Nathan Brannen, Chief Product Officer at Restb.ai, is one of real estate’s most experienced artificial intelligence veterans. Restb.ai is the leader in computer vision and AI for real estate. Its artificial intelligence technology automatically tags, classifies, and scores property photos to extract real estate-specific insights. With over a decade of proven experience in AI, Nathan specializes in partnering with industry innovators to implement AI technology.

At Restb.ai, we are only scratching the surface of the potential benefits AI will have on improving valuations. We currently analyze more than 2.500 visual insights on each property.

Nathan began his AI career at ABB, a Fortune Global 500 company in the Power and Automation industry. As Global Product Manager, he managed a suite of AI-based energy models and platforms, working with advanced technologies such as Neural Network-based price forecasting tools, MILP-based optimization solvers, and long-term Capex planning models. He has managed teams across the US and Europe while serving clients globally, including Saudi Aramco, one of the world’s most valuable firms. Nathan holds an MBA from IESE Business School (Barcelona) and a Bachelor of Science from Georgia Institute of Technology.

For those unfamiliar with Restb.ai, how would you describe its role in real estate today?

At Restb.ai we believe there is a wealth of information hidden within property photos. When you look at an image, there are so many important details that you instantly understand. What features does the property have? Is it in good or bad condition? Are the materials high quality? Does it have a particular style or aesthetic? Do you like that aesthetic? Unfortunately, due to the near-infinite nature of things that may appear in images, much of this valuable information remains uncaptured in typical listing details. Or if it exists, it does so in such a way that makes the data difficult to assign a value to. If a property has a patio, how do we determine whether it’s a small, paved landing or an expansive brick patio that could easily host a 20-person gathering? Our goal at Restb.ai has always been to use AI to analyze these millions of images and generate structured, consistent, and valuable insights to be able to analyze and compare properties at scale.

What drove your deeper investment into the valuation side of the industry, and what are you seeing in terms of demand?

Valuations were always a natural fit for the insights we provide. Valuation companies have an unquenchable desire for more data, higher-quality data. When we began working with AVMs, we learned that many had already been trying to create proxies for condition and quality by parsing through listing remarks. Despite the inconsistencies in an approach relying on agents to accurately describe the property they’re selling, they knew that these data points were valuable. We knew we could provide a better solution to a problem they were already trying to solve.

Furthermore, the consequences of inaccurate valuations are high. Incremental improvements in valuation accuracy are essential. As more companies have incorporated these new visual insights into their processes and models, we have seen a bit of a herd mentality from their competitors. Incorporating this type of data is quickly becoming table stakes in the industry.

The modernization efforts from the GSEs have accelerated all of this. As the appraisal process undergoes its most meaningful change since the introduction of the 1004 more than 20 years ago, many companies are confronted with a need to change their processes, particularly in the face of new AI-first market entrants.

You recently launched a Condition/Quality (C/Q) report. What is it, and why does the industry need it now more than ever?

Our recent white paper focused on whether the value adjustments made for condition and quality in appraisals are reliable and consistent. It analyzes how common it is for an appraisal to contain unwarranted adjustments and how frequently adjustments should have been made but were omitted. We felt this was timely for three main reasons:

• Unlike more objective property characteristics such as living area, lot size, and construction materials, condition and quality are inherently subjective and complex.
• According to the GSE’s most recent updates, issues related to condition and quality are the most frequent issues found in appraisals.
• In the high-profile bias lawsuits that have recently been in the headlines, the comparables used to justify the property valuation were of vastly different condition and quality compared to the subject property.

Until our report, this has been a difficult concept to quantify and something we felt many in the industry were choosing to ignore.

What are the C/Q report’s biggest takeaways for appraisers and the appraisal industry?

Simply put, more than 33% of appraisals have a high risk of either having an unwarranted condition or quality adjustment or omitting an adjustment that the photos would indicate is needed for at least one comparable property. These issues could easily lead to a repurchase request, which has an estimated cost of $32,288 to the lender. Considering all the appraisals that occur each year, this means there is a potential risk of more than $27 billion in repurchase costs.

How can lenders, AVMs, and valuation providers benefit from the findings of the C/Q report? What do they need to do next?

In many ways, I understand why these issues are so common. Consistently analyzing condition and quality can be difficult. Properties may be in varying states of renovation, making it challenging to consistently capture the differences. Furthermore, our analysis showed 81.1% of properties were deemed a C3 or C4, and 97.5% of properties were scored as a Q3 or a Q4. Given all the other considerations appraisers are looking at when selecting comparables, defaulting to a common score may be convenient.

Furthermore, these issues are incredibly time consuming for either AMC or lender review processes to catch. An appraisal only includes a front photo of each comparable property. For a reviewer to validate that condition and quality adjustments are appropriate, they would need to manually search for each property on a portal and scroll through dozens of photos. As such, many companies have admitted they don’t have a consistent process to look for these inconsistencies.

Fortunately, AI provides a game-changing opportunity to automate this process. Rather than asking reviewers to look at the photos of every comparable property, they can rely on AI to do a first pass and only analyze the properties with a potential issue. Given that one-third of appraisals have a potential issue, and the average appraisal has five comparable properties, this automation reduces the number of properties a reviewer needs to look at by over 90%.

Even more importantly, catching these issues in the first review reduces turn times and decreases repurchase costs.

Appraisal bias and modernization are huge industry concerns. How does your Restb.ai technology directly support these industry pain points?

Bias is a complex topic in the appraisal world. Many appraisers are understandably offended by being called racists or anything less than “Objective, Independent, and Unbiased”, as required by USPAP, while at the same time, there have been several high-profile cases that point to valid bias concerns. However, I think it’s much simpler to look at how our technology can stop bad appraisals. As mentioned before, the GSEs have highlighted that condition and quality issues are some of the most common challenges they encounter. These issues have been at the center of some of the recent bias cases on how the appraiser landed on a “biased” valuation. With AI, these discrepancies in condition and quality would be flagged automatically, preventing this from ever becoming a conversation about bias.

When looking at appraisal modernization, one of the key differences in the new report is the requirement to break out condition and quality into an exterior and interior score. This reflects the importance of these components in a home’s value. Still, it does not solve the subjectivity in determining these values, nor the likely adjustment period it will take for appraisers to learn how to assess them consistently. Fortunately, our technology already breaks out condition and quality into these more granular components, providing robust guardrails to ensure accurate valuations.

What are the longer-term implications for the appraisal and valuation industry on the kind of data that AI can unlock, like the findings in the C/Q report?

At Restb.ai, we are only scratching the surface of the potential benefits AI will have on improving valuations. We currently analyze more than 2.500 visual insights on each property. Many of these insights are not included in listings, public records, or appraisals, but that doesn’t mean they don’t have an impact on value.

In fact, having worked with many AVM companies, we have seen that the presence or absence of certain features can have a quantifiable impact. For example, how do homes with galley kitchen layouts compare to more contemporary open layouts? What is the price difference when someone repaints their kitchen as opposed to completely replacing the cabinetry? Now that this data can be tracked and analyzed at scale, we can determine the impacts of all these questions and more importantly, how those impacts vary depending on which market you’re in. As more and more companies utilize this data, the better we will truly understand how to value properties accurately.

The post Q&A with Nathan Brannen, Chief Product Officer at Restb.ai appeared first on citybiz.

This contant was orignally distributed by citybiz. Blockchain Registration, Verification & Enhancement provided by NewsRamp™. The source URL for this press release is Q&A with Nathan Brannen, Chief Product Officer at Restb.ai.