Latest Headlines

PropTech Valuation Models: AVMs, Data, and the Future of Property Appraisal

An educational guide to automated valuation models and the data that powers them, examining how algorithmic property valuation is changing the appraisal process, the accuracy and limitations of AVMs, and the evolving role of the human appraiser

Automated valuation models use algorithms and large datasets to estimate property values instantly, reducing the cost and time of the traditional appraisal process. This educational guide explains how AVMs work, the data they rely on, their accuracy and limitations, and the evolving relationship between algorithmic valuation and the human appraiser in the mortgage lending process.

Ebonie Beaco, Publisher & Mortgage Strategist
Ebonie Beaco

Mortgage Strategist | Senior Mortgage Loan Officer

Nov 6, 2026 8 minutes
PropTech Valuation Models: AVMs, Data, and the Future of Property Appraisal

Automated valuation models, commonly referred to as AVMs, are algorithmic systems that estimate the market value of a residential property by analyzing large datasets of property characteristics, recent comparable sales, market trends, and other relevant data, producing a value estimate in a fraction of the time and at a fraction of the cost of a traditional human appraisal. The AVM has become one of the most widely used and most impactful technologies in the PropTech ecosystem, serving as the valuation engine behind the instant property value estimates that consumers access on platforms such as Zillow and Redfin, the valuation tool that lenders use for portfolio monitoring and for appraisal waivers on certain mortgage transactions, and the pricing algorithm that underlies the iBuyer instant offer model. The proliferation of AVMs has fundamentally changed the landscape of property valuation, creating new efficiencies and new capabilities while also raising important questions about accuracy, transparency, and the appropriate role of algorithmic valuation in the mortgage lending process.

The methodology that underlies an automated valuation model typically involves a combination of statistical and machine learning techniques that are trained on historical transaction data to identify the relationships between property characteristics and sale prices, and that are then applied to the subject property to generate a value estimate. The most common approach is the hedonic regression model, which estimates the contribution of each property characteristic, such as square footage, lot size, number of bedrooms and bathrooms, age, and location, to the overall property value, and then applies the estimated contributions to the subject property characteristics to generate the value estimate. More sophisticated AVMs incorporate additional data sources, including tax assessment records, prior sale histories, neighborhood demographic data, school district information, and macroeconomic indicators, and they use advanced machine learning techniques such as gradient boosting and neural networks to capture non-linear relationships and interaction effects that simpler models may miss.

The comparable sales analysis that is the foundation of most AVMs mirrors the approach that a human appraiser uses, with the model identifying recent sales of properties that are similar to the subject property in location, size, age, and characteristics, and adjusting the sale prices of the comparable properties to account for the differences between each comparable and the subject property. The AVM automates the comparable selection and adjustment process, applying algorithms to identify the most relevant comparables from the transaction database and to calculate the adjustments based on the estimated contribution of each differing characteristic. The advantage of the automated approach is that it can process a much larger set of potential comparables than a human appraiser, and it can apply adjustments consistently and without the subjective bias that can affect human judgment. The disadvantage is that the automated approach may not capture the qualitative factors that a human appraiser observes during a property inspection, such as the condition of the property, the quality of the finishes, or the impact of neighborhood features that are not captured in the data.

The data infrastructure that supports AVMs is a critical determinant of the accuracy and the reliability of the value estimates that the models produce, as the models are only as good as the data on which they are trained and on which they operate. The primary data sources for AVMs include county recorder and assessor data, which provides property characteristics, ownership records, tax assessments, and recorded sale prices, and multiple listing service data, which provides detailed property descriptions, listing histories, and sale prices for properties that were listed and sold through the MLS. The quality, completeness, and timeliness of this data vary significantly across jurisdictions, as the data collection and reporting practices of county recorders and assessors differ from one county to another, and the MLS data coverage varies by market. AVMs that operate in markets with high-quality, comprehensive, and timely data generally produce more accurate value estimates than AVMs that operate in markets with sparse, incomplete, or delayed data.

The accuracy of automated valuation models, as measured by the median error rate between the AVM estimate and the actual sale price of properties that subsequently sell, varies significantly across markets, property types, and market conditions. In markets with homogeneous housing stock, high transaction volumes, and stable market conditions, the median error rate of a well-calibrated AVM can be as low as 3 to 5 percent, meaning that half of the AVM estimates are within 3 to 5 percent of the actual sale price. In markets with heterogeneous housing stock, low transaction volumes, or rapidly changing market conditions, the median error rate can be 10 percent or higher, and the error rate for individual properties can be substantially larger, particularly for properties that are unusual, that have features not well represented in the training data, or that are located in areas with limited comparable sales activity. The error rate is a critical consideration for any application of the AVM that has financial consequences, as a 5 percent error on a $500,000 property represents a $25,000 valuation error.

The use of AVMs in the mortgage lending process has expanded significantly as lenders and the government-sponsored enterprises have sought to reduce the cost and the time of the traditional appraisal process, which typically requires a licensed appraiser to inspect the property and to prepare a detailed appraisal report, a process that can take one to two weeks and that costs several hundred dollars per loan. Fannie Mae and Freddie Mac have implemented appraisal waiver programs that allow certain loans to close without a traditional appraisal, relying on the AVM value estimate and other risk factors to determine eligibility for the waiver. The appraisal waiver programs have been expanded over time to cover a broader range of transactions, and the GSEs have reported that the waiver programs have reduced the appraisal cost and the cycle time for eligible loans while maintaining acceptable loan performance. The expansion of the appraisal waiver programs has been one of the most significant changes in the mortgage appraisal landscape, and it has been enabled by the demonstrated accuracy and reliability of the GSEs proprietary AVMs.

The hybrid appraisal, which combines elements of the traditional appraisal and the AVM, represents an intermediate approach that is gaining traction in the mortgage lending industry as a way to reduce the cost and the time of the appraisal process while maintaining a higher level of accuracy and property-specific assessment than a pure AVM provides. In a hybrid appraisal, a property inspection is conducted by a third-party inspector who is not a licensed appraiser, and the inspection data, including photos, measurements, and condition assessments, is combined with AVM data and comparable sales analysis by a licensed appraiser who prepares the final appraisal report without personally inspecting the property. The hybrid approach reduces the cost of the appraisal by using a lower-cost inspector for the property visit, and it reduces the cycle time by allowing the appraiser to focus on the valuation analysis rather than the property visit. The hybrid appraisal is seen by many in the industry as a pragmatic evolution of the appraisal process that leverages technology while preserving the role of the licensed appraiser in the valuation.

The limitations of AVMs are well documented and are an important consideration for any user of the technology, as the models have inherent constraints that affect their applicability and their reliability in certain situations. The reliance on historical transaction data means that AVMs may lag in rapidly changing markets, as the most recent comparable sales may not yet be reflected in the data, and the model may not fully capture the current market dynamics. The reliance on property characteristic data means that AVMs may not accurately value properties with features that are not captured in the data, such as recently completed renovations, exceptional condition, or unique architectural features. The reliance on comparable sales means that AVMs may struggle in markets with few transactions, as the statistical basis for the value estimate is weaker when the comparable pool is small. These limitations do not make AVMs unreliable in general, but they mean that the AVM value estimate should be evaluated in the context of the property, the market, and the intended use of the valuation.

The role of the human appraiser in a technology-enabled valuation environment is evolving, as the expansion of AVMs and hybrid appraisals changes the volume and the nature of the traditional appraisal work that is available. The human appraiser continues to play a critical role in the valuation of properties that are not suitable for AVM valuation, including complex properties, unique properties, properties in markets with limited comparable data, and properties involved in transactions where the lender or the buyer requires the level of property-specific assessment and professional judgment that only a licensed appraiser can provide. The human appraiser also plays a role in the quality control and the audit of AVM value estimates, as lenders may order a traditional appraisal to verify the AVM estimate for loans that are near the AVM eligibility boundary or that present risk factors that warrant additional valuation scrutiny. The future of the appraisal profession is likely to involve a combination of traditional appraisal work, hybrid appraisal assignments, and AVM quality control and review functions, with the appraiser leveraging technology to enhance the efficiency and the quality of the valuation process.

The regulatory framework for automated valuation models is evolving as the use of the technology expands, with regulators seeking to ensure that the AVMs used in the mortgage lending process meet standards for accuracy, transparency, and fair lending compliance. The Federal Financial Institutions Examination Council has issued guidance on the use of AVMs in real estate lending, addressing the need for lenders to validate the accuracy of the AVMs they use, to understand the limitations of the models, and to maintain appropriate controls over the use of AVM value estimates in lending decisions. The fair lending implications of AVMs are a particular area of regulatory attention, as the models rely on historical transaction data that may reflect historical patterns of discrimination in housing markets, and the regulators have signaled that they will scrutinize the use of AVMs to ensure that they do not produce valuations that discriminate against properties in minority neighborhoods or that otherwise produce disparate outcomes across protected classes.

The consumer-facing AVMs that are available on platforms such as Zillow, Redfin, and Realtor.com, which provide instant property value estimates to anyone who enters a property address, have transformed the transparency of the property valuation landscape, giving consumers access to value estimates that were previously available only to real estate professionals and appraisers. The consumer-facing AVMs have been widely used by homeowners to track the value of their own property, by prospective buyers to assess the fairness of asking prices, and by sellers to set their listing prices, and they have become one of the most widely used and most influential PropTech tools. The accuracy of the consumer-facing AVMs has been the subject of significant analysis and debate, with the AVM operators publishing median error rates and providing confidence ranges around their estimates, and with critics noting that the error rates for individual properties can be substantial, particularly in markets or for property types where the AVM data is sparse or the market conditions are changing rapidly.

The future of property valuation technology is likely to involve the continued improvement of AVM accuracy through the incorporation of additional data sources, the advancement of the machine learning models, and the integration of new types of property data that are not currently captured in the standard AVM data infrastructure. The use of computer vision to analyze property photos and to identify and assess property features, conditions, and quality is an emerging area that could significantly enhance the AVM ability to account for the condition and quality factors that are currently assessed only by a human appraiser or inspector. The use of internet-of-things sensors and smart home data to capture real-time information about property condition and usage is another emerging area that could provide the AVM with more current and more granular property data. The integration of geospatial data, including satellite imagery, topographic data, and environmental hazard data, could enhance the AVM ability to assess the impact of location-specific factors on property value.

The implications of AVM advancement for the mortgage lending process are significant, as more accurate and more reliable AVMs could support the further expansion of appraisal waiver programs, reducing the cost and the cycle time of the mortgage origination process for a larger share of transactions. The reduction in appraisal costs and cycle times is a benefit for borrowers, who pay less in appraisal fees and experience a faster loan process, and for lenders, who reduce the cost and the time of origination. The expansion of AVM-based valuation also has implications for the appraisal profession, as the volume of traditional full appraisals declines and the nature of appraisal work shifts toward complex property valuation, hybrid appraisal review, and AVM quality control. The industry will need to manage this transition in a way that maintains the quality and the integrity of the valuation process, that ensures that the AVMs used are accurate and fair, and that preserves the professional infrastructure that is needed to handle the property valuations that AVMs cannot reliably perform.

The training and professional development of appraisers in a technology-enabled valuation environment is an important consideration for the future of the appraisal profession, as the skills and the knowledge that appraisers need are evolving alongside the technology that is reshaping the valuation process. Appraisers who wish to remain competitive in the evolving market are increasingly expected to be proficient not only in the traditional appraisal methods but also in the use of AVM tools, the interpretation of data analytics, and the integration of technology into the appraisal workflow. The Appraisal Institute and other professional organizations are developing continuing education programs and certifications that address the technology dimensions of the appraisal profession, equipping appraisers with the skills needed to work effectively with hybrid appraisal platforms, to perform AVM review and quality control functions, and to leverage data and analytics tools in their valuation practice. The professional development of the appraisal workforce is essential to ensuring that the industry has the human capital needed to perform the complex and high-stakes valuations that AVMs cannot handle, and to maintaining the quality and the integrity of the valuation process as the technology continues to advance and to expand its role in the lending transaction.

In conclusion, automated valuation models have become an integral and indispensable component of the property valuation landscape, providing instant value estimates that support consumer decision-making, lender portfolio management, and the streamlining of the mortgage origination process. The accuracy and the reliability of AVMs have improved significantly as the data and the models have matured, and the use of AVMs in the lending process is expected to continue to expand as the technology continues to advance. The AVM is not, however, a complete replacement for the human appraiser, as the models have inherent limitations that make them unsuitable for certain properties, certain markets, and certain transaction types. The future of property valuation is likely to be a hybrid model in which AVMs handle the routine valuations that they can perform reliably, and in which human appraisers focus on the complex, unique, and high-stakes valuations that require professional judgment, property-specific assessment, and the qualitative analysis that the models cannot provide. Understanding the capabilities, the limitations, and the appropriate use of AVMs is essential for anyone involved in the real estate transaction, from the homeowner tracking their property value to the lender underwriting a mortgage loan. This article is for educational purposes only and does not constitute professional advice. Readers should consult qualified real estate and appraisal professionals for guidance specific to their individual situation.

Ebonie Beaco — Mortgage Strategist

Have Questions About Your Mortgage Options?

Speak with Ebonie Beaco, Senior Mortgage Loan Officer (NMLS #2389954).

Mortgage Financing & Mortgage Education

Explore Mortgage Financing Options

Visit MortgageStrategist.net for professional services and resources.

Visit Site

This is a promotional link to an affiliated business of the publisher, separate from the magazine's editorial content.

automated valuation modelsavmproperty appraisalvaluation technologyhybrid appraisalproptech valuation
Share

Editorial Disclosure: This article is published by Mortgage & Real Estate Strategist Magazine for informational and educational purposes. It does not constitute legal, tax, financial, or investment advice. Mortgage and real estate professionals should consult appropriate experts and verify current program guidelines before making decisions based on this content.

Ebonie Beaco, Publisher & Mortgage Strategist
Ebonie Beaco

Publisher & Mortgage Strategist

Ebonie Beaco is the founder and publisher of Mortgage & Real Estate Strategist Magazine. She is a licensed mortgage loan officer (NMLS #2389954) affiliated with NEXA Lending (NMLS #1660690), providing lending services in multiple states. She covers mortgage strategy, lending programs, and real estate investing.

Contributing AuthorProfessional Website