naturally firms are looking to smart automation as a solution.

One of the benefits to AVM products is not only the value they provide, but additional information that can be gleaned from the report. For example, First American’s AVM leverages our data to deliver relevant property information, market intelligence and the valuation components. Investment companies should look at all data available and ensure it is all being put into good practice.

There are AVM solutions available for any budget – ranging from manual single order from a vendor’s website to APIs that are fully integrated into custom LOS systems. It is key to work with partners that can help maximize the investment and fit the overall strategy. Used properly, AVMs can significantly lower costs and turn-around times for users and, also consumers, while providing a sound tool for analysis.