Why Appraisal Reviews Need Better Property Intelligence

For lenders, AMCs, and valuation professionals, reviewing property condition and quality is an important part of appraisal quality control. But assessing these attributes consistently across a subject property and its comparable properties can be difficult, especially when reviews need to happen at scale.

Last year, Restb.ai published a white paper examining the reliability of appraisal condition and quality adjustments and their potential impact on appraisal accuracy. The study analyzed 1,271 appraisals and 6,495 comparable properties, using AI-powered analysis of property imagery to evaluate condition and quality.

The research found that 33.6% of the appraisals analyzed were classified as high risk for an improper condition or quality adjustment. In other words, the study identified a significant number of appraisals where the condition or quality adjustments applied to comparable properties could potentially be unsupported by the properties' visual characteristics.

That finding becomes particularly significant when considered alongside mortgage repurchase risk. According to a study by Reggora cited in the white paper, the average mortgage loan repurchase rate was 0.49%, with an average cost of $32,288 per repurchased loan.

Using a conservative estimate of 2.5 million appraisals completed annually, the white paper estimates that the 33.6% high-risk figure could represent $27.1 billion in potential repurchase risk.

So, where does this potential risk come from, and what did the research find?

That's what the white paper set out to explore.

Read the Full White Paper: The Impact of Condition and Quality on Appraisal Accuracy

Why Property Condition Can Be Difficult to Compare

Condition and quality are important components of an appraisal, but assessing them consistently across a subject property and its comparable properties can be challenging.

The UAD six-point scale provides a standardized framework for describing property condition and quality. However, the research found that ratings are heavily concentrated around the middle of the scale. In the Appraisal-Level Public Use File, 81.1% of properties were classified as C3 or C4, while 97.5% were classified as Q3 or Q4.

This concentration means that properties with potentially meaningful differences in their physical characteristics can still fall within the same UAD category. As a result, determining whether a difference is significant enough to justify an adjustment can be challenging.

The whitepaper describes this as a challenge of translating a property's condition and quality, which exist on a spectrum, into a single categorical rating.

Where AI Can Add More Granularity in Appraisal Reviews

One opportunity explored in the whitepaper is using AI-powered computer vision to analyze property imagery and provide more granular condition and quality assessments.

Instead of relying only on whole-number UAD ratings, AI-generated scores can provide decimal-level assessments, helping identify smaller differences between properties. The research also examines individual property components, including the kitchen, bathrooms, interior, and exterior.

This additional level of detail can help make differences between properties easier to identify during appraisal review.

When Adjustments May Not Tell the Full Story

The study also looked at whether condition and quality adjustments applied to comparable properties were supported by the differences identified through AI analysis.

The research identified potential issues in both directions: adjustments that may be missing when differences exist, and adjustments that may be applied when the underlying property differences do not support them.

One example in the study shows an appraisal where the subject and comparable properties were all appraiser-rated C3. However, AI-generated scores showed differences in condition across the properties, while also highlighting how similar some of the properties were. Condition adjustments were applied to two comparables, and the whitepaper classified this scenario as a high risk of potential undervaluation.

How AI Can Support Appraisal Quality Control

For lenders and AMCs, manually reviewing condition and quality across every subject and comparable property can be difficult to do consistently at scale. Comparable properties may also have limited imagery, making it challenging for quality control teams to validate every reported condition and quality assessment.

The research highlights an opportunity for AI-powered analysis to flag potential condition and quality issues for closer review. Rather than replacing professional judgment, AI can provide an additional layer of property intelligence to help valuation teams focus their attention where it may be most needed.

This can help teams:

• Identify potential inconsistencies between properties
• Surface missing or potentially unwarranted adjustments
• Prioritize files for closer review
• Add more consistent property intelligence to quality control workflows

Looking to explore more AI-powered solutions for appraisal and inspection workflows? Explore Restb.ai's AI solutions for appraisals and inspections or Get in touch with our team.

Explore the Full Research

This article provides a brief look at the findings. The full whitepaper includes the study methodology, AI scoring approach, adjustment analysis, risk thresholds, and detailed appraisal examples.

Want to go deeper? Read the Full Whitepaper: The Impact of Condition and Quality on Appraisal Accuracy