A lower property valuation fee is easy to compare. However, the risk behind a property value takes more work to understand. That distinction runs through recent conversations with the Veros Real Estate Solutions (Veros) and Valligent Sales team. In late October 2026 interviews, Sarah Bell, Alex Vargas and Mike Courtney described clients looking for ways to control expenses, serve existing borrowers and make valuation decisions with fewer manual steps. They also described housing finance professionals’ reluctance to change a process that already works for them, even if it is not ideal and can use improvement.
For mortgage lenders evaluating automated valuation models (AVMs), the practical question is how to determine whether a model’s results are dependable for the properties and decisions in front of them. Accuracy testing, confidence scores, condition information and exception handling each answer a different part of that question.
Cost and risk belong in the same conversation
“I don’t think they’ve changed. I think they’ve intensified what they’ve already had,” said Sarah Bell, Sales Director for Veros and Valligent, when asked whether client priorities had shifted over the previous six months.
Bell described a focus on keeping expenses down while understanding collateral risk, both for new originations and loans already on the books. Her observation offers a useful starting point for AVM evaluation: define the decision before comparing the fee.
A model used to screen a portfolio for further review serves a different purpose from one used in an individual lending decision. The cost of an inaccurate result, and the opportunity to catch it, can differ too.
Before testing providers, write down the intended use, the properties in scope and what happens when the result falls outside policy. Otherwise, a low price per valuation can conceal a process that sends too many files into manual review or leaves important exceptions unexplained.
Test the properties you actually need to value
Bell said clients are asking what distinguishes one AVM from another and what makes a model more or less risky. Those questions deserve evidence that resembles the lender’s business.
Start with a representative sample of the properties you expect to value. Then examine results by geography, property type and other segments that matter to the intended use. Review both accuracy and the share of properties receiving a usable result under your acceptance criteria. A strong overall result can still leave a weak spot in a market where you lend heavily.
The same care applies to the benchmark. Veros’ discussion of the MISMO AVM testing framework explains the importance of benchmarks that reflect actual use, including appraisals and closed sales prices, and recent valuation estimates with controls against benchmark data leakage, which occurs when information about the benchmark influences the valuation being tested.
For a lender, the useful follow-up is simple: What does this test tell us about the decisions we will make, and which decisions does it leave untested?
Find out what the confidence score means
An AVM confidence score can help determine when a valuation needs more attention. It only helps if the team understands its meaning. Proprietary scores from different AVM providers may use different scales and methods. Matching numbers do not necessarily represent matching reliability. The MISMO Common Confidence Score, or CCS, provides a standardized measure tied to P10 performance, the share of valuations within plus or minus 10% of the benchmark. Users still need to validate that a provider’s scores correspond to observed performance.
Ask the AVM provider to show what happens across score ranges. As confidence falls, how does measured accuracy change? How much usable coverage remains at the threshold you propose?
That turns a score into a workflow decision. Your team can establish when to accept a result under policy, seek more information or escalate to another valuation method. A score without a defined next step leaves the reviewer to make that decision from scratch.
Ask what the model knows about condition
Two homes can have similar square footage, age and location while differing substantially in upkeep or renovation. Understanding how a model handles those differences belongs in the evaluation.
Bell reported recent market interest in VeroVALUE Elite, including conversations about what sets AVMs apart. VeroVALUE Elite incorporates property condition intelligence using available imagery and sales-based evidence, alongside enhanced modeling and daily updates. Many industry professionals are excited about this new AVM.
The useful questions go beyond whether a condition score exists. What evidence supports it? How recent is that evidence? What happens when recent photos are unavailable?
A condition assessment based on available data is not confirmation of what the property looks like today. If the lending decision depends on current physical condition, determine whether an inspection or other additional evidence is needed. Knowing the limits of the information is part of using it well.
Turn the test results into a decision rule
“People are interested, they want to learn, they want to hear things, but they’re really reluctant to make that change,” said Alex Vargas, Sales Director for Veros and Valligent.
A useful AVM evaluation should give the team clear acceptance criteria and an escalation path. Identify which results meet policy, which require more property information and which call for another valuation method. Include the reviewers who will handle those exceptions when setting the criteria.
Before adopting a model, ask the provider to demonstrate those decisions on a representative set of properties. Your team should be able to explain why it accepted a result, what would cause it to question that result and what it would do next.
For a broader look at what lenders recently discussed in Chicago, read our MBA Annual 2026 takeaways.
How does your AVM perform on the properties you lend on?
Talk with the Veros team about evaluating AVM accuracy, confidence scores and property condition intelligence for your lending needs.
