Veros launches VeroVALUE Elite: computer vision that reads a home's condition into the valuation

Veros Real Estate Solutions announced today the launch of VeroVALUE Elite, a next-generation automated valuation model (AVM) that folds computer vision and market-based evidence directly into the valuation process. The Santa Ana, California company says the model turns property condition — long the softest input in automated valuation — from an assumption into a measurable signal, and claims a 3%+ improvement in nationwide P10 accuracy over its existing VeroVALUE AVM.

All performance figures below come from Veros's own launch announcement, released today via Business Wire and on the company's site. They have not been independently verified — but the claims are unusually specific for an AVM launch, and the feature set is concrete.

The condition problem#

Traditional AVMs lean on property characteristics, comparable sales, and other structured data. Property condition has been the stubborn gap: two homes with identical square footage and identical comps can be worth very different amounts if one was just renovated and the other is falling apart. Yet measuring condition consistently at scale has been historically difficult — every valuation either assumed a home was in "fair and marketable" condition or relied on an appraiser's visit.

"Property condition is a significant element in understanding the difference between what a home is worth and what traditional models historically have had to assume was always fair and marketable condition," said Darius Bozorgi, President and Chief Executive Officer of Veros. "By combining our expertise in housing finance, advanced modeling, and now proprietary computer vision technology, we can now account for property specific conditions at scale in a way that was not previously possible."

How VeroVALUE Elite works#

At the core is Veros's proprietary VeroVISION technology, which analyzes available property imagery and combines those insights with sales and market-based evidence. The output feeds into the Veros Home Score, a proprietary condition measure that captures how a property's condition influences its value. Where data permits, the model also generates a condition adjustment — an estimate of the premium or discount a property's condition commands relative to its local market.

The new model pairs this with enhanced valuation models, condition-adjustment intelligence, and daily data refreshes — valuations recomputed every 24 hours as new data and condition intelligence arrive, replacing the traditional weekly valuation cadence.

The numbers, per Veros#

Testing against the already industry-leading VeroVALUE AVM, per the company's announcement, demonstrated:

• More than a 3% improvement in nationwide P10 accuracy for VeroVALUE Elite.
• Over 5% improvement in P10 accuracy for properties in poor condition — the cases where the old "assume fair" default was doing the most damage.
• Broader coverage in thin markets: in rural and other thin or more heterogeneous markets, the model confidently values more homes, and values them more accurately.

The technology operates at significant scale: condition adjustments are available across approximately 100 million properties, with condition-score coverage for 90% of properties sold.

"The important part of this work is not simply identifying whether a property looks better or worse than another property, but quantifying how that difference can influence value," said Eric Fox, Chief Economist and Executive Vice President of Analytics at Veros. "Our testing demonstrates that incorporating VeroVISION condition intelligence into the valuation model can produce measurable improvements in accuracy, particularly for properties where condition is a significant differentiator."

Availability#

VeroVALUE Elite is configurable to client needs. It ships as a standalone, core condition-adjusted AVM report, or as an enhanced module that adds condition ratings for the subject property and its comparables, condition adjustments, supporting photos, and area-level condition detail.

Delivery channels include Offline Match & Append, the VeroSELECT UI and API, the Snowflake Data Marketplace via private data sharing, and Valligent via the Acuity platform and its integrations. Existing VeroVALUE API users can receive the new model through the existing schema and report — no new integration work required.

Why this matters#

Veros is not a startup chasing a trend. Operating since 2001, the company is the primary architect and technology provider of the GSEs' Uniform Collateral Data Portal (UCDP) — the system Fannie Mae and Freddie Mac use for appraisal delivery — and supports the FHA's Electronic Appraisal Delivery portal and the Department of Veterans Affairs' appraisal streamlining. In other words, the company already sits at the plumbing layer of U.S. mortgage collateral valuation.

That context makes this launch more than a feature release. If computer vision can turn condition from a default assumption into a measured, continuously refreshed input — at 100-million-property scale — the margin between automated and human appraisals narrows further, especially for the thin markets and distressed properties where manual judgment has been hardest to replace. The usual vendor caveat applies: these are Veros's numbers, independently unverified. But the direction is clear — valuation AI is moving from structured data into the pixels.