By Nathan Brannen on 27 May 2021
Determining the value of a property is essential for different real estate, investment and insurance activities. Regardless of whether the goal is to determine an asking price for a property sale, to make an investment decision, or to optimize a home insurance policy’s premium, accurately understanding the value of a home is a challenging task that requires a huge amount of data.
To calculate the best valuation, real estate professionals pull property data from a variety of public and private data sources. However, depending on the purpose of the valuation, the amount of time and effort dedicated to research might vary.
For real estate brokers, it is common to rely on just a few comparable properties when issuing a BPO (broker opinion of value), while appraisers will thoroughly inspect a property first-hand to truly understand its condition, quality and value.
Challenges of Current Valuation Methods
Nowadays, there are countless approaches that use varying types of data to produce the most accurate property valuation. However, despite many advances, valuation processes still face several challenges that are difficult to overcome.
1. Data inconsistencies
Property data is essential to perform accurate valuations. Details like a home’s location, size and age have a huge impact in determining a home’s worth. Unfortunately, collecting all the relevant property information can be difficult as it is often dispersed and scattered throughout varying data sources.
Although many real estate and investment companies have tried to centralize information regarding property features, changes in ownership, and other various insights, there are still many cases where details may differ when looking at two different sources. Making things more difficult, American homes are only sold an average of once every ten years. This makes it virtually impossible for any single company to keep a comprehensive set of up to date information on the 140+ million U.S. residential properties.
This is evident when comparing current Automated Valuation Models (AVMs), which use dozens of public and private data points to calculate a property price. Competing AVMs are constantly searching for reliable data that can improve their models, but obtaining consistent information on all the aspects that influence a property’s price is challenging. The differences in the quality of each AVM’s property info and the design of their algorithms can ultimately lead to vastly different valuations, which is why some real estate websites, like realtor.com, provide users with up to 3 different valuations.
2. Lack of information
Every single property is different and has its own story. For example, let’s take two identical condos, built in the same year, in the same building, with the exact same amenities, number of rooms and square footage. Let’s even say that both have only had one owner since they were originally built. You would think that this would make pricing these properties simple, right?
No matter how complete your datasets are, there are always valuable, yet inaccessible, insights that can tremendously impact the price of a property. Continuing our example from above, we discover the first property has seen tremendous wear and tear over the past 20 years and lacks any of the typical renovations that are common among neighboring properties. On the other hand, the second condo was recently remodeled before its new renter began living there. For many AVMs, the properties would be similarly valued, but to anyone who could open their doors and peek inside, the difference would be immediately apparent.
An AVM's blindspot: Before and after photos of a kitchen renovation
3. Human bias
Property appraisals are by far the most thorough way to understand the value of a property. The Uniform Residential Appraisal Report from Fannie Mae requires appraisers to review a detailed list of items to ensure every aspect of a property is considered in its evaluation. One of the report’s key components requires the appraiser to find three recently sold similar comparable properties to help provide support for the target property’s value.
However, it is quite difficult to find a property’s perfect twin, and even harder to find three. Each property has its own unique features, quality and condition. Even the market conditions when a sale was made can have a large impact on the suitability of using a particular comparable. As a result, it is often necessary for the appraiser to make adjustments to reconcile the differences in prices between the properties.
The difficulty in finding perfect comparables leads to a certain level of bias. In fact, several research studies have shown that appraised values are below the agreed purchase price in only 10% of cases. This means 90% of homes are being appraised either exactly at the sales price or at a higher price point, which implies appraisals are commonly being used to confirm the sales price. Fannie Mae calls this the contract price confirmation bias and is actively looking for ways to help reduce this imbalance and provide more objectivity in home appraisals.
How to improve valuation accuracy using AI
A simple step to help overcome the current objectivity and data issues that valuations currently face would be to consider integrating an AI that can consistently examine a property’s condition and quality.
One of the reasons appraisals are the most accurate method to value properties is that an appraiser visits the property in person. Unfortunately, appraisals are an expensive and time-consuming process that is not feasible for many use cases, such as with AVMs that need to value all the properties in the market. However, using AI would allow companies to analyze property imagery at scale and instantly provide insights on each home’s features, condition and quality. In turn, AVM’s could provide more accurate valuations that would benefit investors, iBuyers and even individual property owners who wish to know how much their property is worth.
Beyond providing condition and quality insights at scale, AI could also help reduce confirmation bias. While we are a ways off from becoming comfortable with AI replacing the human element of an appraisal, it can be an invaluable tool for appraisers to do their job more quickly and efficiently. AI can be used to find more suitable comparable properties or potentially highlight information the appraiser may need to verify and double-check.
Finally, AI could help verify data points and complement data from other sources. This would be beneficial for companies in other industries that also rely on property information, such as home insurance. Currently, carriers fix their insurance premiums by looking at the existing data which is not always as comprehensive as it should be. Having access to this extra AI-generated information could help insurance players optimize policy premiums and reduce their exposure to risk.