As the real estate industry continues to evolve, new technologies have evolved with it. One that has gained significant attention are automated valuation models (AVMs), which provide a value estimate of a property’s worth.
When using a website or app like Xome® to find the worth of your house or to buy a new property, you will see the Xome Value® displayed on the listing. This estimate is generated by our very own AVM, including both the low and high ends of the estimate.
AVMs like ours have become a popular tool for quick and accurate valuations used by lenders, investors, appraisers, and other real estate professionals. But you might wonder — how trustworthy are they?
AVM estimates are only as good as the data available and do not consider the human element of real estate, but that does not mean that you should not trust them. They are not meant to completely replace appraisers and are best used as tools to aid in the appraisal process.
Considering estimates provided by one AVM versus another may differ slightly, so it is important to get familiar with the AVM you are using to understand how the estimates are calculated. The more familiar you are with how AVMs work, the more you will be able to trust them.
1. AVMs are data-driven and objective.
While it may initially seem like the opposite, the fact that AVMs are missing the human influence is actually a benefit for their intended use. AVMs are objective tools for property valuations due to their use of comprehensive real estate data sets, sophisticated algorithms, and standardized methodologies.
They analyze a vast amount of data to come to an objective value estimate, including recent property sales transactions, property characteristics, and market trends, to generate valuations. For example, one of the sources that Xome’s valuation model uses to pull this information includes the Home Price Index for price trends.
Since AVMs do not typically include subjective data such as opinions that often arise from a human-generated appraisal, they can potentially help mitigate bias during the appraisal process depending on the AVM provider and how it is used.
Just like many AVMs, the Xome model doesn’t use any subjective data in its model.
2. AVMs are continuously updated, along with their data.
AVMs rely on comprehensive databases that contain a wide range of real estate data. To ensure that they incorporate the most recent and correct data for property valuations, they are regularly updated with new data from multiple sources, such as public records, tax assessments, and other reliable data providers.
At Xome, our valuation model uses a leading real estate data lake designed to handle large volumes of national data. We use various systems and processes to gather and integrate this data into our data lake, ensuring that the information used by the Xome Value is current and reflective of the market.
Additionally, the Xome team is continuously refining and enhancing our model to improve performance and capture any changes or shifts in the real estate market. By regularly updating both the data and the underlying models, our model strives to provide users with the most accurate and up-to-date valuations possible.
3. AVMs follow specific standards and best practices.
AVMs incorporate specific standards and methodologies that aim to increase their trustworthiness. Among these are the International Association of Assessing Officers’ (IAAO) Standard on Automated Valuation Models for measuring fairness, quality, equity, and accuracy of AVMs. The standard provides basic principles, guidance, and best practices for the development and use of AVMs in real-world property valuations.
Included in the IAAO standard are:
- Data quality best practices: AVMs adhere to specific data quality best practices such as data quality reviews to analyze completeness and consistency of data, as well as outlier identification and removal.
- Quality assurance standards: AVMs are tested to determine if they meet required standards after implementation and before public use. This is accomplished through statistical tests in which value estimates are compared to actual values for the same properties.
AVMs may also be subject to regulatory guidelines and standards set by industry authorities or government agencies. Compliance with these guidelines, such as those outlined by the Uniform Standards of Professional Appraisal Practice (USPAP) in the United States, helps ensure that AVMs adhere to recognized standards of valuation.
4. AVMS go through rigorous testing and validation against actual market transactions.
As a part of the machine learning lifecycle, AVMs are subject to rigorous testing and validation processes. Among these are historical testing and out-of-sample testing.
Historical testing is when AVMs are tested against historical sales data to evaluate their accuracy retrospectively. The AVM’s estimates are compared to the actual sales prices for a large sample of properties across different time periods. This analysis helps assess the AVM’s ability to capture market trends and fluctuations accurately.
Out-of-sample testing is when AVMs are validated against a set of properties that were not used in the development or training of the model. This test ensures that the AVM performs well on new or unseen properties.
AVMs are also tested across different market segments to further evaluate their performance. Market segments may include property types (e.g., single-family homes or condos) or geographic regions (e.g., urban areas or rural areas). This segmentation analysis helps identify any variations in performance based on different property characteristics or locations.
5. AVMs offer wide coverage across various geographic areas.
AVMs use both local and national-level data to provide broader coverage across geographic areas and more accessibility to property valuations in both broad regions and local markets.
They also use comparable sales to identify and analyze recently sold properties that are like the subject property in terms of location, size, and other characteristics. By using this approach, AVMs can provide estimates even in areas with limited data availability.
However, it is important to note that the accuracy and reliability of AVMs may vary based on the availability and quality of data in different areas. Estimates may be less accurate in rural areas where properties are more scattered unlike suburban areas where sale activity is more frequent.
At Xome, our teams perform regular data updates to help capture changes in property values and market conditions across different geographic areas.
While AVMs are useful when providing a quick valuation, they are not perfect. They also come with a few limitations and potential risks.
Since AVMs rely on data accuracy and availability, if the data used by an AVM is outdated, incomplete, or unreliable, it can impact the accuracy of the valuation. Additionally, AVMs may struggle to account for unique property features or factors that require human judgment and expertise.
AVMs can be particularly useful for certain purposes, such as initial screening or providing a rough estimate of a property’s value. But for more complex transactions or properties with unique characteristics, it is often advisable to start with an AVM and consult with a professional who can provide a comprehensive and accurate valuation.
