Financial risk from flawed appraisals runs into the billions

Serious oversights or inconsistencies appear in one-third of traditional property assessments conducted by human appraisers, resulting potentially several billion dollars worth of financial penalties for lenders. 

Almost 34% of appraisals had a serious unwarranted condition or quality adjustment that wasn't justified after further examination by artificial intelligence tools, according to a white paper from technology firm Restb.ai. The errors subject lenders to repurchases of property-collateralized loans that could cost them between $27.1 billion and $59.7 billion.

Overall, three out of four appraisals showed inconsistencies that could lead to incorrect valuations, the company also said. 

Among issues found by Restb.ai, which specializes in using AI-backed computer vision tools to determine valuation, were a challenging classification system, lack of transparency and human-imposed limits on how outcomes might be reviewed by appraisal management companies that allow errors to slip through. 

"The scale of flawed condition and quality adjustments in appraisals is bigger than most people realize," said Restb.ai chief product officer Nathan Brannen in a press release. "Most AMCs and lenders simply don't have a quick and easy way to check for these issues, so they ignore the problem and hope for the best."

In the study that analyzed over 1,200 completed assessments against almost 6,500 comparable units, researchers noted limitations in the uniform appraisal dataset in grading a home's condition and quality as a contributor to inconsistency. 

Limited granularity in the UAD scale led the overwhelming majority of properties to fall in the middle range of values when looking at a home's condition and quality. 

Even when both a professionally assessed property and AI came in with similar ratings, appraisers still made adjustments on almost 12% for condition and 5% for quality, raising questions about transparency.  

Restb.ai's research also pointed to a 2024 Fannie Mae study, which cited an inadequate selection and adjustments of comparables and inaccurate reporting of appraised homes contributing to inaccuracy. 

Using AI for property assessments

Technology-focused companies like ICE Mortgage Technology and Restb.ai have turned to artificial intelligence to help with valuation, with AI allowing for quick examination of a property's images or photographs.  

"While some appraisers remain skeptical of AI, its value is in its ability to immediately flag potential issues for closer review rather than waiting for discrepancies to be found later in the appraisal process," the paper said. 

Using artificial intelligence in the valuation process  would also help remove some subconscious biases that emerge when a professional appraiser does the work, Restb.ai also said. Another advantage is its ability to consistently analyze properties repeatedly.  

"AI is trained over property imagery independently of that property's price, region, owners, or any other aspect that is more difficult for a human to abstract," the firm said.