Veros Real Estate Solutions has launched VeroVALUE Elite, an automated valuation model that incorporates computer vision-based property condition intelligence into home valuations. The new AVM combines imagery, market evidence, enhanced modeling and daily data updates to address one of the longstanding limitations of automated property valuation: measuring a home’s physical condition consistently at scale.

Automated valuation models have become an important part of mortgage lending, servicing and real estate analytics, but one variable has remained difficult to quantify consistently: property condition.

Veros Real Estate Solutions is attempting to address that gap with VeroVALUE Elite, a new automated valuation model (AVM) that uses computer vision and market-based evidence to incorporate a property’s observed condition into its estimated value.

The product builds on Veros’ existing VeroVALUE AVM and adds its proprietary VeroVISION technology, which analyzes available property imagery and converts visual information into a measurable condition signal. That information is combined with sales data and other market evidence before being incorporated into the valuation model.

The significance is straightforward: traditional AVMs can process large volumes of structured information such as property characteristics and comparable sales, but physical condition is harder to represent in a standardized numerical format.

A home with an updated kitchen, renovated bathrooms and well-maintained exterior can command a different price from a broadly comparable property requiring significant repairs. If both properties look similar in a conventional property-data record, an AVM may have limited information with which to distinguish their values.

VeroVALUE Elite introduces a condition layer designed to address that problem. VeroVISION identifies rooms and property features from available imagery, scores observable condition and aggregates those assessments into the Veros Home Score, a 1.0-to-5.0 measure. The resulting information can then contribute directly to the AVM and, where sufficient evidence is available, generate a condition adjustment representing an estimated premium or discount relative to the local market.

Veros says testing against its existing VeroVALUE model produced more than a 3% improvement in nationwide P10 accuracy, with more than a 5% improvement for properties in poor condition. The company also says the model can value more properties in rural and other thin or heterogeneous markets. Those figures are company-reported testing results rather than an independent industry benchmark.

The scale of the data layer is another important component. Veros says condition adjustments are available across approximately 100 million properties, while condition scores cover roughly 90% of properties sold. Its current product documentation reports more than 100 million properties with condition adjustments and says condition information can be supplemented through market-based inference when photographs are unavailable.

That ability to work with incomplete imagery is significant because computer vision cannot simply assume that every property will have a current, comprehensive photographic record. VeroVALUE Elite can use market-based evidence where imagery is unavailable and update its assessment when new images become available.

The timing also reflects a changing regulatory environment for automated valuation. In 2024, federal banking and housing regulators finalized quality-control standards for AVMs used in certain mortgage-related credit and securitization decisions. The rule requires institutions to maintain controls intended to promote confidence in AVM estimates, guard against data manipulation, address conflicts of interest, conduct random-sample testing and comply with applicable nondiscrimination requirements. The rule took effect October 1, 2025.

That regulatory framework makes transparency and model governance increasingly important as lenders use automated property valuations for more decisions. An AVM that can expose additional evidence behind an estimate—including property condition information—has a different operating profile from a model that produces only a single value.

The broader mortgage industry already operates on large standardized datasets. The Federal Housing Finance Agency’s Uniform Appraisal Dataset, for example, provides aggregate statistics derived from millions of appraisal records, covering property characteristics, geography and valuation information. FHFA says the public dataset is intended to improve transparency around home valuation.

Computer vision adds another type of information to that ecosystem: visual evidence.

For lenders and mortgage investors, more condition-sensitive valuation could potentially improve collateral risk assessment, particularly when properties differ materially from local comparables. Servicers and investors can also use updated valuations for portfolio monitoring, while insurers, brokerages and PropTech platforms may have their own use cases for property-condition intelligence.

The 24-hour refresh cycle is another departure from slower valuation workflows. Veros says VeroVALUE Elite updates valuations daily as new market data and condition intelligence become available. That creates a more current valuation layer for applications where property values need to be monitored repeatedly rather than assessed only at origination.

The product is also being distributed through several data and technology channels, including VeroSELECT UI/API, Snowflake Data Marketplace private sharing, Valligent’s Acuity platform and existing VeroVALUE API infrastructure. That API and marketplace availability is important because the technology can become a component inside larger mortgage, lending, servicing and real estate workflows rather than remaining a standalone valuation report.

The development places Veros within a broader PropTech and fintech trend toward using machine learning, computer vision and alternative data to automate traditionally manual real estate processes. The challenge is not simply generating an automated estimate; it is producing an estimate that can incorporate more of the information a human appraiser would consider.

That distinction will matter as lenders increasingly combine automated valuations with appraisal and collateral-risk workflows. Federal rules do not eliminate the use of AVMs, but they raise the importance of quality control, testing and governance around them.

For financial technology, mortgage technology and AI-powered risk management, VeroVALUE Elite therefore represents a shift from data-heavy AVMs toward multimodal valuation models that combine structured market information with visual evidence.

The broader question is whether property imagery can consistently improve valuation accuracy across different housing markets without introducing new forms of model bias or uncertainty. Veros’ initial testing provides company-reported evidence of improvement, but independent validation across diverse markets and use cases will remain important as condition-aware AVMs become more widely adopted.

Market Landscape

Automated property valuation is evolving from models based primarily on structured property and transaction data toward systems that can incorporate computer vision, image-derived features and continuously refreshed market information.

The regulatory backdrop makes model transparency and quality control increasingly important. Federal regulators’ AVM rule requires covered institutions to establish controls around confidence, data integrity, testing, conflicts of interest and nondiscrimination.

Meanwhile, the scale of available housing data is expanding. FHFA’s UAD aggregate statistics provide millions of appraisal-derived observations, while mortgage datasets such as Freddie Mac’s Single-Family Loan-Level Dataset cover approximately 56 million mortgages originated between 1999 and March 2026.

Within that environment, the competitive differentiation for AVM providers is increasingly likely to involve not just coverage and model accuracy, but the breadth of evidence incorporated into valuations, update frequency, explainability and integration into mortgage technology workflows.

Top Insights

• VeroVALUE Elite adds computer vision-derived property condition intelligence to an established automated valuation model.
• Veros reports more than 3% improvement in nationwide P10 accuracy and over 5% improvement for properties in poor condition.
• The model combines imagery with sales and market evidence to generate a condition score and potential valuation adjustment.
• Federal AVM quality-control rules make testing, data integrity, transparency and nondiscrimination increasingly important for mortgage valuation technology.
• Daily valuation refreshes and API-based distribution position condition-aware AVM data for use across lending, servicing and real estate technology workflows.

Get in touch with our fintech expert