Restb.ai Valuation General Manager Tony Pistilli recently shared some of the latest updates pertaining to appraisal technology, including the company’s Computer Vision platform, while offering insight on the efficiency of manual reviews of appraisals, property conditions and renovation estimates.
“So much of the review process is what I call ‘stare and compare.’ meaning a human looking at one part of the appraisal and matching it to another part of the appraisal or an external data source to validate the information on an appraisal,” Pistilli told Valuation Review. “Computer vision allows technology to assist and often replace this function from the review process. Examples of this are looking at the floor plan, then the photos and then the appraisal to validate the number of bedrooms, baths, etc. Computer vision can match all three data points.
“Validating the condition of the comparables requires searching public sites like Trulia, Zillow, etc. and then opening the listing and looking at the photos, if they are still available,” he added. “Computer vision can present an objective and unbiased condition score on the subject property and every comparable used in the appraisal in less than half of a second, thus eliminating all of the searching.”
As to some of the faulty lending decisions that can put lenders and their AMCs at risk, Pistilli told us he is seeing a dramatic increase in overvaluation and undervaluation risk and “appraisal bias” issues that stem primarily from poor comparable selection.
He also noted that computer vision can identify appropriate condition and quality comparables from the market area to assist in letting the user determine if the most appropriate comparables were used in the appraisal.
Pistilli reviewed updates to the Computer Vision platform that includes providing assistance to appraisers, lenders and AMCs.
“Every week we have new use cases being deployed by our lender, AMC and appraiser clients,” he said. “From the manufacturing of an appraisal, uses in data collection to prepopulate forms, to the validation of images using the insights provided during the review process. We are continuously updating our models and adding new models to provide comprehensive solutions to the industry. We identify over 100 damage items, things like peeling paint, cracked foundation walls, mold, water damage and missing cabinetry in the home from the photos provided and have recently released the ability to identify material and immaterial damage.”
But how has streamlining of workflow increased accuracy?
Pistilli responded by noting that computer vision “sees” what everyone sees. And it does it with consistency and objectivity. What used to take several minutes can be done in seconds, often eliminating the need for human intervention, he added.
“We have seen clients reduce review times by up to 50 percent when incorporating computer vision into their review process while eliminating revisions from clients,” he shared.
Pistilli has noticed the cost to store data continues to go down and he does not believe it will be an issue for appraisers, lenders or AMCs going forward.
Pistilli pointed out technological advancement brings unprecedented levels of automation, analytics, and precision.
“I recall doing my first appraisal on a carbon paper form with Polaroid film after looking through books of MLS sales. Since then, there have been many advances in technology but we the appraisers have always been the ones inputting the data,” he told us. “I see that changing and appraisers in the future will be reviewing data provided to them, not the ones inputting or providing the data.
“With the integration of technology, we will soon see appraisers inspecting the property using digital photography, floor plan mapping software and submitting the appraisal from their car using web-based appraisal software within moments after the inspection,” Pistilli added.
In a recent article, Pistilli made a reference to appraisers in the residential financing industry as lending’s “unsung heroes.”
“Appraisers are the only truly independent party in the homebuying or refinancing process. Their responsibility in the process and their knowledge of USPAP is integral to ensuring confidence in the home valuation process,” he said.
“I believe these two things actually work together – as technology such as computer vision gets more ingrained into operational processes, the quality and efficiency both grow together,” Pistilli added, referencing the GSE’s leveraging technology as it relates to the processes. “This works just like doing math by hand or using a calculator; both can arrive at the same outcome, but the calculator will do it much quicker and without error.”
Combining technology and appraisers in the manufacturing process allows a six sigma-like process to be developed that would result in fewer, if any, errors and enhanced data integrity, he added.
The technology can also provide more clarity when it comes to GSE guidelines for evaluating photos.
“As an example, computer vision can detect a picture of a scuttle or peeling paint, and it can alert the appraiser or reviewer that these items need attention. This applies to many of the Fannie Mae, Freddie Mac, FHA and VA requirements,” he said. “Computer vision can improve the quality of third-party data collections by ensuring photos that are blurry, too light or too dark are not delivered to appraisers.
“Computer vision can identify people, lawn signs, religious items and other things related to fair housing that can assist with compliance to the GSE guidelines,” Pistilli added.
The veteran appraiser also shared some details as to how appraisers, lenders and AMCs can balance scale with quality from the technological standpoint.
Technology allows appraisers, lenders and AMCs to scale without losing quality, Pistilli said, and he would actually suggest that quality and efficiency is enhanced from the use of technology.
Computer vision can present comparables in different condition and quality scores, Pistilli told us, allowing appraisers and lenders to access appropriate comparables instantly.
“As an example, comparables can be returned at an ‘as-is’ condition and quality and then at an ‘at-repaired’ condition and quality,” he said. “Additionally, computer vision can help determine average price per square foot values for different markets based on different property conditions, providing up to date, market-specific values for any property. This is a huge time-savings and can also lead to increased accuracy in the appraisal process.”
How does this technology better inform the appraiser as it relates to deeper market trends and insights?
“We have run our models on the past five years of listings. This allows us to track items like the style of kitchen cabinetry, the color of the woodwork, and literally thousands of other items inside and outside of a home,” Pistilli said. “It can also identify the sales price by condition and quality bands in a specific market and so many other items that are not currently able to be tracked by traditional data points. The technology allows the user to identify the prevalence and trending over time of certain features like an ADU, outdoor kitchen, pole buildings in a particular market area at scale.”
We also wanted to know Pistilli’s opinion, as it relates to where the industry currently stands in terms of appraisers truly embracing technology and taking full advantage of the opportunities before them.
“We have only scratched the surface of technology,” he said. “Appraisers are seen as slow to adopt technology, but I believe there are many who have been and will continue to be on the leading edge of technology in appraising.
“I often compare the transformation of auto mechanics that worked on cars in the 1980’s when things were far more mechanical to mechanics today that work on cars that are far more technical,” he added. “Mechanics that never learned the new computers and technical items in cars are no longer working on cars. Appraisers will be much the same. Technology will not replace appraisers, but appraisers that embrace technology will replace those that do not.”